U.S. flag

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

  • Publications
  • Account settings
  • Advanced Search
  • Journal List
  • Comput Intell Neurosci
  • v.2022; 2022

Logo of cin

This article has been retracted.

Internet of medical things (iomt)-based smart healthcare system: trends and progress, jyoti srivastava.

1 Department of Computer Science and Engineering, School of Engineering, Indrashil University, Rajpur, Mehsana, Gujarat, India

Sidheswar Routray

Sultan ahmad.

2 Department of Computer Science, College of Computer Engineering and Sciences, Prince Sattam Bin Abdulaziz University, Alkharj 11942, Saudi Arabia

Mohammad Maqbool Waris

3 Department of Mechanical Engineering, Adama Science and Technology University, Adama, Ethiopia

Associated Data

The data used to support the findings of this study are available from the first author upon request ( moc.liamg@886avatsavirsitoyj ).

Internet of Medical Thing (IoMT) is the most emerging era of the Internet of Thing (IoT), which is exponentially gaining researchers' attention with every passing day because of its wide applicability in Smart Healthcare systems (SHS). Because of the current pandemic situation, it is highly risky for an individual to visit the doctor for every small problem. Hence, using IoMT devices, we can easily monitor our day-to-day health records, and thereby initial precautions can be taken on our own. IoMT is playing a crucial role within the healthcare industry to increase the accuracy, reliability, and productivity of electronic devices. This research work provides an overview of IoMT with emphasis on various enabling techniques used in smart healthcare systems (SHS), such as radio frequency identification (RFID), artificial intelligence (AI), and blockchain. We are providing a comparative analysis of various IoMT architectures proposed by several researchers. Also, we have defined various health domains of IoMT, including the analysis of different sensors with their application environment, merits, and demerits. In addition, we have figured out key protocol design challenges, which are to be considered during the implementation of an IoMT network-based smart healthcare system. Considering these challenges, we prepared a comparative study for different data collection techniques that can be used to maintain the accuracy of collected data. In addition, this research work also provides a comprehensive study for maintaining the energy efficiency of an AI-based IoMT framework based on various parameters, such as the amount of energy consumed, packet delivery ratio, battery lifetime, quality of service, power drain, network throughput, delay, and transmission rate. Finally, we have provided different correlation equations for finding the accuracy and efficiency within the IoMT-based healthcare system using artificial intelligence. We have compared different data collection algorithms graphically based on their accuracy and error rate. Similarly, different energy efficiency algorithms are also graphically compared based on their energy consumption and packet loss percentage. We have analyzed our references used in this study, which are graphically represented based on their distribution of publication year and publication avenue.

1. Introduction

The Internet of Things (IoT) deals with various interconnected computing devices, machines, objects, humans, or animals with unique IDs and is capable of transferring data within the network without human intervention [ 1 ]. It includes monitoring and controlling systems that enable smart homes, for example, thermostats, heating, ventilation, and air conditioning devices, including IoT. IoT can also be used in other domains like transportation, healthcare, industrial automation, and energy response to natural and man-made disasters. Various IoT applications in different domains are illustrated in Figure 1 . Verma et al. [ 2 ] proposed a data congestion monitoring system having a sharp area structure, where IoT helps in convincing the control of the leading body in traffic area via advanced systems. Fuqaha et al. [ 3 ] presented the use of IoT for checking environmental conditions with the help of disappointment figures, sullying control, and alarm trigger under crisis.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.001.jpg

Overall view emphasizing the role of IoT in different domains.

The applications of IoT in healthcare are the most demanding areas of research as per the current scenario. The Internet of Medical Thing (IoMT) is playing a crucial role within the healthcare industry to increase the precision, consistency, and throughput of the electronic devices as presented by Joyia et al. [ 4 ]. Because of the current pandemic situation, it is highly risky for an individual to visit the doctor for every small problem. Hence, using IoMT devices, we can easily monitor our day-to-day health records, and thereby initial precautions can be taken on our own.

An IoMT-based smart healthcare system is a collection of various smart medical devices connected within the network through the internet. An IoMT framework-based smart healthcare is formed of various phases. Firstly, medical data will be collected from the patient's body using smart sensors integrated within smart wearable or implanted devices that are connected together via a body sensor network (BSN) [ 5 ] or wireless sensor network (WSN) [ 6 ]. Then, this data will be transferred over the internet to the next component dealing with the prediction and analysis phase. After receiving the medical data, analysis can be done using a proper AI-based data transformation and interpretation technique [ 7 ]. In case of serious problems, doctors or other medical requirements can be approached with the help of smart AI-based applications in smartphones [ 8 ]. In nonserious cases, self-preventive measures can be taken.

AI provides the capability to a computer or robot, which is controlled by a computer system for performing tasks that are usually done by humans via their intelligence. Within a smart healthcare system with proper data interpretation techniques, a machine can also monitor health parameters using the implanted/wearable sensors on the body of the person under observation. Real-time disease management and prevention with improved user-end experience can be achieved using AI. SHS deals with very sensitive medical data of the person under observation. Hence, providing essential security measures in IoMT-based SHS is a very crucial task. AI can also be used for providing security in IoMT by detecting network intrusion [ 9 ] and intermediate security attacks within the IoMT systems [ 10 ], performing web-based security assessment using an IoMT-SAF device [ 11 ], etc. In an emergency situation, an automatic alert can be given to different parties using AI, which will help in saving a life by taking immediate actions [ 12 ]. Hence, doctors can easily manage patients' records and can also provide off-time medical services using AI. Blockchain can also be used for providing security in an IoMT network. It is a distributed database that maintains secure and decentralized information electronically in a digital format, which will guarantee the security and fidelity of data. Hence, it generates trust without the involvement of a third party. Blockchain can be used in IoMT for providing security in medical servers with electronic health records like MedRec that can be used for permission and access control management of medical data [ 13 ].

Various advancements in the smart home technology provide a healthy life and enhanced healthcare quality, especially for the handicapped and elderly personalities, and these advancements provide a comfortable lifestyle for patients in homecare, thus avoiding their admittance to hospitals, nursing facilities, or other confinement facilities [ 14 , 15 ]. SHS will improve healthcare facilities for humans from various locations outside the hospital [ 16 ], thereby reducing depression, stress, and loneliness inside hospital wards. Doctors can also monitor and diagnose patients' health parameters and provide medicine prescriptions accordingly from any location [ 17 ]. Also, the exponential improvement of various new software and hardware technologies in SHS helps people, especially the disabled ones, to easily access certain home appliances using various smart devices, such as smartphones, laptops, tablets, etc. SHS is made of various computing devices that act proactively on behalf of persistent users [ 18 ]. Hence, for making good decisions in SHS, we require essential features, for example, users' preferences need to be considered for finding their choice of interest in certain scenarios [ 19 – 24 ]. Here, user preferences deal with the information used for describing the situation of a person considering the physical medical status or requirements. In modern SHS, we measure and record specific health parameters, such as blood pressure (BP), body temperature, pulse, glucose level, etc. We can also send a reminder in SHS to patients for medications based on some prior input provided by the user.

Therefore, we are motivated to do a comparative analysis of various research challenges faced by different researchers while developing an IoMT-based SHS. Considering the sensitivity of the real-time medical environment, we are encouraged to work in the direction of an artificial intelligence-based smart healthcare system using the IoMT framework.

The major contributions of the research work are as follows:

  • To analyze various IoMT architectures used in AI-based smart healthcare system
  • To present a comparative analysis of various data collection techniques to improve the accuracy of collected medical data
  • To present a comprehensive analysis of various energy-efficient techniques to optimize energy consumption by IoMT devices in SHS
  • To explore various health domains of the IoMT framework along with their application in the smart healthcare system, including the types of sensors used for each domain
  • To propose various research challenges that need to be considered while creating an IoMT-based smart healthcare system

The rest of the paper is organized as follows: Section 2 represents the distribution of referenced papers based on the publication year and avenue, graphically. Section 3 discusses various existing IoMT architectures used by various authors for SHS, gives a comparative analysis of various data collection techniques used in smart healthcare systems, and provides a comparison of different energy-efficient algorithms using various parameters. Section 4 defines the health domain with its application in SHS. Section 5 describes the findings of literature survey in terms of the concerned challenges during the design of an IoMT network. Section 6 is the conclusion of the research work done through this paper.

2. Statistical Distribution of Publications Referred

Figure 2 compares the referenced papers according to the publication venue. Figure 2 highlights the distribution of referenced papers based on the type of journal. 67 papers from the total referenced papers are primary research papers from reputed journals, whereas 31 papers originate from conferences.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.002.jpg

Distribution by the type of papers.

Figure 3 represents the frequency of papers concerning the smart healthcare system. We have reviewed only 7 papers that are published on or before the year 2014 because of our interest in recent technologies in this era. We have referenced 4 papers from the year 2015, whereas the year 2016 contains 10 research papers on energy efficiency, security, and accuracy of healthcare data. We expect modern technology-based quality paper growth in the IoMT system from 2017, with the papers becoming freely available from various reputed journals for reference. As it can be seen, we have referred to most of the recent papers to get updated regarding the current tools and technologies in this era.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.003.jpg

Distribution of reference papers by the year of publication.

3. Related Work

This section gives the comparative and comprehensive analysis of work done by various authors in IoMT-based Smart Healthcare systems regarding different IoMT architectures, data collection techniques, their comparative analysis, and a comparison of various energy-efficient algorithms.

3.1. IoMT Architectures

An IoMT-based smart healthcare system is a collection of various smart medical devices connected within the network through the internet [ 25 ]. An IoMT framework-based smart healthcare is formed of various phases. Firstly, medical data will be collected from the patient's body using smart sensors integrated within the smart wearable or implanted devices that are connected together via BSN or WSN [ 26 ]. Then, this data will be transferred over the internet to the next component dealing with the prediction and analysis phase. After receiving the medical data, analysis can be done using a proper AI-based data transformation and interpretation technique [ 27 ]. In case of serious problems, doctors or other medical requirements can be approached with the help of smart AI-based applications in smartphones [ 28 ]. In nonserious cases, self-preventive measures can be taken.

Sun et al. [ 29 ] explained that IoMT architecture mainly consists of 3 layers, which are as follows: the application layer, perceptual layer, and network layer. They are demonstrated in Figure 4 . The bottom layer, i.e., the perceptual layer, deals with the collection of data from the source and making important viewpoints from the collected data. Now, the perception layer consists of 2 sublayers, i.e., the data access sublayer and data acquisition sublayer. Perception from the collected data is the main task done by the data acquisition sublayer, for which it utilizes various medical perception equipment and signals acquisition equipment. Graphic code, RFID, GPRS, etc., can be considered the major signal acquisition methods. The data access sublayer connects the collected data from the data acquisition layer to the network layer through short-range data transfer techniques, such as Bluetooth, Wireless Fidelity (Wi-Fi), ZigBee, etc.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.004.jpg

IoMT architecture overview.

The middle layer, i.e., the network layer, deals with providing various platform and interface-related services and provides various data transmission techniques. This layer is formed of 2 subsequent layers, namely, the service layer and the network transmission layer. The network transmission sublayer uses mobile communication networks, wireless sensor networks, internet, etc., for transmitting the data received from the perception layer in a precise, consistent, real-time, and barrier-free way. However, the service layer realizes the integration of various networks, information description formats, data warehouses, etc. For such integrations, it provides open interface services and various other platform-related services.

The application layer utilizes the information gathered from the network layer to manage the medical record by means of various applications. This layer again consists of 2 sublayers, namely, the medical information decision-making application layer and the medical information application layer. The medical information application layer contains various health care equipment and other materials related to information for maintaining patient information, such as inpatient, outpatient, medical treatment, etc., records, whereas the medical information decision-making application layer deals with the analysis of various pieces of information, such as patients, disease, medication, diagnosis, treatment, etc.

Sun et al. [ 1 ] explained another three-tier architecture with medical server level, sensor level, and personal server level. Sensor level contains various sensors and medical devices in the form of a local network like a body sensor network (BSN) using low power wireless technology (such as BLE, NFC, or RFID) to transfer data.

The personal server level has few personal servers that can internally process and store from smart wearable devices (like a smartwatch) or off-body devices (like routers). It is required in situations where either a network connection is lost or the user needs the patient's data remotely. The last layer is the medical server layer, which consists of an algorithm or program for early diagnosis, rehabilitation progress assessment, or continuous patient monitoring (for example, MobiCare and BSN-Care [ 30 ]). The problem stated here is security negligence.

Kumar et al. [ 31 ] proposed an end-to-end architecture named mHealth System that connects the IoT smart sensors directly with SHS. This architecture contains three layers, i.e., the data processing layer, data collection layer, and data storage layer. The bottom layer, i.e., the data collection layer, consists of IoT devices that can sense and collect medical parameters. The next layer, i.e., the data storage layer, stores medical data on wide-scale and high-speed storage racks. The topmost layer, i.e., the data processing layer, involves various techniques to analyze collected sensor data.

Abdulmohsin Hammood et al. [ 32 ] proposed the four-tier architecture of an Internet of Medical Thing health-based model, where the first tier is the WBSN tier in which sensors like ECG (Electrocardiography) are directly connected to the human body. Fetched data from these sensors are transferred to the coordinator node via wireless 802.15.6 standard, which is then transmitted to the next tier. Tier 2 is the Smart\Wireless technology interface tier, where smart devices are utilized for data inspection and analysis and then transfer this data to tier 3 either by smart devices or wireless communication technologies. Tier 3 is the infrastructure internet tier that provides various communication technologies. Tier 4 is the care-services tier, where the received data are forwarded to the intelligent server (IS), where the data are stored, analyzed, and forwarded for smart medical services.

Here, we have seen 4 architectures for an IoMT-based smart healthcare system, where most of them have three layers. The last architecture alone, proposed by Abdulmohsin Hammood et al. [ 32 ], has a four-tier architecture. Upon comparing all these architectures, we can generalize that the bottom-most layer will have sensors in direct contact with the human body. In the middle of the architecture, we need a few layers for the inception, storage, and processing of data. The topmost layer will be used for providing services to the end-users.

3.2. Technologies Used for the Collection of Sensor-Based Medical Data

IoMT-based SHS uses various techniques to collect and transfer sensor data to servers, such as BSN, WSN, or RFID [ 33 ]. BSN is an IOT-based technology in a healthcare system that deals with monitoring the health of patients using a collection of various wireless sensor nodes with low-weight and low-power consumption [ 34 , 35 ]. BSN-based social insurance systems can be used for therapeutic administration systems to accomplish various security essentials [ 36 ]. Since BSN nodes collect sensitive information and may operate in a heterogeneous environment, they require strict security mechanisms like BSN-Care [ 30 ].

RFID is a contactless technique for the automatic identification of targets using radiofrequency with 2-way data communication in various zones identified by their unique names [ 37 , 38 ]. RFID consists of 3 parts, namely, the reader, database management system, and radio frequency electronic tag [ 39 ]. It can be used for identifying locations, the management of medical equipment and assets, waste tracking, personal identification, and for the collection of vital sign data of patients, such as ECG and blood pressure data [ 40 , 41 ]. The advantage of using RFID is that without any human intervention, it can recognize objects at long distances with strong anti-interference. Flexible RFID tags can be used to expand their reading range [ 31 ]. A low-cost inkjet-printed RFID tag antenna can be used in remote healthcare applications [ 42 ]. We can also work upon the middleware providing an interface between the reader-writer and backend application. It will capture data from the sensing device and conduct proofreading, filtering, processing, and transferring them to RFID [ 43 ]. It will make healthcare more affordable and convenient to use.

Wireless sensor network (WSN) is a network of different monitoring sensors located in a homogeneous or heterogeneous environment. WSN can be used in IoMT for monitoring the real-time physiological condition of the person under observation [ 44 ]. Also, there are sensors that can measure the pressure level by examining the body's perspiration, speed of movement, and temperature of the patient's body [ 45 ]. Dhunna et al. [ 46 ] proposed a smart grid monitoring application for providing security in WSN with very low energy consumption. Yadav et al. [ 47 ] proposed a clustering algorithm for minimizing the energy consumption within the WSN network.

3.3. Comparison of Data Collection Techniques

A smart healthcare system will work precisely only when it will get correct and accurate data [ 48 ]. Hence, this section elaborates on a comparative analysis of different smart healthcare data collection techniques to maintain the accuracy of collected medical data [ 49 ]. In Table 1 , we have compared different research works done on the techniques that can be used for collecting sensitive medical data using parameters, such as accuracy, error rate, and correlation prediction. Tekieh et al. [ 50 ] used a survey to demonstrate the uses of data mining in the healthcare system. The main problem is to maintain the quality and security of a large amount of health-related medical data, which is progressively increasing every day [ 55 ]. To overcome the problem, they have discussed 3 data mining processes in brief, i.e., association, clustering, and classification. They have discussed 4 applications of these techniques of data mining in SHS, i.e., the health of a population, health administration and policies, biomedicines and genetics, and clinical decision-taking [ 56 ].

Comparison of different data collection techniques.

Shahin et al. [ 51 ] proposed an advanced reduction technique named dynamic rough sets attribute reduction (DRSAR) with multiple classifiers for a random forest (RF) in the healthcare information systems (HIS). This model will be helpful to overcome the most critical challenges, i.e., for extracting relevant information from a large amount of medical data that needs to support the proper working of the system. The efficiency of the model is examined using 4 case studies (namely, premature birth, coronary heart disease, osteoporosis, and acute appendicitis). They have also provided web interfaces so that patients can calculate the level of risk involved with every medical case.

Yang et al. [ 52 ] proposed an association rule remining algorithm, multimode, and high-value association rule mining (MH-ARM) based on both the characteristics of data and the user's intention and knowledge as shown in Figure 5 . They have considered more metrics, such as Kulczynski (KULC) and imbalanced ratio (IR), for the measurement of the support-confidence framework. They have taken 2 threshold values, i.e., the minimum support and minimum confidence, and they can be adapted as per the need of the user.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.005.jpg

MH-ARM framework.

3.3.1. Multimode Association Rule Mining

Let A and B be the attributes that can be shared for every instance belonging to the same class or unshared specific attribute varying with every instance. Then, multimode association rule mining can be given by Figure 6 .

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.006.jpg

Multimode association rule mining.

Here, four parameters are used, namely, support, confidence, correlation, and novelty, to extend the support-confidence framework of association rules given in equation ( 1 ). KULC and IR will provide the support and confidence for the algorithm, whereas novelty is calculated in equation ( 4 ), and the overall weight for the association rule remining algorithm, multimode, and highvalue association rule mining (MH-ARM) is calculated in equation ( 5 ).

where sup represents the support of the algorithm and is calculated with the help of IR. Similarly, conf represent the confidence of the algorithm and is calculated with the help of KULC.

KULC: given 2 item sets X and Y , and KULC of X and Y can be calculated as follows:

From ( 2 ), we can say that KULC is subjective to the conditional probability of P ( X|Y ) and P ( Y|X ) and is independent of the number of records. The value of KULC will be from 0 to 1. The higher value of KULC signifies greater relevance between X and Y .

IR: the imbalanced ratio for X and Y can be calculated using

If both X and Y are in the same direction, then IR ( X , Y ) is 0. Otherwise, the increase in the difference of two directions gives a higher IR value. As it can be seen from equation ( 3 ), IR is independent of zero transaction and number of records.

Novelty: novelty rules neither inferred by others nor known to the users. Rule x ⇒ y will be treated as novel when P ( XY ) cannot be inferred by P ( X ) or P ( Y ). Novelty can be refined using

For the determination of weights, they have used analytic hierarchy process (AHP) using “support—confidence—correlation—novelty” as a comparison parameter, with the weights of C, S, K, and N being 0.482, 0.11, 0.19, and 0.218, respectively. Therefore, the complete evaluation coefficient obtained is given in

Electronic health record system helps in the storage and organization of data. Mdaghri et al. [ 53 ] used this technique with data mining approaches to gain accuracy during the extraction of information from huge raw data. They elaborate on the application of data mining in healthcare, especially for the following four categories: the health of a population, health administration and policies, biomedicines and genetics, and clinical decision making. Roy et al. [ 54 ] proposed a variation of the decision tree model using the correlation ratio (CR) concept for smart healthcare datasets with many attributes, and each attribute contains various values. They have applied this model to various healthcare datasets to prove that the correlation ratio-based approach is unbiased toward a number of attributes, thereby giving more accuracy to the result.

Suppose l tuples are available in a dataset and the number of times y ∈ Y (where Y is set of outcomes) occurs is l y , then the dataset partitioned by their outcomes is given by

where S y is the set of tuples with the outcome y , and j is the value for the i − th attribute of the j − th tuple among all the l y tuples with the outcome y . Equation ( 7 ) shows the average of the i − th attribute from all the tuples in each outcome class.

Equation ( 8 ) gives the overall average for the i − th attribute of all tuples.

The square of CR between the i − th attribute and outcome (class attribute) is given by

Now, this CR will be able to find nonlinear dependencies, which will reflect the biasness, and thereby improve the accuracy of the collected data, whereas in paper [ 34 ], a detailed analysis of the physician and hospital rating data was done using a toolkit based on open-source modules, which are the publicly available datasets of USA.

3.4. Comparison of Energy Efficiency Measurement Techniques

Energy efficiency determines the size, lifetime, and usability of IoMT-based medical devices used in SHS [ 57 , 58 ]. Implant devices should have a battery life minimum of 10 years to 15 years to avoid repetitive surgery as it results in physical and financial loss [ 59 ]. As far as wearable devices are considered, frequent battery changes reduce device usability [ 60 ]. Energy efficiency can be measured through various parameters, such as the amount of energy consumed, packet drop ratio, delivery time, data leakage, energy discharge, battery lifetime, packet loss, QoS (Quality of Service), power drain, network throughput, end-to-end delay, transmission rate, outage probability, internode distance, path-loss, and antenna gain. As shown in Table 2 , Rehman et. al. [ 7 ] have compared their energy-efficient IoT e-health model with the attribute-based encryption (ABE) and privacy-enhanced data fusion system (PDFS) model using parameters EC (energy consumption), PDR (packet drop ratio), DT (delivery time), and DL (data leakage). For reducing latency by minimizing the number of hops, they use the heuristic formula, h ( n )= d + t d , for each node, where t d is the delivery time and distance d is comprised of distance from source node i to corresponding neighbor n i , which is denoted by d ′. d _edge is the distance of the neighbor to the corresponding network edges edge i , and thereby, the mobility ratio for the network edge is given in

Comparison of recent energy efficiency measurement techniques.

Now, the calculation of t d includes delay time and data reception fluctuations, denoted by d recp . This model sets a threshold value to determine the strong s and weak ω transmission channel c , which is given as follows:

They proved through simulation results that the proposed model, when compared with ABE and PDFS, respectively, has improved the efficacy by 13% and 15% for data latency, 19% and 21% for packet drop ratio, 16% and 18% for energy consumption preround, 21% and 28% for delivery time, and 12% and 14% for data breaches.

Sodhro et al. [ 63 ] proposed an energy-efficient algorithm (EEA) that mainly focuses on data transmission and connectivity increase with a reduced interruption during information transfer. The authors compared the proposed algorithm with battery recovery-based lifetime enhancement (BRLE) using parameters, such as energy dissipation and charge dissipation. The discharge curve for the battery is defined by the voltage function, which includes the state of charge (SOC) with exponential decay. ( 12 ) and ( 13 ) show the discharge curve by SOC, which is equal to st , and S gives the remaining capacity/total capacity.

where F ( V ) denotes the voltage function of battery, S is the state of charge, t denotes the time duration for battery discharge, β is the parameter used for battery diffusion, t k is the time duration of task k , t f is the time for turning ON the load, and t i is the time for turning OFF the load. They have shown through MATLAB simulation that EEA dissipates 89.7 J of energy, while BRLE dissipates more energy up to 95.68 J, and the charge dissipation of EEA is only 16,657.1409 mC·mint, while that of BRLE is 18,742.6591 mC·mint.

Lazarevska et al. [ 65 ] proposed a routing protocol for low power and lossy networks (RPL) to provide energy efficiency while accounting for the mobility of sensor nodes in WSNs with both static and mobile nodes. The proposed model objective function considers 5 parameters: EC (energy consumption), PDR (packet delivery ratio), duty cycle, total control overhead, and network lifetime. For calculating network lifetime, they used the power tracker tool for online monitoring of real-time duty cycle providing average simulated radio duty cycles of the transmission (Tx) and reception (Rx) of data for each node in (%) using

Using ( 16 ), the energy consumption of every single node and of whole network can be estimated.

where E is energy, P is power, V is voltage, I is current, and t is the total time spent in a state. From equations ( 14 ), ( 15 ), and ( 16 ), we can reach

Here, the predefined values for voltage, transmission, and reception current are 3 Volts, 8.5 mA, and 19.7 mA, respectively. The total energy consumption is given by the sum of the independent energy consumption of Tx , Rx , CPU (central processing unit) and LPM (low power CPU model). Now, E _cpu and E _lpm are relatively very small, and hence, they can be neglected easily for the final formula of the total average energy consumption as

Tanzila et al. [ 66 ] proposed a secure and energy-efficient e-healthcare (SEF-IoMT) framework using the Internet of Medical Things (IoMT) and compared it with a simplified energy-balanced alternative-aware routing algorithm (SEAR), energy-efficient routing protocol (EERP), and critical routing data (CRD) using network simulator NS3. For measuring energy efficiency, they used five parameters, namely, packet loss rate, network throughput, energy consumption (EC), E2E (end-to-end) delay, and link breakages. The formula for calculating energy consumption is given in

where E tx shows transmitting energy, E elect gives energy consumption per data bit, E fs is energy for transmitted amplifier, k denotes data bits, and d shows the distance between the sensor nodes. In this algorithm, biosensors are interconnected through a undirected graph by the cost function f ( c ), which includes the weighted residual energy ( WRE ), number of sink hops h c , distance to neighborhoods N i , and queuing delay Q d factors. The network throughput can be measured using

where w 1, w 2, w 3, w 4 are weighted coefficients, and their summation is 1. Link breakage and packet loss can be calculated using

where C e 1 , C e 2 ,…, C en is the estimated energy consumed with neighboring nodes, e init is the initial energy, e net shows network energy, and e tx ( k ) denotes the energy required for transmitting k data bits over a periodic time interval Δ t . The delay can be calculate using Q d = a r + t c D i , (23)

where Q d is the queuing delay, a r denotes arrival data packets D i to sensor node i , and t c shows the transmission capacity of the link.

Abdulmohsin Hammood et al. [ 32 ] proposed inter-WBAN cooperation in the IoMT environment (IWC-IoMT) for providing communication between wireless body sensor networks (WBSN) and those that are beyond their communication range. Efficiency comparison between the proposed algorithm and noninter-WBAN cooperation, namely, two hops in IoMT environment (TH -IoMT) and direct transmission in IoMT environment (DT-IoMT), is done. The formula for calculating the efficiency of DT-IoMT is as shown in

where β i , j is the rate of data transmission from node i to j , P i , j tot is the total power consumption, and the calculating formula is given in

where P amp shows power consumption by amplifier for transmission, and P tx and P rx show power consumption by an internal circuit for transmission and reception, respectively.

The formula for calculating the efficiency of DT-IoMT is as shown in

Here, β TH is the rate of data transmission in DT-IoMT, P TH tot is the total power consumption, and the calculating formula is given in

Finally, the energy efficiency for IWC-IoMT of the 1 st sensor in the network is given by

Here, (1 − P s 1, cn 1 out )(1 − P S 1, cn 2 out ) represents the probability of successful transmission from s 1 to cn 1, and from s 1 to cn 2, respectively, and (1 − P cn 1, T 2 out )(1 − P cn 2, T 2 out ) represents the successful transmission probability from cn 1 to T 2, and from cn 2 to T 2, respectively.

The first term total power, P amp + P tx +2 P rx , contains two nodes for receiving data and one single node for transmitting data, however, in the second term total power, 2 P amp + P tx + P rx contains two data transmission nodes and a single data reception node. Now, β FH shows the rate of data transmission from sensors to coordinators (first phase), and β SH shows the data transmission rate from coordinators to T 2 (second hop).

3.5. Performance Comparison

Figure 7 shows the comparison of the percentage of accuracy achieved by 6 decision tree classification models, namely, J48, iterative dichotomiser 3 (ID3), random forest (RF), correlation ratio (CR), information gain (IG), and gain ratio (GR). Here, the accuracy of RF is 97.71%, ID3 is 71.8%, J48 is 88.95%, IG is 70.83%, GR is 72.26%, and CR is 71.09%. It is clear from the graph that the multiclassifier random forest is giving the highest accuracy among all six algorithms.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.007.jpg

Comparison for accuracy.

Figure 8 shows the comparison of error rates for 3 machine learning decision tree classification algorithms, namely, J48, iterative dichotomiser 3 (ID3), and random forest (RF). Here, the error rate for ID3 is 28.19%, J48 is 11.04%, and RF is 2.28%. The graph clearly shows that the multiclassifier random forest gives a minimum error rate in comparison to other two algorithms.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.008.jpg

Comparison of error rates.

Figure 9 shows a comparison of improvement in the percentage of energy consumed by 3 methodologies (secure and energy-efficient framework using Internet of Medical Things (IoMT) for e-healthcare (SEF-IoMT), routing protocol for low power and lossy networks (RPL) with a new objective function (NEWOF), and an energy efficient IoT e-health model using AI with homomorphic secret sharing). As the graph shows, we have maximum improvement in SEF-IoMT, i.e., 29%, followed by the energy efficient IoT e-health model using AI with homomorphic secret sharing with 17% of improvement in energy consumption, and the last one is the routing protocol for low power and lossy networks (RPL) with a new objective function (NEWOF), which gives 1.45% of energy consumption improvement.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.009.jpg

Comparison graph for energy consumption.

Figure 10 shows the comparison of the improvement in the percentage of packet loss during transmission by 2 methodologies, namely, SEF-IoMT and energy efficient IoT e-health model using AI, with homomorphic secret sharing. As the graph shows, we have a maximum improvement in SEF-IoMT, i.e., 42%, followed by the second methodology, with a 17% improvement in packet loss during transmission within the network.

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.010.jpg

Comparison for packet loss.

4. Health Domain and Its Application

As shown in Figures ​ Figures11 11 and ​ and12, 12 , there are mainly three modules that need to be monitored in a smart healthcare system, namely, homecare [ 15 ], selfcare, and acute care [ 31 ]. In a selfcare system, a person can monitor and access his own fitness through different wearable devices and take necessary actions to prevent diseases in the future [ 8 , 71 ]. In the homecare system, the healthcare providers measure patients' health remotely, and if any problem arises, an alarm will be triggered to alert the doctor and the patient, and both of them collaboratively decide the action that needs to be performed [ 27 , 72 ]. Acute care deals with critical situations, where urgent responses are required. It is usually used for elderly care wearable/implanted devices [ 30 , 73 ].

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.011.jpg

Acute care and selfcare example [ 31 , 70 ].

An external file that holds a picture, illustration, etc.
Object name is CIN2022-7218113.012.jpg

Home care example [ 8 , 31 ].

Each domain uses different types of sensors or a combination of one or more sensors. Table 3 shows various sensors that can be used in a smart healthcare system to detect vital parameters of the client. The first sensor is the accelerometer, which belongs to the selfcare domain. It is used to measure the change in the linear velocity, and it is helpful to detect the blood glucose level of the patient or the person under observation [ 8 ] and the change in the position of the patient [ 70 ] or any other body part of the patient [ 74 ]. A gyroscope detects the angular velocity, which will help in detecting human tilt, and it uses an alarm for the professionals to gain their attention whenever required [ 75 ]. The magnetometer detects the magnetic field and relative orientation. It is mostly used in elderly care devices in conjunction with gyroscopes and accelerometers [ 30 ]. The LM35 sensor changes its voltage according to the change in temperature and generally measures the body temperature of the individual under observation [ 30 , 77 ], whereas DHT11 is used to measure the environmental temperature and humidity [ 76 , 78 ]. LM35 consumes more energy compared to DHT11. A small chip named AD8232 analyzes the pumping stroke of heart muscles, which results in ECG (Electrocardiogram) [ 80 , 82 ]. ECG analyzes heart signals, irrespective of the body state of the person under examination [ 71 , 79 ], whereas MAX 30105 is an integrated optical sensor with 2 LEDs in a single photodetector, processing low noise analog signals in combination with Ardunio to monitor the heart rate between 1.8 V and 3.3 V [ 71 , 79 , 81 ]. ADXL335 is a body position sensor used to check the proper shoulder position to prevent various complications, such as pain, swelling, respiratory problems, etc.

List of sensors and their application in the smart healthcare system.

As we have discussed in this section, various types of sensors can be used in the IoMT network based on the requirement of the system. Besides the selection of accurate application-specific sensor, there are various other aspects that are to be considered while developing an IoMT network, which we are going to discuss in Section 5 .

5. Challenges within a Smart Healthcare System to be Considered during IoMT Network Design

AI provides the capability of a computer or robot, which is controlled by a computer system for performing tasks that are usually done by humans via their intelligence [ 28 , 83 ]. In a smart healthcare system with proper data interpretation techniques, a machine can also monitor health parameters using the implanted/wearable sensors on the body of the person under observation [ 84 , 85 ]. Real-time disease management and prevention with improved user-end experience can be achieved using AI [ 86 , 87 ]. Designing an IoMT-based smart network is very complex because of the below-mentioned challenges that influence the designing techniques at every edge [ 88 ]. The routing protocol will govern the exchange of data between routers and gives information, enabling route selection between nodes [ 89 , 90 ]. In a smart healthcare system, we collect very sensitive patient data using small and ultralow power IoMT devices [ 91 , 92 ]. Hence, the mentioned challenges cannot be tackled within the implanted/wearable IoMT devices, however, they can be balanced in the network and protocol designing techniques with the consideration of effective network topology, power conservation, and channel effectiveness. Hence, the few points that are to be considered especially in IoMT network designs are as follows:

  • Body movements: the real challenge arises when there are changes in network topology because of the movement of the user under observation with on-body sensors or medical devices [ 93 ]. Hence, the routing protocol in IoMT must be adaptable to deal with such unpredictable challenges without compromising the quality of communication strength.
  • Temperature change: the main cause of the rise in the temperature of the IoMT devise is the absorption of radiations by the antennas and the power consumed by node circuitry [ 94 ]. This rise in the temperature of the wearable or implant devices can result in damage to tissues or other body organs of the user under observation. Hence, the considerations of (transmission and computation) the power consumption of IoMT devices are essential.
  • Energy efficiency: energy efficiency determines device size, lifetime, and usability. Hence, the routing protocol should optimize the energy consumption by the IoMT device. Implant devices should have a battery life minimum of 10 to 15 years to avoid repetitive surgery as it results in physical and financial loss [ 59 ]. As far as wearable devices are concerned, frequent battery changes reduce device usability.
  • Range of transmission: when the range of data transmission is short, having postural body movements leads to disconnection and repartitioning among sensor nodes in the IoMT system [ 5 ].
  • Heterogeneous environment: the routing protocol for SHS must be capable of handling challenges because of the heterogeneous environment of BSN applications (for example DexterNet) [ 83 , 95 ].
  • QoS: when we deal with real-time BSN applications, such as ECG, it is very sensitive for data loss, and it is time critical [ 96 ]. Hence, accordingly, the quality of service requirements should be made to deal with such situations. Now, implanted smart sensors have fixed memory and computational capabilities. Hence, the routing protocol should adopt QoS measures [ 35 ].
  • Security: users' data is stored in the cloud for more accurate and faster responses to the patients being monitored using IoMT devices, however, this advancement can lead to the risk of user data being stored or abused [ 1 , 97 , 98 ].

6. Conclusion and Future Scope

This research paper gives an overview of the Internet of Medical things (IoMT) with an emphasis on various enabling techniques used in smart healthcare systems (SHS). Here, we have discussed various methodologies used in smart healthcare systems, such as radio frequency identification (rfid), artificial intelligence (ai), blockchain, etc. This paper provides a detailed description and comparison of various IoMT architectures being used by multiple authors for AI-based smart healthcare systems. A smart healthcare system will work precisely only when it will get correct and accurate data. Hence, we are presenting a comparative analysis of different smart healthcare data collection techniques to maintain the accuracy of collected medical data. For collecting these medical data, we are using implant/wearable IoMT devices on the body of the patient. Implant devices should have a battery life minimum of 10 years to 15 years to avoid repetitive surgery as it results in physical and financial loss. As far as wearable devices are concerned, frequent battery changes reduce device usability. Energy efficiency determines the size, lifetime, and usability of the IoMT devices. Hence, we are focusing on techniques used for energy optimization by the IoMT device. This paper provides a detailed comparison through both tabular and graphical methods showing the recent work done by various authors to maintain the energy efficiency of an IoMT network. For calculating the efficiency of a system, different parameters are being used, such as the amount of energy consumed, packet delivery ratio, battery lifetime, quality of service, power drain, network throughput, delay, transmission rate, etc. In this paper, we are providing different correlation-based equations for finding accuracy and efficiency within the IoMT-based healthcare system. We are also discussing various health domains of the IoMT framework, including the list of sensors with their application in measuring the health of the person under evaluation.

In this paper, we have presented seven key protocol design challenges that need to be considered during the implementation of an IoMT network-based smart healthcare system, namely, the regular body movement of the patient, change in the temperature of the health monitoring device, energy efficiency of the network, transmission range of the device, performance of the IoMT device in a heterogeneous environment, quality of service, and security. In this work, we have compared and elaborated work for the efficient use of energy, which is only one of the key challenges, and the other six challenges need to be explored and analyzed in the future. Considering the sensitivity of medical data, a deep analysis and future enhancement must be done for providing security to the system.

Data Availability

Conflicts of interest.

The authors declare that they have no conflicts of interest.

Advertisement

Advertisement

Systematic Mapping Study of AI/Machine Learning in Healthcare and Future Directions

  • Survey Article
  • Published: 16 September 2021
  • Volume 2 , article number  461 , ( 2021 )

Cite this article

  • Gaurav Parashar   ORCID: orcid.org/0000-0003-4869-1819 1 ,
  • Alka Chaudhary 1 &
  • Ajay Rana 1  

3969 Accesses

9 Citations

3 Altmetric

Explore all metrics

This study attempts to categorise research conducted in the area of: use of machine learning in healthcare , using a systematic mapping study methodology. In our attempt, we reviewed literature from top journals, articles, and conference papers by using the keywords use of machine learning in healthcare . We queried Google Scholar, resulted in 1400 papers, and then categorised the results on the basis of the objective of the study, the methodology adopted, type of problem attempted and disease studied. As a result we were able to categorize study in five different categories namely, interpretable ML, evaluation of medical images, processing of EHR, security/privacy framework, and transfer learning. In the study we also found that most of the authors have studied cancer, and one of the least studied disease was epilepsy, evaluation of medical images is the most researched and a new field of research, Interpretable ML/Explainable AI, is gaining momentum. Our basic intent is to provide a fair idea to future researchers about the field and future directions.

Similar content being viewed by others

research papers in medical domain

What Is Machine Learning?

research papers in medical domain

The role of artificial intelligence in healthcare: a structured literature review

Silvana Secinaro, Davide Calandra, … Paolo Biancone

research papers in medical domain

Revolutionizing healthcare: the role of artificial intelligence in clinical practice

Shuroug A. Alowais, Sahar S. Alghamdi, … Abdulkareem M. Albekairy

Avoid common mistakes on your manuscript.

Introduction

Artificial intelligence (AI) can be defined as a field in which the machine demonstrates intelligence by learning itself. It can be done by deploying various techniques & algorithms to understand human intelligence but does not confine to—John McCarthy. Even though if we do not specifically program the machine and still it can automatically learn and improve itself this defines an intelligent behaviour of machine. Machine learning (ML) is a specific field of AI which relates to techniques that can automatically learn from experience.

The use of machine learning in healthcare had shown many promising solutions which had created confidence in the field. Researchers had used ICT tools with ML in developing solutions for increasing the effectiveness of the earlier methods or procedures. The field of healthcare had also shown tremendous improvement after the use of Big Data, ICT, and AI/machine learning (ML) in precision and speed. These tools have greatly helped physicians and healthcare professionals in their day-to-day working, research, testing the effect of biomedicine on humans using simulations. Every single detail of the patient gets recorded by the doctors with the other information like clinical notes, prescriptions, medical test results, diagnosis, X-rays, MRI scan, sonographic images, etc. This data becomes huge repository of information, which if churned, could give us better insights of treatment, fruitful suggestions and recommendations in diagnosis, progressive pattern of one disease could be correlated to another disease and may lead to new procedure for treatment of a disease and many more. There may be a chance that healthcare professional overlooked a symptom, which if not addressed early could lead to loss of life. Therefore, tools like AI/ML could help in better healthcare services.

The use of tools like IBM Watson Footnote 1 and Google DeepMind [ 1 ] have shown impressive results in healthcare. On top of these tools researchers and developers have designed applications, which harness the capabilities of these tools, to provide personalised patient care, better drug discovery, and improved healthcare organisational performance. According to Wired Footnote 2 Google DeepMind was used to identify protein structures associated with SARS-CoV-2 and understand how the virus functions. One of the oldest scientific puzzle of ’protein folding problem’ was also solved and paved the way for faster development of drugs, better treatment by Google DeepMind. Other contributions in the field of healthcare are use of association rules(AR) which helped analyse malaria in Brazil [ 2 ], according to [ 3 ] diagnosing images of X-rays revealed respiratory condition of the patient and helped in better healthcare services.

ML can be applied to varied fields like defence, automation, finance, automobile, and manufacturing in performing tasks like classification, clustering, and forecasting. It can be categorised into three types supervised, unsupervised, and reinforcement learning.

In supervised learning, algorithms learn from the labeled datasets and prepare a model. After training, we give data to the model, which the model has not seen earlier and belongs to the same category so that it can correctly classify it. In unsupervised learning, algorithms themselves learn, by analysing data and the model then prepares a model, which can be used to correctly cluster the elements. Lastly, in reinforcement learning, machine learns itself from its mistakes or maximising rewards and by reducing penalties.

In this paper, we aim to categorise papers based on the healthcare and machine learning.With the use of AI and ML in healthcare, there have been significant changes in the life of healthcare professionals. The accuracy of medical diagnostic has increased, healthcare professionals have an assistant on which they can rely, they can predict diseases like pneumonia, cancer, heart diseases, Tumour, COVID-19, and many more with better accuracy & precision than before.

In this paper, we attempt to categorise research done in use of machine learning in healthcare. According to best of our knowledge this type of categorisation has not been done earlier . This attempt will become the basis of future research in the field. We attempt to categorise them on the basis of the objective of the study, methodology adopted, type of problem attempted, etc. We discussed in section “ Research Methodology ”, in section “ Literature Survey ”, in section “ Results ” and section “ Conclusion ”.

Research Methodology

This section describes the systematic mapping procedure adopted to study the use of AI/ML in the healthcare domain. This study was conducted using the keywords “Machine Learning” OR ”Healthcare” . The search was conducted on Google scholar and considered only results from Nature, Wiley periodicals, Elsevier, Taylor and Francis, IEEE transactions, ACM, SVN, IET, and ArXiv. Following steps were carried out: (1) Definition of research questions (2) conduct search for primary studies (3) screening of papers for inclusion and exclusion (4) keywording using abstracts (5) data extraction and mapping of studies. The above steps were proposed by [ 4 ].

figure 1

The systematic mapping process

The Systematic Mapping Process

We have adopted the systematic mapping process from [ 4 ] and applied it to the study conducted on use of ML in the healthcare .

Systematic Mapping Process is a well defined, comprehensive overview study done on a particular research topic. According to [ 5 ] it helps researchers do verifiable, unbiased literature review, find research gaps by critical examination of research literature, helps collate evidence, reduce reviewer selection bias & publication bias with transparent inclusion and exclusion criteria.

The process is described here:

We first define research questions and scope of the study.

With respect to the questions framed from the previous step now search is conducted and literature is collected.

Proper screening is done to check whether the selected literature is related to the research question and scope of the research.

Abstract and keywords are scanned for critical survey of the content

In a spreadsheet, collected data is mapped with the RQs.

In the following (see Fig.  1 ) we had shown the process which we had implemented in the study.

Definition of Research Questions

The main intent of the study is to find out the use of ML in the field of healthcare. To start with we have formulated three research questions(see Table  1 ) which are based on the topic of the study. Major goals of systematic mapping study are:-

Find review of the research area

Find the quantity, type of research, and result

Find journals of the published research topic

Therefore, on the basis of the above goals following research questions have been formed.

What type of research has been conducted in the field of use of AI/ML in healthcare? Rationale: This question aims to find the type of research, which has been conducted under the field of the healthcare domain. We need to find out papers published under the topic.

What are the broad categories of papers published under the topic? Rationale: The rationale for this question arises from the outcome of the RQ1. FQ1 gives research papers, then we need to find out the broad category under which the paper lie.

What are the different diseases which have been studied and total number of total paper published under it? Rationale: After categorising the papers, we need to find out different diseases which are being studied in the research done by other researchers. The main intent is to find the least studied disease, which can become a starting point for new research.

Conduct Search for Primary Studies

To conduct the search we followed the steps:

Prepare the search string w.r.t to different databases(as described in Table 2 ). Since we have used only Google scholar therefore we had used a broad search string to cover all papers containing the keywords healthcare and machine learning .

Execute the search and collect the results (see Fig.  2 ).

Categorise the results by studying the papers and grouping them together on the basis of the disease studied (As mentioned in the Table  6 ) & intent of the paper (As mentioned in the Table  5 ).

figure 2

Raw text from Google scholar results

We took 1400 search results from google scholar and transferred them to spreadsheet based on the query mentioned in Table  2 . This data of around 1400 entries will be further drilled down in next section by excluding the entries which are not related to the study.

Screening of Papers for Inclusion and Exclusion (Relevant Papers)

In this step we exclude all the papers that are not relevant in the study. By this we also mean that the papers which are not related to the RQs (Refer Table  5 ), papers which do not from Nature, Wiley periodicals, Elsevier, Taylor and Francis, IEEE transactions, ACM, SVN, IET, and ArXiv are excluded from the final list.

Using the above criteria we retained those entries, which are based on Inclusion criteria (Refer Table  3 ). After using the above exclusion criteria we drilled down the entries which we finally considered were 42 .

Keywording Using Abstracts

For our study we followed the systematic process of classifying the results from Google Scholar. For Keywording we followed these steps:

The result collected from the previous step are analysed by surveying abstract.

Abstract are surveyed for keywords and content. Then context of the study is evaluated.

Group the result on the basis of context and keywords (Refer Table  4 ).

Data Extraction and Mapping of Studies

We collected all the information in a spreadsheet with the information like s.no., paper title, abstract, keywords, year of publication, authors, name of publisher, name of periodical/journal/conference, major findings, major shortcomings. After that we mapped the RQs (see Table 1 ) to each entry.

Literature Survey

Interpretable machine learning.

Interpretable models are those which explains itself. Interpretable models are linear regression, logistic regression and decision trees. For instance, if we use decision tree model then we can easily extract decision rules as explanations for the model.

In [ 6 ] authors referred to the use of ML in healthcare with an emphasis on Interpretability. Interpretable ML refers to models which can provide rationale on predictions made by the model. The basic impediment in the adoption of ML in healthcare is its BlackBox nature. Since we have to develop ML as a tool that can act as an assistant to physicians, therefore, we need to make its output more explainable. Mere providing metrics like AUC, recall, precision, F-Score may not suffice. We need to develop more interpretable models that themselves can provide explanations of their predictions. The authors [ 7 ] proposed a model which adds important value to features and make the output interpretable. Authors [ 8 ] developed reasoning through the use of visual indicators making the model interpretable. In [ 9 ] authors proposed an interpretable tree from a decision forest making understandable by humans. As proposed in [ 10 , 11 ] interpretable ML models helps develop a reasonable and data-driven decision support system that results in personalised decisions.

Authors [ 12 ] applied deep learning on medical data of patients for developing interpretable predictions for decision support.

Evaluation of Medical Images

In this category, the authors discussed evaluation of medical images for better diagnosis using machine learning models.

In [ 13 , 14 , 15 , 16 , 17 ] authors used deep learning models, Neural Network to classify different diseases, organ segmentation and compared it with the diagnosis of health care professionals for diagnostic accuracy. In [ 18 ] authors proposed a novel colour deconvolution for stain separation and colour normalisation of images. In [ 19 ] authors performed a comparison of five colour normalisation algorithms and found stain colour normalisation algorithms performed better, which had high stain segmentation accuracy and low computational complexity. In their review paper authors [ 20 ] did a comparison of different image segmentation methods and related studies with medical imaging for AI-assisted diagnosis of COVID-19. In [ 21 ] authors explained AI, ML, DL, and CNN and the use of these techniques in imaging. [ 22 ] discussed image enhancements method with noise suppression by enhancing low light regions.

Processing of Electronic Health Record (EHR)

In this category, we had compiled papers that had processed electronic health records of patients.

In the paper [ 17 ] authors proposed diagnostic of pneumonia in a resource-constrained environment. The authors of [ 23 , 24 , 25 ] discussed the processing of electronic health records and used ML algorithms to categorise disease. The authors [ 26 ] trained their proposed model on large dataset and performed regression and classification to check their effectiveness and accuracy. In [ 27 ], a medical recommendation system was proposed using Fast Fourier transformation coupled with a machine learning ensemble model. The model uses this model for disease risk prediction to provide medical recommendations like medical tests and other recommendations for chronic heart disease patients. In [ 28 ] authors proposed the use of graphical structure of electronic health records and find hidden structure from it. In [ 29 ] proposed a model that provides help to physicians to evaluate the quality of evidence for better decision making. Authors used risk of bias assessment in textual data using Natural Language Processing.

Security/Privacy Framework

Under this category, we will summarised papers related to the security and privacy framework for safeguarding health records transferred over network or internet.

Authors of [ 30 ] researched on novel design of smart and secure healthcare information system by adopting machine learning. It also employed advanced security mechanism to handle the big data of the healthcare industry. This framework used many security tools to secure the data like encryption, monitoring the activity, access control, and many other mechanism. This paper [ 31 ] discussed the privacy-preserving collaborative model using ML tools for medical diagnosis systems.

Most of the privacy protection methods are centralized. There is a need for a decentralized system that can help in mitigating several challenges like single-point-of-failure, modification of records, privacy preservation, improper information exchange that may lead to risk of patient’s life. To protect, many researchers have proposed different algorithms [ 32 , 33 , 34 , 35 ]. Models like VERTIGO, GLORE, and WebDISCO were designed for privacy preservation and predictive modelling. These models aimed to preserve privacy by sending partially-trained machine learning models rather than patient data. This way the information is preserved, and develop trust between different parties.

Many other distributed privacy-preserving models were developed those were based on Blockchain technology. They use the technology to update models as in Blockchain like ModelChain, EXPLORER, Distributed Autonomous Online Learning sequentially.

Secure multiparty computation(SMC) for privacy preservation that do computations on encrypted data with personally identifiable information had opened a new dimension. Data is a very precious commodity, therefore techniques like privacy preserving scoring of Tree Ensembles [ 36 ] are designed to provide a framework that provides cryptographic protocols for sending data securely.

Transfer Learning

In this category, we summarised research papers related to transfer learning. Transfer learning is a technique in which we gain knowledge from one problem and use the same knowledge to solve different but related problem. In [ 37 ] authors proposed a technique for handling missing data using transfer learning perspective. The proposed classifier learn weights and then complete portion of the dataset and then transfer it to the target domain. In [ 38 ] authors used transfer learning approach to predict breast cancer using model trained on a task some other task. A model trained on ImageNet databases containing 1.2 million images is used as feature extractor. The model is combined with other components to perform classification. [ 39 , 40 ] uses data generated by different wearable devices using federated learning, and then builds machine learning model by transfer learning. The study was applied to diagnose Parkinson’s disease.

After the systematic mapping process we got categories of research literature as mentioned in Table  5 . This table describes category and total number of papers under the category.

From Table  5 we can clearly observe that most research is being done in Processing of Medical Images this might be due to the availability of the dataset for research purpose. In case of Processing of EHR, which is second most researched category, might be again due to availability of the dataset. In case of Interpretable ML, since it is a new field and slowly gaining momentum therefore, researchers are taking interest because this gives a rationale to the outcome of the model result. This is a vey important attribute, when it comes to certain domains where high stakes are at risk. Like for example in healthcare, defence, and finance. Lastly, In case of Transfer Learning, it is a field which talks about using the domain knowledge of one domain and use it in another related domain. So according to us researchers use this technique to apply for testing the results. Therefore it has a very limited number of research.

From Table  6 it is clearly evident that most researched disease is Cancer, which is 38 and Pneumonia with frequency as 4, Alzheimer as 3, Parkinson as 2 and Epilepsy as 2 are the least researched diseases. These results have been extracted from 1400 papers downloaded from Google Scholar.

In this paper, we have provided a brief overview of the directions of research in the healthcare domain using Machine Leaning. As described earlier, these papers can show researchers path where they can work. The result is based on literature review done on around 1400 papers and filtered down to 42 papers. As described in section “ Literature Survey ”, we categorised the research into 5 broad areas and found most of the research is done in the field of Evaluation of Medical Images in which authors researched many diseases like cancer, heart disease, COVID-19, Parkinson, etc. Authors used different kinds of dataset like images, voice, and electronic health records. Using these dataset they predicted these diseases using machine learning/AI. As described in section “ Processing of Electronic Health Record (EHR) ” second major contribution is done in this category. We would like to conclude that in section “ Interpretable Machine Learning ” very little research has been done so this area can be chosen for further research.

https://www.healthcareglobal.com/technology-and-ai-3/four-ways-which-watson-transforming-healthcare-sector

https://www.wired.co.uk/article/ai-healthcare-boom-deepmind

Powles J, Hodson H. Google DeepMind and healthcare in an age of algorithms. Health Technol. 2017;7(4):351.

Article   Google Scholar  

Baroni L, Salles R, Salles S, Guedes G, Porto F, Bezerra E, Barcellos C, Pedroso M, Ogasawara E. An analysis of malaria in the Brazilian Legal Amazon using divergent association rules. J Biomed Inf. 2020;108:103512.

Rajpurkar P, Irvin J, Zhu K, Yang B, Mehta H, Duan T, Ding D, Bagul A, Langlotz C, Shpanskaya K, et al. 2017. arXiv:1711.05225 .

Petersen K, Feldt R, Mujtaba S, Mattsson M. In: 12th international conference on evaluation and assessment in software engineering (EASE) 12; 2008. pp. 1–10.

Haddaway NR, Westgate MJ. Predicting the time needed for environmental systematic reviews and systematic maps. Conserv Biol. 2019;33(2):434.

Ahmad MA, Eckert C, Teredesai A. In: Proceedings of the 2018 ACM international conference on bioinformatics, computational biology, and health informatics; 2018. pp. 559–60.

Lundberg S, Lee SI. 2017. arXiv:1705.07874 .

Yu F, Ip HH. Semantic content analysis and annotation of histological images. Comput Biol Med. 2008;38(6):635.

Sagi O, Rokach L. Explainable decision forest: transforming a decision forest into an interpretable tree. Inf Fusion. 2020;61:124.

Stiglic G, Kocbek P, Fijacko N, Zitnik M, Verbert K, Cilar L. Interpretability of machine learning-based prediction models in healthcare. Wiley Interdiscipl Rev Data Min Knowl Disc. 2020;10(5):e1379.

Google Scholar  

Ribeiro MT, Singh S, Guestrin C. In: Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining; 2016. pp. 1135–44.

Rebane J, Samsten I, Papapetrou P. Exploiting complex medical data with interpretable deep learning for adverse drug event predictio. Artif Intell Med. 2020;109:101942.

Liu X, Faes L, Kale AU, Wagner SK, Fu DJ, Bruynseels A, Mahendiran T, Moraes G, Shamdas M, Kern C, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. Lancet Dig Health. 2019;1(6):e271.

Panayides AS, Amini A, Filipovic ND, Sharma A, Tsaftaris SA, Young A, Foran D, Do N, Golemati S, Kurc T, et al. AI in medical imaging informatics: current challenges and future directions. IEEE J Biomed Health Inf. 2020;24(7):1837.

Oktay O, Ferrante E, Kamnitsas K, Heinrich M, Bai W, Caballero J, Cook SA, De Marvao A, Dawes T, Oregan DP, et al. Anatomically constrained neural networks (ACNNs): application to cardiac image enhancement and segmentation. IEEE Trans Med Imaging. 2017;37(2):384.

Jeyaraj PR, Nadar ERS. Deep Boltzmann machine algorithm for accurate medical image analysis for classification of cancerous region. Cogn Comput Syst. 2019;1(3):85.

Harmon SA, Sanford TH, Xu S, Turkbey EB, Roth H, Xu Z, Yang D, Myronenko A, Anderson V, Amalou A, et al. A systematic review of antibody mediated immunity to coronaviruses: kinetics, correlates of protection, and association with severity. Nature Commun. 2020;11(1):1.

Zheng Y, Jiang Z, Zhang H, Xie F, Shi J, Xue C. Adaptive color deconvolution for histological WSI normalization. Comput Methods Progr Biomed. 2019;170:107.

Hoffman RA, Kothari S, Wang MD. In: 2014 36th annual international conference of the IEEE engineering in medicine and biology society, IEEE; 2014. pp. 194–7.

Feng S, et al. 2020. arXiv:2004.02731 .

Currie G, Hawk KE, Rohren E, Vial A, Klein R. Machine learning and deep learning in medical imaging: intelligent imaging. J Med Imaging Radiat Sci. 2019;50(4):477.

Xia W, Chen EC, Peters T. Endoscopic image enhancement with noise suppression. Healthcare Technol Lett. 2018;5(5):154.

Capotorti A. Probabilistic inconsistency correction for misclassification in statistical matching, with an example in health care. Int J Gener Syst. 2020;49(1):32.

Article   MathSciNet   Google Scholar  

Li JP, Haq AU, Din SU, Khan J, Khan A, Saboor A. Heart disease identification method using machine learning classification in e-healthcare. IEEE Access. 2020;8:107562.

Naydenova E, Tsanas A, Casals-Pascual C, De Vos M. In: 2015 IEEE global humanitarian technology conference (GHTC), IEEE; 2015. pp. 377–84.

Haq AU, Li JP, Memon MH, Malik A, Ahmad T, Ali A, Nazir S, Ahad I, Shahid M, et al. Feature selection based on L1-norm support vector machine and effective recognition system for Parkinson disease using voice recordings. IEEE Access. 2019;7:37718.

Zhang J, Lafta RL, Tao X, Li Y, Chen F, Luo Y, Zhu X. Coupling a fast fourier transformation with a machine learning ensemble model to support recommendations for heart disease patients in a telehealth environment. IEEE Access. 2017;5:10674.

Choi E, Xu Z, Li Y, Dusenberry M, Flores G, Xue E, Dai A. In: Proceedings of the AAAI conference on artificial intelligence, vol. 34; 2020. pp. 606–13.

Pereira RG, Castro GZ, Azevedo P, Tôrres L, Zuppo I, Rocha T, Júnior AAG. In: 2020 IEEE 33rd international symposium on computer-based medical systems (CBMS), IEEE; 2020. pp. 1–6.

Kaur P, Sharma M, Mittal M. Big data and machine learning based secure healthcare framework. Procedia Comput Sci. 2018;132:1049.

Wang F, Zhu H, Liu X, Lu R, Hua J, Li H, Li H. Privacy-preserving collaborative model learning scheme for E-healthcare. IEEE Access. 2019;7:166054.

Wu Y, Jiang X, Kim J, Ohno-Machado L. Grid Binary LOgistic REgression (GLORE): building shared models without sharing data. J Am Med Inf Assoc. 2012;19(5):758.

Kuo TT, Ohno-Machado L. Modelchain: decentralized privacy-preserving healthcare predictive modeling framework on private blockchain networks. 2018.

Wang S, Jiang X, Wu Y, Cui L, Cheng S, Ohno-Machado L. Expectation propagation logistic regression (explorer): distributed privacy-preserving online model learning. J Biomed Inf. 2013;46(3):480.

Li Y, Jiang X, Wang S, Xiong H, Ohno-Machado L. Vertical grid logistic regression (vertigo). J Am Med Inf Assoc. 2016;23(3):570.

Fritchman K, Saminathan K, Dowsley R, Hughes T, De Cock M, Nascimento A, Teredesai A. In: 2018 IEEE international conference on big data (Big Data); 2018. pp. 2413–22. https://doi.org/10.1109/BigData.2018.8622627 .

Wang G, Lu J, Choi KS, Zhang G. A transfer-based additive LS-SVM classifier for handling missing data. IEEE Trans Cybern. 2018;50(2):739.

Dey N, Das H, Naik B, Behera HS. Big data analytics for intelligent healthcare management. Cambridge: Academic Press; 2019.

Chen Y, Qin X, Wang J, Yu C, Gao W, Chen Y, Qin X, Wang J, Yu C, Gao W. Fedhealth: a federated transfer learning framework for wearable healthcare. IEEE Intell Syst. 2020;35(4):83.

Download references

Author information

Authors and affiliations.

AIIT, AMITY University, Noida, Uttar Pradesh, India

Gaurav Parashar, Alka Chaudhary & Ajay Rana

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Gaurav Parashar .

Ethics declarations

Conflict of interest.

The authors declare that they have no conflict of interest.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This article is part of the topical collection “Intelligent Systems” guest edited by Geetha Ganesan, Lalit Garg, Renu Dhir, Vijay Kumar and Manik Sharma.

Rights and permissions

Reprints and permissions

About this article

Parashar, G., Chaudhary, A. & Rana, A. Systematic Mapping Study of AI/Machine Learning in Healthcare and Future Directions. SN COMPUT. SCI. 2 , 461 (2021). https://doi.org/10.1007/s42979-021-00848-6

Download citation

Received : 04 August 2021

Accepted : 01 September 2021

Published : 16 September 2021

DOI : https://doi.org/10.1007/s42979-021-00848-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Machine learning (ML)
  • Transfer learning (TL)
  • Interpretable ML
  • Electronic health records (EHR)
  • Security framework
  • Privacy framework
  • Find a journal
  • Publish with us
  • Track your research

A large language model from Google Research, designed for the medical domain.

research papers in medical domain

Introduction

Med-PaLM is a large language model (LLM) designed to provide high quality answers to medical questions. Our second version, Med-PaLM 2, is one of the research models that powers MedLM – a family of foundation models fine-tuned for the healthcare industry. MedLM is now available to Google Cloud customers who have been exploring a range of applications, from basic tasks to complex workflows.

Med-PaLM harnesses the power of Google’s large language models, which we have aligned to the medical domain and evaluated using medical exams, medical research, and consumer queries. Our first version of Med-PaLM, preprinted in late 2022 and published in Nature in July 2023 , was the first AI system to surpass the pass mark (>60%) in the U.S. Medical Licensing Examination (USMLE) style questions. Med-PaLM also generates accurate, helpful long-form answers to consumer health questions, as judged by panels of physicians and users.

We introduced Med-PaLM 2 at Google Health’s annual event, The Check Up, in March 2023. Med-PaLM 2 was the first to reach human expert level on answering USMLE-style questions. According to physicians, the model's long-form answers to consumer medical questions improved substantially.

research papers in medical domain

Med-PaLM 2 reached 86.5% accuracy on the MedQA medical exam benchmark in research

Since medicine is inherently multimodal, we have also introduced research on a multimodal version of Med-PaLM, called Med-PaLM M . We are also exploring a wide range of other techniques to build medical AI systems that can bring information together from a wide range of data modalities.

Medical question–answering: a grand challenge for AI

Progress in AI over the last decade has enabled it to play an increasingly important role in healthcare and medicine . Breakthroughs such as the Transformer have enabled LLMs – such as PaLM –  and other large models to scale to billions of parametersletting generative AI move beyond the limited pattern-spotting of earlier AIs and into the creation of novel expressions of content, from speech to scientific modeling. 

Developing AI that can answer medical questions accurately has been a long-standing challenge with several research advances over the past few decades. While the topic is broad, answering USMLE-style questions has recently emerged as a popular benchmark for evaluating medical question answering performance.

Above is an example USMLE-style question. You are presented with a vignette containing a description of the patient, symptoms, and medications.

Answering the question accurately requires the reader to understand symptoms, examine findings from a patient’s tests, perform complex reasoning about the likely diagnosis, and ultimately, pick the right answer for what disease, test, or treatment is most appropriate. In short, a combination of medical comprehension, knowledge retrieval, and reasoning is necessary to do well. It takes years of training for clinicians to be able to accurately and consistently answer these questions.

The generation capabilities of large language models also enable them to produce long-form answers to consumer medical questions. However, ensuring model responses are accurate, safe, and helpful has been a crucial research challenge, especially in this safety-critical domain.

research papers in medical domain

In a pairwise study, Med-PaLM 2 answers were preferred to physician answers across eight of nine axes considered.

Evaluating answer quality

We assessed Med-PaLM and Med-PaLM 2 against a benchmark we call ‘MultiMedQA’, which combines seven question answering datasets spanning professional medical exams, medical research, and consumer queries. Med-PaLM was the first AI system to obtain a passing score on USMLE-style questions from the MedQA dataset, with an accuracy of 67.6%. Med-PaLM 2 improves on this further with state of the art performance of 86.5%.

Importantly, in this work we go beyond multiple-choice accuracy to measure and improve model capabilities in medical question answering. Our model’s long-form answers were tested against several criteria — including scientific factuality, precision, medical consensus, reasoning, bias, and likelihood of possible harm — which were evaluated by clinicians and non-clinicians from a range of backgrounds and countries. Both Med-PaLM and Med-PaLM 2 performed encouragingly across three datasets of consumer medical questions. In a pairwise study, Med-PaLM 2 answers were preferred to physician answers across eight of nine axes considered.

Check out how Med-PaLM 2 answers medical questions

*Examples only. Med-PaLM 2 is currently being evaluated to ensure safe and responsible use.

Extending Med-PaLM 2 Beyond Language

The practice of medicine is inherently multi-modal and incorporates information from images, electronic health records, sensors, wearables, genomics and more. We believe AI systems that leverage these data at scale using self-supervised learning with careful consideration of privacy, safety and health equity will be the foundation of the next generation of medical AI systems that scale world-class healthcare to everyone.

Building on the “PaLM-E” vision-language model, we designed a multimodal version of Med-PaLM, called Med-PaLM M . This system can synthesize and communicate information from images like chest X-rays, mammograms, and more to help doctors provide better patient care. Within scope are several modalities alongside language: dermatology, retina, radiology (3D and 2D), pathology, health records and genomics. We’re excited to explore how this technology can benefit clinicians in the future.

research papers in medical domain

*Example only. This image reflects early exploration of Med-PaLM M's future capabilities.

Limitations

While Med-PaLM 2 reached state-of-the-art performance on several multiple-choice medical question answering benchmarks, and our human evaluation shows answers compare favorably to physician answers across several clinically important axes, we know that more work needs to be done to ensure these models are safely and effectively deployed.

Careful consideration will need to be given to the ethical deployment of this technology including rigorous quality assessment in different clinical settings with guardrails to mitigate against risks. For example, the potential harms of using a LLM for diagnosing or treating an illness are much greater than using a LLM for information about a disease or medication. Additional research will be needed to assess LLMs used in healthcare for homogenization and amplification of biases and security vulnerabilities inherited from base models.

We dive into many important areas for further research in our Med-PaLM and Med-PaLM 2 papers.

In the press

Scientific American: AI Chatbots Can Diagnose Medical Conditions at Home. How Good Are They?

CNBC: Google’s working on an updated version of its medical A.I. that can answer health questions

Med Page Today: Google AI Performs at 'Expert' Level on U.S. Medical Licensing Exam

New Scientist: Google's AI is best yet at answering medical and health questions

The Economist: A bioethicist and a professor of medicine on regulating AI in healthcare

Advisory Board: Are AI doctors on the horizon?

STAT: Google will let health care customers test its generative AI model, ramping up rivalry with GPT-4

MobiHealthNews: Google to offer limited access to medical LLM

Forbes: How Tech Leaders Compete In The Battle Of Healthcare AI

Google Cloud Blog: MedLM

Acknowledgements

Med-PaLM research:  

Karan Singhal*, Shekoofeh Azizi*, Tao Tu*, S. Sara Mahdavi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tanwani, Heather Cole-Lewis, Stephen Pfohl, Perry Payne, Martin Seneviratne, Paul Gamble, Chris Kelly, Abubakr Babiker, Yu-Han Liu, Nathanael Schärli, Aakanksha Chowdhery, Philip Mansfield, Dina Demner-Fushman, Blaise Agüera y Arcas, Dale Webster, Greg S. Corrado, Yossi Matias, Katherine Chou, Juraj Gottweis, Nenad Tomasev, Yun Liu, Alvin Rajkomar, Joelle Barral, Christopher Semturs, Alan Karthikesalingam**, and Vivek Natarajan**

Med-PaLM 2 research:

Karan Singhal*, Tao Tu*, Juraj Gottweis*, Rory Sayres*, Ellery Wulczyn, Le Hou, Kevin Clark, Stephen Pfohl, Heather Cole-Lewis, Darlene Neal, Mike Schaekermann, Amy Wang, Mohamed Amin, Sami Lachgar, Philip Mansfield, Sushant Prakash, Bradley Green, Ewa Dominowska, Blaise Aguera y Arcas, Nenad Tomasev, Yun Liu, Renee Wong, Christopher Semturs, S. Sara Mahdavi, Joelle Barral, Dale Webster, Greg S. Corrado, Yossi Matias, Shekoofeh Azizi**, Alan Karthikesalingam**, Vivek Natarajan**

Med-PaLM M research:

Tao Tu*, Shekoofeh Azizi*, Danny Driess, Mike Schaekermann, Mohamed Amin, Pi-Chuan Chang, Andrew Carroll, Chuck Lau, Ryutaro Tanno, Ira Ktena, Basil Mustafa, Aakanksha Chowdhery, Yun Liu, Simon Kornblith, David Fleet, Philip Mansfield, Sushant Prakash, Renee Wong, Sunny Virmani, Christopher Semturs, S Sara Mahdavi, Bradley Green, Ewa Dominowska, Blaise Aguera y Arcas, Joelle Barral, Dale Webster, Greg S. Corrado, Yossi Matias, Karan Singhal, Pete Florence, Alan Karthikesalingam** and Vivek Natarajan**

* - equal contributions

** - equal leadership

Additional contributors:  

Preeti Singh, Kavita Kulkarni, Jonas Kemp, Anna Iurchenko, Lauren Winer, Will Vaughan, Le Hou, Jimmy Hu, Yuan Liu, Jonathan Krause, John Guilyard, Divya Pandya.

We thank Michael Howell, Boris Babenko, Naama Hammel, Cameron Chen, Basil Mustafa, David Fleet, Douglas Eck, Simon Kornblith, Fayruz Kibria, Gordon Turner, Lisa Lehmann, Ivor Horn, Maggie Shiels, Shravya Shetty, Jukka Zitting, Evan Rappaport, Lucy Marples, Viknesh Sounderajah, Ali Connell, Jan Freyberg, Dave Steiner, Cian Hughes, Brett Hatfield, SiWai Man, Gary Parakkal, Sudhanshu Sharma, Megan Jones-Bell, Susan Thomas, Martin Ho, Sushant Prakash, Bradley Green, Ewa Dominowska, Frederick Liu, Kate Weber, Annisah Um’rani, Laura Culp, and Xuezhi Wang for their assistance, insights, and feedback during our research. 

We are also grateful to Yossi Matias, Karen DeSalvo, Zoubin Ghahramani, James Manyika, and Jeff Dean for their support throughout this project.

Discover More

A responsible path to generative AI in healthcare

Our latest health AI research updates

The next generation of AI for developers and Google Workspace

Google Research, 2022 & beyond: Health

Google Health + blog

Help | Advanced Search

Computer Science > Computer Vision and Pattern Recognition

Title: a survey on domain generalization for medical image analysis.

Abstract: Medical Image Analysis (MedIA) has emerged as a crucial tool in computer-aided diagnosis systems, particularly with the advancement of deep learning (DL) in recent years. However, well-trained deep models often experience significant performance degradation when deployed in different medical sites, modalities, and sequences, known as a domain shift issue. In light of this, Domain Generalization (DG) for MedIA aims to address the domain shift challenge by generalizing effectively and performing robustly across unknown data distributions. This paper presents the a comprehensive review of substantial developments in this area. First, we provide a formal definition of domain shift and domain generalization in medical field, and discuss several related settings. Subsequently, we summarize the recent methods from three viewpoints: data manipulation level, feature representation level, and model training level, and present some algorithms in detail for each viewpoints. Furthermore, we introduce the commonly used datasets. Finally, we summarize existing literature and present some potential research topics for the future. For this survey, we also created a GitHub project by collecting the supporting resources, at the link: this https URL

Submission history

Access paper:, references & citations.

  • Google Scholar
  • Semantic Scholar

BibTeX formatted citation

BibSonomy logo

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

  • Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

This paper is in the following e-collection/theme issue:

Published on 19.2.2024 in Vol 26 (2024)

Media Use Behavior Mediates the Association Between Family Health and Intention to Use Mobile Health Devices Among Older Adults: Cross-Sectional Study

Authors of this article:

Author Orcid Image

Original Paper

  • Jinghui Chang 1 * , PhD   ; 
  • Yanshan Mai 2 *   ; 
  • Dayi Zhang 2   ; 
  • Xixi Yang 1   ; 
  • Anqi Li 1 , MSc   ; 
  • Wende Yan 2   ; 
  • Yibo Wu 3 , PhD   ; 
  • Jiangyun Chen 1 , PhD  

1 School of Health Management, Southern Medical University, Guangzhou, China

2 School of Public Health, Southern Medical University, Guangzhou, China

3 School of Public Health, Peking University, Beijing, China

*these authors contributed equally

Corresponding Author:

Jiangyun Chen, PhD

School of Health Management

Southern Medical University

Number 1023, South Shatai Road

Baiyun District

Guangzhou, 510515

Phone: 86 1 858 822 0304

Email: [email protected]

Background: With the advent of a new era for health and medical treatment, characterized by the integration of mobile technology, a significant digital divide has surfaced, particularly in the engagement of older individuals with mobile health (mHealth). The health of a family is intricately connected to the well-being of its members, and the use of media plays a crucial role in facilitating mHealth care. Therefore, it is important to examine the mediating role of media use behavior in the connection between the family health of older individuals and their inclination to use mHealth devices.

Objective: This study aims to investigate the impact of family health and media use behavior on the intention of older individuals to use mHealth devices in China. The study aims to delve into the intricate dynamics to determine whether media use behavior serves as a mediator in the relationship between family health and the intention to use mHealth devices among older adults. The ultimate goal is to offer well-founded and practical recommendations to assist older individuals in overcoming the digital divide.

Methods: The study used data from 3712 individuals aged 60 and above, sourced from the 2022 Psychology and Behavior Investigation of Chinese Residents study. Linear regression models were used to assess the relationships between family health, media use behavior, and the intention to use mHealth devices. To investigate the mediating role of media use behavior, we used the Sobel-Goodman Mediation Test. This analysis focused on the connection between 4 dimensions of family health and the intention to use mHealth devices.

Results: A positive correlation was observed among family health, media use behavior, and the intention to use mHealth devices (r=0.077-0.178, P<.001). Notably, media use behavior was identified as a partial mediator in the relationship between the overall score of family health and the intention to use mHealth devices, as indicated by the Sobel test (z=5.451, P<.001). Subgroup analysis further indicated that a complete mediating effect was observed specifically between family health resources and the intention to use mHealth devices in older individuals with varying education levels.

Conclusions: The study revealed the significance of family health and media use behavior in motivating older adults to adopt mHealth devices. Media use behavior was identified as a mediator in the connection between family health and the intention to use mHealth devices, with more intricate dynamics observed among older adults with lower education levels. Going forward, the critical role of home health resources must be maximized, such as initiatives to develop digital education tailored for older adults and the creation of media products specifically designed for them. These measures aim to alleviate technological challenges associated with using media devices among older adults, ultimately bolstering their inclination to adopt mHealth devices.

Introduction

The 2022 United Nations report on “World Population Prospects” predicted that by 2050, the global population will reach 9.7 billion. Within this demographic shift, 1.5 billion individuals aged 65 and above are anticipated, constituting 16% of the total population [ 1 ]. Notably, the trend of population aging is intensifying. In the context of population dynamics, China, as a heavily populated nation, is undergoing significant and intricate transformations. The Seventh National Population Census of China revealed that there are 264 million individuals aged 60 or older in the country, comprising 18.7% of the overall population [ 2 ]. This underscores the profound changes in China’s demographic landscape. The rapidly increasing aging rate in China poses substantial challenges for the future development of the country’s medical services. Over 180 million older adults in China grapple with chronic diseases, and a staggering 75% of them contend with multiple chronic illnesses [ 3 ]. This places older individuals in a high-risk and vulnerable category, imposing considerable financial and operational burdens on China’s medical and health sector.

Mobile health (mHealth) devices typically encompass mHealth programs and wearable devices [ 4 ]. Functioning as portable tools leveraging internet communication technology, these devices continuously monitor diverse physiological conditions. They have the capability to track and record users’ daily lifestyle and health status data in real-time [ 5 ]. These real-time data are instrumental for users to make informed adjustments to their health behaviors, facilitated by prompt feedback on health information [ 6 ]. The utilization of mHealth devices addresses the emerging need for self-monitoring and self-management within the expanding medical service market, aligning with heightened health awareness among consumers. These devices play a pivotal role in enabling early diagnosis, intervention, clinical treatment, and monitoring of various diseases by continuously supervising vital signs in real-time. However, it is noteworthy that despite the potential benefits, mHealth devices are not widely embraced by older individuals [ 7 ]. Consequently, the robust functionalities and inherent advantages of these devices remain underutilized within this demographic group. Emerging as an inevitable outcome of the internet era and the aging society, mHealth holds substantial potential to offer a promising solution to meet the escalating demands for medical services in developing countries [ 8 ]. Recognizing that older individuals constitute the most frequent and substantial users of health services [ 9 ], it becomes imperative to cultivate a new social trend, encouraging the integration of older individuals with mHealth [ 10 ].

Prior research has demonstrated that mHealth can significantly enhance the health, well-being, and longevity of older individuals in the digital era. However, it also introduces a new social governance challenge—the digital divide among older individuals [ 11 , 12 ]. This divide arises from challenges in accessing or utilizing information infrastructure coupled with a lower level of digital education, resulting in difficulties for older individuals to stay abreast of social, economic, and technological advancements [ 13 ]. As outlined in the 50th Statistical Report on the Development of the Internet in China by the China Internet Network Information Center, individuals aged 60 and above constitute the predominant group of non-netizens, comprising 41.6% of this demographic [ 14 ]. A confluence of personal, family, social, and technological factors collectively contributes to the estrangement of older individuals from engaging with new media, such as the internet [ 15 ]. Research indicates that the motivation for older individuals to actively seek health information on the internet is closely tied to their interactions with family or friends [ 16 ]. Older adults primarily rely on their families for social support, and the cohesion within the family unit significantly influences their overall health status [ 17 , 18 ].

Family health represents a collective resource that emerges from the interconnected well-being of each family member, encompassing their health, interactions, capacities, and the family’s overall physical, social, emotional, economic, and medical resources [ 19 ]. As an interdisciplinary concept, evaluating family health necessitates a thorough examination of various factors, including but not limited to family functioning, emotional support, financial resources, and access to external services [ 20 ]. Existing literature demonstrates that family support plays a pivotal role in motivating older individuals to seek medical services [ 21 ]. Additionally, family function and overall health serve as crucial indicators for assessing the mental well-being of older individuals [ 22 ]. Communication within the family, involving interactions with children, grandchildren, and peer groups, influences older individuals’ inclination to adopt smart senior care solutions [ 23 ]. While numerous articles predominantly explore family health from a singular dimension [ 24 - 26 ], there exists a research gap concerning the specific influence of family health on older individuals’ intention to adopt mHealth devices.

The evolution of mHealth is intricately linked to the technical backing of media. Media technology plays a dual role—it not only generates visual data representing health conditions detected by mHealth devices [ 27 ] but also serves as a platform for the public to exchange and share medical information. In the case of older adults, their acceptance of new health services and access to health information are influenced in distinct ways by the utilization of media devices [ 28 , 29 ]. A Chinese empirical analysis revealed a fundamental correlation between media use and the health level of older adults [ 30 ]. Social media communication is considered an intervention measure to alleviate the loneliness experienced by older adults, achieved by enhancing social support and contact levels, thereby fostering positive responses to emerging technologies [ 31 , 32 ]. Furthermore, the utilization of mobile phones and other media significantly influences disparities in medical care. Increasing the frequency of contact and sustained use of media by older individuals can contribute to unlocking the considerable potential of mobile medical technology in the health care of older individuals [ 33 ].

In summary, there is an immediate and practical need to reduce the digital divide among older adults. The willingness of older individuals to embrace mHealth devices, as reflected in surveys, signifies their acceptance of new health technologies and, to a certain extent, their integration into the era of mHealth. Previous research on factors influencing the intention to use mHealth devices among older adults has predominantly centered on understanding the behavioral motivations and mechanisms behind users’ intentions to use, emphasizing the impact of technical and social aspects on actual usage behavior [ 34 ]. Research on influencing factors has primarily delved into age, gender, education level, BMI, income, and health status, among other individual aspects [ 35 - 37 ]. However, there is a paucity of studies examining external environmental factors, notably the influence of family and social dynamics, particularly among the older adult population in China. A previous study indicated that family internet access enhances older adults’ cognitive function and increases the frequency of media use [ 38 ]. Moreover, family support has been identified as a crucial factor aiding older adults in overcoming barriers to the utilization of mHealth services [ 39 ]. Considering the substantial impact of family factors on the proactive health information-seeking behavior of older individuals [ 40 - 43 ], it becomes imperative to delve deeper into the relationship between family health, media use behavior, and the older individual’s intention to use mHealth devices. Additionally, exploring the mediating role of media use behavior between family health and the older individual’s intention to use mHealth devices is crucial. This comprehensive investigation aims to facilitate the integration of older individuals into the “digital age” starting from the family level, foster the adoption of mHealth in the health care sector, enhance societal healthy aging, and contribute to the realization of the objectives outlined in the “Healthy China 2030 Plan.”

In this study, information pertaining to family health, media use behavior, and the intention to use mHealth devices among older adults was gathered from the Psychology and Behavior Investigation of Chinese Residents (PBICR) study. The primary objective of this study was to examine the impact of family health and media use behavior on the intention of older individuals to use mHealth devices in China. Furthermore, the study aimed to assess whether media use behavior acts as a mediating factor in the relationship between family health and the intention to use mHealth devices among older adults. Drawing upon the insights gained from the literature review, the following hypotheses were formulated: (1) family health has a direct impact on the intention to use mHealth devices among older adults; (2) family health exerts an indirect influence on the intention to use mHealth devices through the mediating factor of media use behavior; in other words, media use behavior serves as a mediator in the relationship between family health and the intention to use mHealth devices.

Study Design and Setting

The data for this study were sourced from the PBICR survey, a comprehensive cross-sectional survey initiated by the Peking University School of Public Health in 2022. The survey encompasses 148 cities spanning 23 provinces, 5 autonomous regions, and 4 municipalities directly under the central government in China. Using a multistage sampling approach, the survey uses a stratified sampling method in cities, districts, counties, and communities, and uses a quota sampling method from the community level down to the individual level.

The survey was carried out by adeptly trained investigators. Electronic questionnaires (developed previously [ 44 ]) were distributed directly to the public through one-on-one, face-to-face interactions on-site. Respondents could access the questionnaire by scanning the provided QR code. In situations where face-to-face investigations were impeded due to the constraints of the COVID-19 epidemic, investigators distributed the electronic questionnaire on a one-on-one basis through instant communication tools such as WeChat (Tencent Holdings Ltd.). Additionally, online video investigations were conducted through platforms such as Tencent Meeting (Tencent Holdings Ltd.)and WeChat video [ 45 ].

Within the PBICR survey, investigators underwent comprehensive training in sampling methods, research tools, and quality control. Only those investigators who strictly adhered to the trained survey procedures were deemed qualified and eligible to participate in the study. Furthermore, during the data processing phase, 2 researchers were designated to perform logical checks. Questionnaires that did not meet the predetermined screening criteria were excluded, ensuring the quality and reliability of the data. Additionally, in this study, further screening was implemented to eliminate questionnaires completed in an excessively short time, those containing outliers, or those with missing values.

In the 2022 PBICR survey, a total of 23,414 questionnaires were collected. Following logical checks and the elimination of outliers, 21,916 questionnaires were deemed valid. For the purposes of this study, the focus will be confined to the age group of 60 years and above. Consequently, the final sample size included 3712 older adults after sorting.

Participants

A total of 21,916 questionnaires were collected, with the screening criterion being individuals aged 60 years and above, ensuring the absence of missing data and logic errors. Following a meticulous summary and screening process, 3712 valid survey responses were obtained for analysis in this study.

The inclusion criteria for participants in this study were as follows: (1) age between 18 and 60 years old; (2) possession of the nationality of the People’s Republic of China; (3) status as a Chinese permanent resident with an annual travel time of 1 month or less; (4) willing participation in the study and voluntary completion of the informed consent form; (5) ability to independently complete the questionnaire survey or do so with the assistance of investigators; (6) capacity to comprehend the meaning of each item in the questionnaire.

The exclusion criteria for participants in this study were as follows: (1) individuals with unconsciousness or mental disorders; (2) individuals with cognitive impairment; (3) those currently participating in other similar research projects; and (4) individuals unwilling to collaborate or reluctant to participate in the study.

Ethics Approval

The study adhered to the principles outlined in the Declaration of Helsinki. Ethical approval for all experimental protocols was granted by the ethics research committees of the Health Culture Research Center of Shaanxi (approval number JKWH-2022-02) and Second Xiangya Hospital of Central South University (approval number 2022-K050). The cover page of the questionnaire provided a clear explanation of the study’s purpose and assured participants of anonymity, confidentiality, and the right to refuse participation. Informed consent was obtained from all participants involved in the study.

The questionnaire cover used in this study provided a detailed explanation of the study’s purpose and ensured participants of anonymity, confidentiality, and the right to refuse participation. All participants were required to voluntarily sign an informed consent form before engaging in the study. While respondents did not directly benefit from the survey, their input contributed to a more comprehensive and systematic understanding of the physical and mental health status of the public. The data from this study will be strictly managed and used in accordance with the Statistics Law of the People’s Republic of China. The research data are intended for academic purposes only, and when the research findings are published, no information about individual participants will be disclosed or adversely affected.

Measurements

General situation survey information.

The basic demographic information of the older individuals included gender, age rank, nationality, religion, BMI rank, political status, status of occupation, education level, chronic diseases, and family type (conjugal family, core family, backbone family, and other family).

Family types were defined as follows:

  • Conjugal family: a family consisting of only husband and wife.
  • Core family: a family consisting of parents and unmarried children.
  • Backbone family: a family consisting of parents and married children.
  • Other family: other families including joint families, single-parent families, DINK (dual income, no kids) families, and single families.

Short-Form of the Family Health Scale

The assessment of family health in this study used the Chinese version of The Short-Form of the Family Health Scale (FHS-SF), developed by Crandall et al [ 20 ]. Wang et al [ 46 ] introduced the FHS-SF cross-culturally to create a Chinese version as a quantitative tool for evaluating family health issues in China. The scale comprises 10 items, encompassing 4 dimensions: family social and emotional health processes, family health lifestyle, family health resources, and family external social supports. A 5-point Likert scale was used for each item of the FHS-SF, with response options ranging from 1=strongly disagree to 5=strongly agree. Items with negative wording were scored in reverse. The final score on the scale ranged from 10 to 50, where higher scores indicated higher levels of family health. Wang et al [ 46 ] reported that the Cronbach α for the FHS-SF was .83. Additionally, the Cronbach α for the 4 subscales ranged from .70 to .90, and the retest reliability of the scale was 0.75.

In our study, the composite reliability values for the 4 dimensions were 0.912, 0.848, 0.781, and 0.806, respectively. All these values surpass the reliability threshold of 0.7. The average variance extracted values for the dimensions were 0.775, 0.736, 0.553, and 0.677, respectively, all of which exceed the threshold of 0.5. The Cronbach α of the FHS-SF was .90, and the factor loadings ranged from 0.73 to 0.90, all within an acceptable range.

Media Use Behavior Scale

The frequency of media use in this study was gauged using the Media Use Behavior Scale developed by the PBICR survey of Peking University. The scale encompasses various media channels such as newspapers, radio, television, the internet, and mobile phones. Comprising 6 items related to social contact, self-presentation, social behavior, leisure and entertainment, access to information, and business transactions, the scale uses options that signify the degree of media use frequency, ranging from “1=infrequent” to “5=frequent.” The total score on the scale ranges from 6 to 30, with higher scores indicative of more frequent use of the media [ 45 ].

In this study, the composite reliability for the Media Use Behavior Scale was 0.894, and the average variance extracted was 0.585. The Cronbach α for the Media Use Behavior Scale was .89, indicating strong internal consistency. Additionally, the standardized factor loadings obtained from the validation factor analysis were above 0.50, all falling within acceptable limits.

Intention to Use mHealth Devices

The intention to use mHealth devices in this study was assessed through subjective evaluations. Participants were required to provide a numerical response ranging from 0 to 100 based on their individual subjective awareness. This formed a continuous variable, where a higher numerical value indicated a stronger intention to use mHealth devices.

Data Analysis

Continuous variables were assessed for normality using the Kolmogorov-Smirnov test and presented as the median and IQR. Categorical variables were reported in terms of frequency and percentage. Nonparametric methods were used to test the differences in characteristics related to the total score of the intention to use mHealth devices. Specifically, the Mann-Whitney U test was used for dichotomous variables, while the Kruskal-Wallis H test was used for multicategorical variables. The partial correlation coefficient between family health scores, media use behavior scores, and intention to use mHealth devices scores was calculated using a regression model. Linear regression models were used to assess the association between family health scores and media use behavior/intention to use mHealth devices scores, both with and without adjustment for covariates. The associations between media use behavior and intention to use mHealth devices scores were also examined. The results are reported as coefficients along with 95% CIs. Covariates, determined based on previous studies and general knowledge, were included in the models for adjustment. To examine the mediating role of media use behavior scores in the association between family health scores and intention to use mHealth devices scores, we conducted a Sobel-Goodman Mediation Test. This analysis was performed while controlling for all selected covariates. The significance of the indirect effect, direct effect, and the total effect was determined using the bootstrap algorithm.

All P values were 2-sided, with a significance level (α) of .05 used to define statistical significance. The data were analyzed using IBM SPSS Statistics 26 and R version 4.1.3 (R Foundation).

Subgroup Analysis

Indeed, empirical studies have consistently indicated a positive association between education and health. Individuals with higher levels of education often exhibit a tendency to adopt healthier lifestyles, and their increased income may lead to greater investment in health-related expenses [ 47 ]. Furthermore, education is closely linked to varying levels of internet participation. Generally, individuals with higher educational attainment are more likely to use online platforms for accessing health-related information [ 48 ]. In diverse educational and cultural backgrounds, individuals may exhibit varying levels of concern regarding health risks, subsequently influencing their acceptance of health care technology [ 49 ]. Additionally, preliminary analysis in our study revealed significant differences in the total score of family health across different education levels ( P <.001). Building on the established influence of education on health behavior and media use, as outlined in the existing literature and supported by our results, this paper intends to analyze education level as a subgroup. The aim is to comprehensively explore the mediating role of media use behavior among older adults with different education levels in the relationship between family health and their intention to use mHealth devices.

General Characteristics

A total of 3712 older individuals aged 60 and above participated in this study, with an average age of 69.23 (SD 6.13) years. The majority of older adults (3036/3712, 81.79%) fell within the age range of 60-74 years. Basic demographic data for the 3712 older adult participants are detailed in Table 1 . Among them, 1839 were males (49.54%) and 1873 were females (50.46%). The majority identified as Han nationality (3370/3712, 90.79%) and nonreligious (3416/3712, 92.03%), with the majority expressing mass political views (3151/3712, 84.89%). There were noteworthy differences in the willingness to use mHealth devices among older adults with varying political statuses, occupational statuses, and chronic disease conditions ( P <.001). However, no significant differences were observed in the willingness to use mHealth devices among older adults with different family types ( P =.97; Table 1 ).

a Median (IQR) was used to describe the continuous variable, whereas n (%) was used to describe the categorical variable.

Association Analysis

After adjusting for covariates, the intention to use mHealth devices exhibited a positive correlation with the total score of family health ( r =0.077, P <.001) and the media use behavior score ( r =0.178, P <.001). Additionally, the total score of family health was positively correlated with the media use behavior score ( r =0.079, P <.001; Table 2 ).

a The model was adjusted for various covariates, including religion, BMI rank, political status, occupational status, education degree, and chronic diseases. Variables achieved statistical significance at P ≤.05.

b N/A: not applicable.

Relationship Between Family Health and Media Use Behavior Score/Intention to Use mHealth Devices

In the linear regression models before adjustment, the 4 dimensions of family health (ie, family socialization, family healthy lifestyle, family health resources, and family external social support) and the total score were significantly ( P <.001) associated with media use behavior. Moreover, they were significantly ( P <.001) related to the intention to use mHealth devices, except for family health resources ( P= .15). After adjusting for gender and age rank, as well as political status, nationality, religion, BMI rank, occupation status, education level, family type, and chronic diseases, all dimensions remained statistically significant ( P <.001) except for family health resources ( P= .29; Table 3 ).

a Data were adjusted for gender and age rank, political status, nation, religion, BMI rank, status of occupation, education degree, family type, and chronic diseases.

Relationship Between Media Use Behavior Score and Intention to Use mHealth Devices

In the linear regression models before adjustment, media use behavior was significantly ( P <.001) associated with the intention to use mHealth devices. After adjusting for gender and age rank, as well as political status, nationality, religion, BMI rank, occupation status, education level, family type, and chronic diseases, the association remained statistically significant ( P <.001; Table 4 ).

Mediation Analysis

The family health total score demonstrated a positive association with the intention to use mHealth devices among older adults. Mediation analysis, including media use behavior, revealed that the relationship between the total score of family health and the intention to use mHealth devices was mediated through media use behavior. In this study, media use behavior partially mediated the association between family health and the intention to use mHealth devices. The mediating variable accounted for nearly a quarter (22.46/100) of the association when adjusting for covariates. The total score of family health was associated with media use behavior (β=.088, P <.001) and intention to use mHealth devices (β=.244, P <.001). Additionally, media use behavior was linked to the intention to use mHealth devices (β=.810, P <.001). The final mediation models depicting the independent variable (total score of family health), the mediating variable (media usage behavior), and the dependent variable (intention to use mHealth devices) are illustrated in Figure 1 .

research papers in medical domain

The 4 dimensions of family health were positively associated with the use of mHealth devices among older adults, except for the dimension of family health resources, which had a nonsignificant association ( P= .72). The mediation analysis involving media use behavior indicated that the direct and total effects of family health resources were not significant ( P =.72 and P =.20, respectively). Media use behavior acted as a full mediator when adjusting for covariates. Media use behavior partially mediated the relationship between family social, family healthy lifestyle, family external social support, and the intention to use mHealth devices, with mediating effects of 35.18/100, 31.78/100, and 31.33/100, respectively, under adjusted covariates ( Table 5 ).

a The Sobel-Goodman Mediation Test was applied in adjusted models for religion, BMI rank, political status, occupation status, education level, and chronic diseases.

b The Sobel test was used to assess the hypothesis that the indirect role was equal to 0, adjusting for covariates such as religion, BMI rank, political status, occupation status, education level, and chronic diseases. Values reach statistical significance at P ≤.05.

Subgroup analyses based on education degrees are presented in Table 6 . Among the older adult population with primary school education and below, media use behavior showed no mediating effect between the total score of family health and the intention to use mHealth devices ( z =–0.942; indirect effect=–0.019, P =.35; direct effect=0.252, P =.007). Additionally, the mediating effect of media use behavior between family healthy lifestyles and the intention to use mHealth devices was not significant ( z =1.953, P =.052). Media use behavior fully mediated the association between family health resources scores and intention to use mHealth devices scores in different education degrees among the older adult population: primary school and below degree older adult population ( z =–5.832; indirect effect=–0.331, P <.001; direct effect=0.218, P= .29), middle school/vocational school/high school degree older adult population ( z =–3.439; indirect effect=–0.136, P <.001; direct effect=–0.066, P =.76), and college and above degree older adult population ( z =–2.516; indirect effect=–0.212, P= .01; direct effect=0.026, P =.93).

a The Sobel-Goodman Mediation Test was applied in adjusted models for religion, BMI rank, political status, status of occupation, and chronic diseases.

Principal Findings

Previous studies have consistently demonstrated that family factors play a crucial role in influencing the frequency of media use and the acceptance of mHealth among older adults [ 50 ]. The findings of our study further confirm that family health positively contributes to increasing the willingness of older adults to use mHealth devices. Additionally, a high frequency of media use behavior emerges as a significant driver for the utilization of mHealth devices, a behavior that is profoundly influenced by the state of family health. The results align with previous research on the digital divide among older adults, indicating that those with higher family health scores tend to engage in more frequent media contact behaviors. This heightened connectivity to the internet makes them more adaptable to a big data–based mHealth environment, fostering a greater willingness to use mHealth devices. Before conducting the mediation analysis, the study also observed, through univariate analysis, that older individuals over 90 years and those who were unemployed exhibited a lower willingness to use mobile medical devices. The results confirm the existence of differences in the digital divide among age groups, especially with older age groups experiencing inequalities in social and economic support [ 51 , 52 ]. These disparities may further impact their access to and utilization of media devices.

In addition to the descriptive findings, this study delves into the intricate relationship between family health and the willingness to use mHealth devices, uncovering the mediating role of media use behavior. Primarily, the study supports the positive impact of media use behavior, which partially mediates the influence of overall family health levels on the intention to use mHealth devices. Furthermore, the results indicate that media use behavior serves as a fully mediating variable in the dimension of family health resources. In essence, the findings suggest that older adults lacking family health resources completely lose their willingness to use mHealth devices, primarily due to their challenges in accessing or using media. This underscores the crucial role of family health resources in integrating older adults into the internet sphere and enabling them to benefit from mHealth technology. The study emphasizes the practical importance of addressing resource-related health inequities, with financial support from the family being identified as a critical factor in the daily lives of seniors [ 52 ]. To address the imbalance in the distribution of resources among families in different regions at the societal level, it is crucial for the government to assist socioeconomically disadvantaged older adults in gaining greater access to various devices. This can be achieved through economic empowerment initiatives and the development of policies aimed at bridging the digital divide [ 53 ].

Building upon the crucial role of media contacts in linking family health resources and the willingness to use mHealth devices among the older population, there is an opportunity to further motivate the desire for mHealth device usage. Leveraging the positive influence of family health resources to increase the frequency of media exposure can enhance the motivation of older individuals. Effective communication within the family emerges as a catalyst for improving the technology literacy and information-seeking skills of older adults [ 16 ]. Family members play a crucial role in supporting seniors to build confidence in using internet technology while alleviating their anxiety and fear of new technologies. Encouraging older adults to adapt and learn information technology, such as WeChat and health-related mobile apps, through straightforward and repeated demonstrations can be an effective strategy [ 54 ]. Additionally, family support may help mitigate the economic challenges associated with using health care services by influencing older adults’ subjective perceptions of financial accessibility [ 55 ]. To address financial challenges and enhance older adults’ access to technology, a comprehensive approach can be adopted. This involves leveraging both the financial support within the family and external economic resources. Encouraging family members to provide suitable financial assistance to each other, coupled with ensuring stable financial security for older individuals, can be achieved by gradually increasing pensions for retirees. This approach aims to augment the purchasing power of older adults, enabling them to acquire media devices and enhancing their ability to use technological devices in the health care sector to a greater extent.

The subgroup analysis further indicated that media use behavior did not mediate the relationship between the total family health score and the intention to use mHealth devices among older adults with primary school education or below. However, it did partially mediate the association among those with primary school education and above, aligning with the study hypothesis. Given that the older adult population with low education levels may experience relatively weak cognitive function and lack personal health literacy [ 56 , 57 ], the mechanisms by which they are influenced by family, social, and economic environments in the acceptance of new health technologies become more intricate. Conversely, older adults with a high school education or higher often perceive themselves as having an above-average ability to learn, making them less uncomfortable with the changing social environment brought about by technological developments [ 58 ]. Moreover, older individuals with limited education often lack access to information technology education or the ability to operate mobile devices [ 59 ]. For these individuals, exposure to media devices or mHealth devices is relatively homogeneous. Consequently, they may lack a progressive transition from regular media contact behaviors to the use of mHealth devices.

Disparities in internet participation levels due to education constitute a significant barrier hindering older adults from using media devices to access the mHealth era. To bridge the “digital divide” and enhance the effective use of mHealth devices among older individuals, it is imperative to consider implementing relevant education measures. These measures can focus on improving their ability to use smart technology, thus empowering them to navigate and benefit from the advancements in health care technology. In alignment with the comprehensive “Smart Senior Care” action plan in China [ 60 ], communities can implement health education initiatives through a blend of technology-supported learning and traditional lectures. For instance, using touchscreen tablets for courses on healthy diet and nutrition guidance can enhance the older individual’s interest in the internet while imparting essential health and hygiene knowledge [ 61 ]. This approach serves to bridge the transition from traditional modes of access to mobile health care. Adopting adaptive behaviors and learning strategies can further enhance the efficiency and effectiveness of mobile health care apps [ 62 ]. In the mHealth era, the design of mHealth devices should be tailored to the cognitive abilities and mindset of older individuals. Full consideration should be given to their eHealth literacy, incorporating improvements in usability, emphasizing the responsiveness of operations, and integrating monitoring functions that align with the physical activities of older individuals [ 63 ]. Such considerations aim to enhance the overall satisfaction of older individuals with mobile health care apps [ 64 ]. Moreover, due to prevailing stereotypes about older people, digital platforms often harbor ageist mechanisms that categorize them as users uninterested in technology [ 65 ]. This results in an unfavorable digital environment for older individuals. In general, the development and application of internet technology must not overlook the realistic capacity and objective demands of older individuals [ 66 ]. Digital platforms should strive to create more inclusive algorithms and use statistical models of social digital media practices that cater to all literacy levels [ 65 ]. This may involve reducing complex and lengthy text that is difficult to understand, avoiding in-depth and complex hierarchical options, and adopting simple page designs [ 67 ] to mitigate the impact of technological differences on the accessibility of digital health care for older adults.

Strength and Limitations

This study contributes significantly to the existing literature by evaluating the connection between family health, media use behavior, and the intention to use mHealth devices among older adults, using cross-sectional data from the PBICR survey. The findings of this study support our hypothesis that media use behavior serves as a mediator between family health status and the intention to use mHealth devices among older adults. Furthermore, a subgroup analysis based on education level revealed that the impact of family health on the willingness to use mHealth devices through media use behavior was not significant among older adults with lower education levels, indicating a nuanced mechanism at play. All of the aforementioned studies contribute to the body of research on the digital divide among older individuals.

Despite comprehensive consideration, the results of this study have several limitations. First, due to the exploratory cross-sectional design, no causal inferences can be drawn. Second, the majority of seniors included in this study were in the young-old age group (60 to 74 years old), lacking representation of the entire age spectrum of older adults and potentially neglecting variations in social background associated with age factors. Third, the results obtained in this study may be influenced by economic factors and psychological variables. As mHealth devices represent an evolving component of the health system, their development trajectory is still undergoing exploration. It is possible that various latent factors influencing the relationship between family health, media use behavior, and the intention to use mHealth devices are yet to be uncovered.

Conclusions

In conclusion, this study highlights the substantial impact of family health and media use behavior on the intention of older adults to use mHealth devices. Media use behavior acts as a mediator in the relationship between family health and the intention to use mHealth devices, with more intricate dynamics observed among older adults with lower educational levels. These findings emphasize that robust family health, particularly sufficient family health resources, plays a crucial role in enhancing the media engagement of older individuals, ultimately fostering their interest in embracing mHealth devices. The insights from this work provide valuable recommendations for bridging the gap in digital health adoption among older adults. Furthermore, encouraging teaching by family members can create a supportive environment for seniors to embrace mobile technology, while financial support can enhance their accessibility to health-related mobile devices. Additionally, developing age-specific digital education programs and media products tailored to the needs and preferences of older individuals can contribute to overcoming technological barriers and fostering a positive digital experience for older adults in the realm of mobile health care. These strategies align with the goal of promoting inclusive and user-friendly digital solutions for seniors, ensuring they can benefit from advancements in health technology.

Acknowledgments

This study was conducted with the support of data from the Psychology and Behavior Investigation of Chinese Residents (PBICR). We appreciate all the participants who showed great patience in answering the questionnaires. None of the portions of this article used generative artificial intelligence. This work was supported by the 2023 Guangdong Province Education Science Planning Project (Specialized in Higher Education; 2023GXJK252), the Science and Technology Program of Guangzhou (grant numbers 2023A04J2267 and 2024A04J02668), the Guangdong Basic and Applied Basic Research Foundation (grant number 2021A1515110743), the Health Economics Association of Guangdong Province (grant number 2023-WJMZ-51), the Student Innovation and Entrepreneurship Training Program of Guangdong Province (grant number S202312121283), the Key Laboratory of Philosophy and Social Sciences of Guangdong Higher Education Institutions for Health Policies Research and Evaluation (grant number 2015WSY0010), and the Research Base for Development of Public Health Service System of Guangzhou.

Data Availability

The data sets generated and analyzed during this study are not publicly available because the data still need to be used for other research but are available from the corresponding author on reasonable request.

Authors' Contributions

JHC, YBW, and JYC designed and conducted this study. YBW collected data. YSM, AQL, and XXY participated in the data screening. DYZ and WDY conducted data analysis. JHC and YSM wrote the first draft of the paper. JYC contributed to supervising data analysis and developing the manuscript. All authors made contributions to the critical revision of the manuscript. The authors read and approved the final manuscript.

Conflicts of Interest

None declared.

  • World population porospects 2022: summary of results. United Nations Department of Economic and Social Affairs, Population Division. 2022. URL: https:/​/www.​un.org/​development/​desa/​pd/​sites/​www.un.org.development.desa.pd/​files/​wpp2022_summary_of_results.​pdf [accessed 2022-07-26]
  • Communique of the Seventh National Population Census (No.5). National Bureau of Statistics. May 11, 2021. URL: https://www.stats.gov.cn/sj/zxfb/202302/t20230203_1901085.html [accessed 2021-05-11]
  • Guiding opinions on establishing and improving the health service system for the elderly. National Health Commission/State Council Information Office. Nov 01, 2019. URL: http://www.scio.gov.cn/xwfb/bwxwfb/gbwfbh/wsjkwyh/202307/t20230703_721062.html [accessed 2024-02-11]
  • Wang X, Wu Y, Meng Z, Li J, Xu L, Sun X, et al. Willingness to use mobile health devices in the post-COVID-19 era: nationwide cross-sectional study in China. J Med Internet Res. Feb 17, 2023;25:e44225. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Tran V, Riveros C, Ravaud P. Patients' views of wearable devices and AI in healthcare: findings from the ComPaRe e-cohort. NPJ Digit Med. 2019;2:53. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Piwek L, Ellis DA, Andrews S, Joinson A. The rise of consumer health wearables: promises and barriers. PLoS Med. Feb 2016;13(2):e1001953. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Kekade S, Hseieh C, Islam MM, Atique S, Mohammed Khalfan A, Li Y, et al. The usefulness and actual use of wearable devices among the elderly population. Comput Methods Programs Biomed. Jan 2018;153:137-159. [ CrossRef ] [ Medline ]
  • Li H, Zhang T, Chi H, Chen Y, Li Y, Wang J. Mobile health in China: current status and future development. Asian J Psychiatr. Aug 2014;10:101-104. [ CrossRef ] [ Medline ]
  • Choi N. Relationship between health service use and health information technology use among older adults: analysis of the US National Health Interview Survey. J Med Internet Res. Apr 20, 2011;13(2):e33. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Mo XT, Deng ZH. An empirical study on mobile health service acceptance behavior of middle-aged and elderly users in Wuhan. Chinese Journal of Health Statistics. 2015;32(02):324-327.
  • Yang JH, Liu YF. Longevity bonus of digital era: the capability and endogenous motivation of old people's digital life. Administration Reform. 2022;1(1):26-36. [ CrossRef ]
  • Lu JH, Wei XD. Analysis framework, concept, and pathways of digital divide governance for older adults: from the perspective of digital divide and knowledge gap theory. Population Research. 2021;45(03):17-30.
  • Huang CX. Status, challenges and countermeasures of the digital divide in older adults. People's Tribune. 2020;29:126-128.
  • The 50th Statistical Report on Internet Development in China. China Internet Network Information Center(CNNIC). 2022. URL: https://www.cnnic.net.cn/n4/2022/0914/c88-10226.html [accessed 2022-08-31]
  • Yang B, Jin DC. The manifestation, motivation and solution of elderly digital divide. Academic Journal of Zhongzhou. 2021(12):74-80.
  • Magsamen-Conrad K, Dillon JM, Billotte Verhoff C, Faulkner SL. Online health-information seeking among older populations: family influences and the role of the medical professional. Health Commun. Jul 2019;34(8):859-871. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Kim S, Sok SR. Relationships among the perceived health status, family support and life satisfaction of older Korean adults. Int J Nurs Pract. Aug 2012;18(4):325-331. [ CrossRef ] [ Medline ]
  • Phillips DR, Feng Z. Challenges for the aging family in the People's Republic of China. Can J Aging. Sep 2015;34(3):290-304. [ CrossRef ] [ Medline ]
  • Weiss-Laxer NS, Crandall A, Okano L, Riley AW. Building a foundation for family health measurement in national surveys: a modified Delphi expert process. Matern Child Health J. Mar 2020;24(3):259-266. [ CrossRef ] [ Medline ]
  • Crandall A, Weiss-Laxer NS, Broadbent E, Holmes EK, Magnusson BM, Okano L, et al. The Family Health Scale: reliability and validity of a short- and long-form. Front Public Health. 2020;8:587125. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Yuan B, Zhang T, Li J. Family support and transport cost: understanding health service among older people from the perspective of social-ecological model. Arch Public Health. Jul 19, 2022;80(1):173. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Leung K, Chen C, Lue B, Hsu S. Social support and family functioning on psychological symptoms in elderly Chinese. Arch Gerontol Geriatr. 2007;44(2):203-213. [ CrossRef ] [ Medline ]
  • Zhang Z, Mao YH, Hu YC. A study on willingness to use intelligent elderly care services from the perspective of elderly digital divide. Northwest Population Journal. 2023;21:1-12.
  • Molina-Mula J, Gallo-Estrada J, González-Trujillo A. Self-perceptions and behavior of older people living alone. Int J Environ Res Public Health. Nov 24, 2020;17(23):8739. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Nguyen T, Irizarry C, Garrett R, Downing A. Access to mobile communications by older people. Australas J Ageing. Jun 2015;34(2):E7-E12. [ CrossRef ] [ Medline ]
  • Bisschop MI, Kriegsman DMW, van Tilburg TG, Penninx BWJH, van Eijk JTM, Deeg DJH. The influence of differing social ties on decline in physical functioning among older people with and without chronic diseases: the Longitudinal Aging Study Amsterdam. Aging Clin Exp Res. Apr 2003;15(2):164-173. [ CrossRef ] [ Medline ]
  • Martínez-Pérez B, de la Torre-Díez I, López-Coronado M. Mobile health applications for the most prevalent conditions by the World Health Organization: review and analysis. J Med Internet Res. Jun 14, 2013;15(6):e120. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Li C, Neugroschl J, Zhu CW, Aloysi A, Schimming CA, Cai D, et al. Design considerations for mobile health applications targeting older adults. J Alzheimers Dis. 2021;79(1):1-8. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Devos P, Min Jou A, De Waele G, Petrovic M. Design for personalized mobile health applications for enhanced older people participation. European Geriatric Medicine. Dec 2015;6(6):593-597. [ CrossRef ]
  • Wang H, Sun X, Wang R, Yang Y, Wang Y. The impact of media use on disparities in physical and mental health among the older people: an empirical analysis from China. Front Public Health. 2022;10:949062. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Zhang K, Kim K, Silverstein NM, Song Q, Burr JA. Social media communication and loneliness among older adults: the mediating roles of social support and social contact. Gerontologist. Aug 13, 2021;61(6):888-896. [ CrossRef ] [ Medline ]
  • Ma Y, Liang C, Gu D, Zhao S, Yang X, Wang X. How social media use at work affects improvement of older people's willingness to delay retirement during transfer from demographic bonus to health bonus: causal relationship empirical study. J Med Internet Res. Feb 10, 2021;23(2):e18264. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Zhang C. Smartphones and telemedicine for older people in China: opportunities and challenges. Digit Health. 2022;8:20552076221133695. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Lee SM, Lee D. Healthcare wearable devices: an analysis of key factors for continuous use intention. Serv Bus. Oct 15, 2020;14(4):503-531. [ CrossRef ]
  • Krebs P, Duncan DT. Health app use among US mobile phone owners: a national survey. JMIR Mhealth Uhealth. Nov 04, 2015;3(4):e101. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Li J, Ma Q, Chan AH, Man S. Health monitoring through wearable technologies for older adults: smart wearables acceptance model. Appl Ergon. Feb 2019;75:162-169. [ CrossRef ] [ Medline ]
  • Stühmann LM, Paprott R, Heidemann C, Baumert J, Hansen S, Zahn D, et al. Health app use and its correlates among individuals with and without type 2 diabetes: nationwide population-based survey. JMIR Diabetes. May 20, 2020;5(2):e14396. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Li Y, Han W, Hu M. Does internet access make a difference for older adults' cognition in urban China? The moderating role of living arrangements. Health Soc Care Community. Jul 2022;30(4):e909-e920. [ CrossRef ] [ Medline ]
  • Mizrachi Y, Shahrabani S, Nachmani M, Hornik A. Obstacles to using online health services among adults age 50 and up and the role of family support in overcoming them. Isr J Health Policy Res. Aug 21, 2020;9(1):42. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Tu J, Shen M, Zhong J, Yuan G, Chen M. The perceptions and experiences of mobile health technology by older people in Guangzhou, China: a qualitative study. Front Public Health. 2021;9:683712. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Navabi N, Ghaffari F, Jannat-Alipoor Z. Older adults' attitudes and barriers toward the use of mobile phones. Clin Interv Aging. 2016;11:1371-1378. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Harris T, Cook DG, Victor CR, Beighton C, Dewilde S, Carey IM. Linking survey data with computerised records to predict consulting by older people. Br J Gen Pract. Dec 2004;54(509):928-931. [ FREE Full text ] [ Medline ]
  • Gao M, Li Y, Zhang S, Gu L, Zhang J, Li Z, et al. Does an empty nest affect elders' health? Empirical evidence from China. Int J Environ Res Public Health. Apr 27, 2017;14(5):463. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Wenjuanxing. URL: https://www.wjx.cn/ [accessed 2024-02-07]
  • Wang Y, Kaierdebieke A, Fan S, Zhang R, Huang M, Li H, et al. Study protocol: a cross-sectional study on psychology and behavior investigation of Chinese residents, PBICR. Psychosom Med Res. 2022;4(3):19. [ CrossRef ]
  • Wang F, Wu Y, Sun X, Wang D, Ming W, Sun X, et al. Reliability and validity of the Chinese version of a short form of the family health scale. BMC Prim Care. May 06, 2022;23(1):108. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Wang W, Dong Y, Liu X, Zhang L, Bai Y, Hagist S. The more educated, the healthier: evidence from rural China. Int J Environ Res Public Health. Dec 13, 2018;15(12):2848. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Oh YS, Choi EY, Kim YS. Predictors of smartphone uses for health information seeking in the Korean elderly. Soc Work Public Health. 2018;33(1):43-54. [ CrossRef ] [ Medline ]
  • Nadal C, Sas C, Doherty G. Technology acceptance in mobile health: scoping review of definitions, models, and measurement. J Med Internet Res. Jul 06, 2020;22(7):e17256. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Chew TH, Chin CP, Leau Y. Untangling factors influencing social networking sites use among older adults: a literature review. Univ Access Inf Soc. Mar 17, 2022;22(3):687-698. [ CrossRef ]
  • Hunsaker A, Hargittai E. A review of internet use among older adults. New Media & Society. Jul 16, 2018;20(10):3937-3954. [ CrossRef ]
  • García MF, Hessel P, Rodríguez-Lesmes P. Wealth and inequality gradients for the detection and control of hypertension in older individuals in middle-income economies around 2007-2015. PLoS One. 2022;17(7):e0269118. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Guo W, Chen L, Perez C. Economic status, family dependence, and health outcomes of older people in western rural China. J Gerontol Soc Work. Oct 2019;62(7):762-775. [ CrossRef ] [ Medline ]
  • Shi Y, Ma D, Zhang J, Chen B. In the digital age: a systematic literature review of the e-health literacy and influencing factors among Chinese older adults. Z Gesundh Wiss. 2023;31(5):679-687. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Di X, Wang L, Yang L, Dai X. Impact of economic accessibility on realized utilization of home-based healthcare services for the older adults in China. Healthcare (Basel). Feb 17, 2021;9(2):218. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Liu H, Byles JE, Xu X, Zhang M, Wu X, Hall JJ. Evaluation of successful aging among older people in China: results from China health and retirement longitudinal study. Geriatr Gerontol Int. Aug 2017;17(8):1183-1190. [ CrossRef ] [ Medline ]
  • Quenzel G, Vogt D, Schaeffer D. Differences in health literacy of adolescents with lower educational attainment, older people and migrants. Gesundheitswesen. Nov 2016;78(11):708-710. [ CrossRef ] [ Medline ]
  • Berkowsky RW, Sharit J, Czaja SJ. Factors predicting decisions about technology adoption among older adults. Innov Aging. Jan 2018;2(1):igy002. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Jun W. A study on cause analysis of digital divide among older people in Korea. Int J Environ Res Public Health. Aug 14, 2021;18(16):8586. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • National Health Commission; National Council on the Aging. Notice on the in-depth implementation of the “Smart Help for the Elderly” Action in 2022. National Health Commission. 2022. URL: http://www.nhc.gov.cn/lljks/zcwj2/202206/24a5b60b8789409c9053b38e4aab19e7.shtml [accessed 2022-06-16]
  • Chiu C, Kuo S, Lin D. Technology-embedded health education on nutrition for middle-aged and older adults living in the community. Glob Health Promot. Sep 2019;26(3):80-87. [ CrossRef ] [ Medline ]
  • Yu Y, Yan XD, Z X, Zhou SL. What they gain depends on what they do: an exploratory empirical research on effective use of mobile healthcare applications. Presented at: Hawaii International Conference on System Sciences; January 8-11, 2019, 2019; Maui, HI. URL: https://aisel.aisnet.org/cgi/viewcontent.cgi?article=1477&context=hicss-52 [ CrossRef ]
  • Ma Z, Gao Q, Yang M. Adoption of wearable devices by older people: changes in use behaviors and user experiences. International Journal of Human–Computer Interaction. Aug 31, 2022;39(5):964-987. [ CrossRef ]
  • Zhang X, Yan X, Cao X, Sun Y, Chen H, She J. The role of perceived e-health literacy in users’ continuance intention to use mobile healthcare applications: an exploratory empirical study in China. Information Technology for Development. Mar 09, 2017;24(2):198-223. [ CrossRef ]
  • Rosales A, Fernández-Ardèvol M. Ageism in the era of digital platforms. Convergence (Lond). Dec 2020;26(5-6):1074-1087. [ FREE Full text ] [ CrossRef ] [ Medline ]
  • Zhou X, Chen L. Digital health care in China and access for older people. The Lancet Public Health. Dec 2021;6(12):e873-e874. [ CrossRef ]
  • Gao Q, Ebert D, Chen X, Ding Y. Design of a mobile social community platform for older Chinese people in Urban areas. Hum Factors Man. Jun 27, 2012;25(1):66-89. [ CrossRef ]

Abbreviations

Edited by T de Azevedo Cardoso; submitted 18.06.23; peer-reviewed by R Sun, X Zhang; comments to author 08.08.23; revised version received 29.08.23; accepted 28.01.24; published 19.02.24.

©Jinghui Chang, Yanshan Mai, Dayi Zhang, Xixi Yang, Anqi Li, Wende Yan, Yibo Wu, Jiangyun Chen. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 19.02.2024.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.

  • Share full article

research papers in medical domain

A Columbia Surgeon’s Study Was Pulled. He Kept Publishing Flawed Data.

The quiet withdrawal of a 2021 cancer study by Dr. Sam Yoon highlights scientific publishers’ lack of transparency around data problems.

Supported by

Benjamin Mueller

By Benjamin Mueller

Benjamin Mueller covers medical science and has reported on several research scandals.

  • Feb. 15, 2024

The stomach cancer study was shot through with suspicious data. Identical constellations of cells were said to depict separate experiments on wholly different biological lineages. Photos of tumor-stricken mice, used to show that a drug reduced cancer growth, had been featured in two previous papers describing other treatments.

Problems with the study were severe enough that its publisher, after finding that the paper violated ethics guidelines, formally withdrew it within a few months of its publication in 2021. The study was then wiped from the internet, leaving behind a barren web page that said nothing about the reasons for its removal.

As it turned out, the flawed study was part of a pattern. Since 2008, two of its authors — Dr. Sam S. Yoon, chief of a cancer surgery division at Columbia University’s medical center, and a more junior cancer biologist — have collaborated with a rotating cast of researchers on a combined 26 articles that a British scientific sleuth has publicly flagged for containing suspect data. A medical journal retracted one of them this month after inquiries from The New York Times.

A person walks across a covered walkway connecting two buildings over a road with parked cars. A large, blue sign on the walkway says "Columbia University Irving Medical Center."

Memorial Sloan Kettering Cancer Center, where Dr. Yoon worked when much of the research was done, is now investigating the studies. Columbia’s medical center declined to comment on specific allegations, saying only that it reviews “any concerns about scientific integrity brought to our attention.”

Dr. Yoon, who has said his research could lead to better cancer treatments , did not answer repeated questions. Attempts to speak to the other researcher, Changhwan Yoon, an associate research scientist at Columbia, were also unsuccessful.

The allegations were aired in recent months in online comments on a science forum and in a blog post by Sholto David, an independent molecular biologist. He has ferreted out problems in a raft of high-profile cancer research , including dozens of papers at a Harvard cancer center that were subsequently referred for retractions or corrections.

From his flat in Wales , Dr. David pores over published images of cells, tumors and mice in his spare time and then reports slip-ups, trying to close the gap between people’s regard for academic research and the sometimes shoddier realities of the profession.

When evaluating scientific images, it is difficult to distinguish sloppy copy-and-paste errors from deliberate doctoring of data. Two other imaging experts who reviewed the allegations at the request of The Times said some of the discrepancies identified by Dr. David bore signs of manipulation, like flipped, rotated or seemingly digitally altered images.

Armed with A.I.-powered detection tools, scientists and bloggers have recently exposed a growing body of such questionable research, like the faulty papers at Harvard’s Dana-Farber Cancer Institute and studies by Stanford’s president that led to his resignation last year.

But those high-profile cases were merely the tip of the iceberg, experts said. A deeper pool of unreliable research has gone unaddressed for years, shielded in part by powerful scientific publishers driven to put out huge volumes of studies while avoiding the reputational damage of retracting them publicly.

The quiet removal of the 2021 stomach cancer study from Dr. Yoon’s lab, a copy of which was reviewed by The Times, illustrates how that system of scientific publishing has helped enable faulty research, experts said. In some cases, critical medical fields have remained seeded with erroneous studies.

“The journals do the bare minimum,” said Elisabeth Bik, a microbiologist and image expert who described Dr. Yoon’s papers as showing a worrisome pattern of copied or doctored data. “There’s no oversight.”

Memorial Sloan Kettering, where portions of the stomach cancer research were done, said no one — not the journal nor the researchers — had ever told administrators that the paper was withdrawn or why it had been. The study said it was supported in part by federal funding given to the cancer center.

Dr. Yoon, a stomach cancer specialist and a proponent of robotic surgery, kept climbing the academic ranks, bringing his junior researcher along with him. In September 2021, around the time the study was published, he joined Columbia, which celebrated his prolific research output in a news release . His work was financed in part by half a million dollars in federal research money that year, adding to a career haul of nearly $5 million in federal funds.

The decision by the stomach cancer study’s publisher, Elsevier, not to post an explanation for the paper’s removal made it less likely that the episode would draw public attention or affect the duo’s work. That very study continued to be cited in papers by other scientists .

And as recently as last year, Dr. Yoon’s lab published more studies containing identical images that were said to depict separate experiments, according to Dr. David’s analyses.

The researchers’ suspicious publications stretch back 16 years. Over time, relatively minor image copies in papers by Dr. Yoon gave way to more serious discrepancies in studies he collaborated on with Changhwan Yoon, Dr. David said. The pair, who are not related, began publishing articles together around 2013.

But neither their employers nor their publishers seemed to start investigating their work until this past fall, when Dr. David published his initial findings on For Better Science, a blog, and notified Memorial Sloan Kettering, Columbia and the journals. Memorial Sloan Kettering said it began its investigation then.

None of those flagged studies was retracted until last week. Three days after The Times asked publishers about the allegations, the journal Oncotarget retracted a 2016 study on combating certain pernicious cancers. In a retraction notice , the journal said the authors’ explanations for copied images “were deemed unacceptable.”

The belated action was symptomatic of what experts described as a broken system for policing scientific research.

A proliferation of medical journals, they said, has helped fuel demand for ever more research articles. But those same journals, many of them operated by multibillion-dollar publishing companies, often respond slowly or do nothing at all once one of those articles is shown to contain copied data. Journals retract papers at a fraction of the rate at which they publish ones with problems.

Springer Nature, which published nine of the articles that Dr. David said contained discrepancies across five journals, said it was investigating concerns. So did the American Association for Cancer Research, which published 10 articles under question from Dr. Yoon’s lab across four journals.

It is difficult to know who is responsible for errors in articles. Eleven of the scientists’ co-authors, including researchers at Harvard, Duke and Georgetown, did not answer emailed inquiries.

The articles under question examined why certain stomach and soft-tissue cancers withstood treatment, and how that resistance could be overcome.

The two independent image specialists said the volume of copied data, along with signs that some images had been rotated or similarly manipulated, suggested considerable sloppiness or worse.

“There are examples in this set that raise pretty serious red flags for the possibility of misconduct,” said Dr. Matthew Schrag, a Vanderbilt University neurologist who commented as part of his outside work on research integrity.

One set of 10 articles identified by Dr. David showed repeated reuse of identical or overlapping black-and-white images of cancer cells supposedly under different experimental conditions, he said.

“There’s no reason to have done that unless you weren’t doing the work,” Dr. David said.

One of those papers , published in 2012, was formally tagged with corrections. Unlike later studies, which were largely overseen by Dr. Yoon in New York, this paper was written by South Korea-based scientists, including Changhwan Yoon, who then worked in Seoul.

An immunologist in Norway randomly selected the paper as part of a screening of copied data in cancer journals. That led the paper’s publisher, the medical journal Oncogene, to add corrections in 2016.

But the journal did not catch all of the duplicated data , Dr. David said. And, he said, images from the study later turned up in identical form in another paper that remains uncorrected.

Copied cancer data kept recurring, Dr. David said. A picture of a small red tumor from a 2017 study reappeared in papers in 2020 and 2021 under different descriptions, he said. A ruler included in the pictures for scale wound up in two different positions.

The 2020 study included another tumor image that Dr. David said appeared to be a mirror image of one previously published by Dr. Yoon’s lab. And the 2021 study featured a color version of a tumor that had appeared in an earlier paper atop a different section of ruler, Dr. David said.

“This is another example where this looks intentionally done,” Dr. Bik said.

The researchers were faced with more serious action when the publisher Elsevier withdrew the stomach cancer study that had been published online in 2021. “The editors determined that the article violated journal publishing ethics guidelines,” Elsevier said.

Roland Herzog, the editor of Molecular Therapy, the journal where the article appeared, said that “image duplications were noticed” as part of a process of screening for discrepancies that the journal has since continued to beef up.

Because the problems were detected before the study was ever published in the print journal, Elsevier’s policy dictated that the article be taken down and no explanation posted online.

But that decision appeared to conflict with industry guidelines from the Committee on Publication Ethics . Posting articles online “usually constitutes publication,” those guidelines state. And when publishers pull such articles, the guidelines say, they should keep the work online for the sake of transparency and post “a clear notice of retraction.”

Dr. Herzog said he personally hoped that such an explanation could still be posted for the stomach cancer study. The journal editors and Elsevier, he said, are examining possible options.

The editors notified Dr. Yoon and Changhwan Yoon of the article’s removal, but neither scientist alerted Memorial Sloan Kettering, the hospital said. Columbia did not say whether it had been told.

Experts said the handling of the article was symptomatic of a tendency on the part of scientific publishers to obscure reports of lapses .

“This is typical, sweeping-things-under-the-rug kind of nonsense,” said Dr. Ivan Oransky, co-founder of Retraction Watch, which keeps a database of 47,000-plus retracted papers. “This is not good for the scientific record, to put it mildly.”

Susan C. Beachy contributed research.

Benjamin Mueller reports on health and medicine. He was previously a U.K. correspondent in London and a police reporter in New York. More about Benjamin Mueller

Advertisement

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • View all journals
  • Explore content
  • About the journal
  • Publish with us
  • Sign up for alerts
  • Review Article
  • Published: 17 July 2023

Large language models in medicine

  • Arun James Thirunavukarasu   ORCID: orcid.org/0000-0001-8968-4768 1 , 2 ,
  • Darren Shu Jeng Ting 3 , 4 , 5 ,
  • Kabilan Elangovan   ORCID: orcid.org/0000-0002-7711-7368 6 ,
  • Laura Gutierrez   ORCID: orcid.org/0000-0001-7416-2350 6 ,
  • Ting Fang Tan 6 , 7 &
  • Daniel Shu Wei Ting 6 , 7 , 8  

Nature Medicine volume  29 ,  pages 1930–1940 ( 2023 ) Cite this article

29k Accesses

101 Citations

73 Altmetric

Metrics details

  • Patient education
  • Translational research

Large language models (LLMs) can respond to free-text queries without being specifically trained in the task in question, causing excitement and concern about their use in healthcare settings. ChatGPT is a generative artificial intelligence (AI) chatbot produced through sophisticated fine-tuning of an LLM, and other tools are emerging through similar developmental processes. Here we outline how LLM applications such as ChatGPT are developed, and we discuss how they are being leveraged in clinical settings. We consider the strengths and limitations of LLMs and their potential to improve the efficiency and effectiveness of clinical, educational and research work in medicine. LLM chatbots have already been deployed in a range of biomedical contexts, with impressive but mixed results. This review acts as a primer for interested clinicians, who will determine if and how LLM technology is used in healthcare for the benefit of patients and practitioners.

This is a preview of subscription content, access via your institution

Access options

Access Nature and 54 other Nature Portfolio journals

Get Nature+, our best-value online-access subscription

24,99 € / 30 days

cancel any time

Subscribe to this journal

Receive 12 print issues and online access

195,33 € per year

only 16,28 € per issue

Rent or buy this article

Prices vary by article type

Prices may be subject to local taxes which are calculated during checkout

research papers in medical domain

Esteva, A. et al. A guide to deep learning in healthcare. Nat. Med. 25 , 24–29 (2019).

CAS   PubMed   Google Scholar  

Aggarwal, R. et al. Diagnostic accuracy of deep learning in medical imaging: a systematic review and meta-analysis. NPJ Digit. Med. 4 , 65 (2021).

PubMed   PubMed Central   Google Scholar  

Liddy, E. Natural language processing. In Encyclopedia of Library and Information Science (eds Kent, A. & Lancour, H.)(Marcel Decker, 2001).

Khurana, D., Koli, A., Khatter, K. & Singh, S. Natural language processing: state of the art, current trends and challenges. Multimed. Tools Appl. 82 , 3713–3744 (2023).

PubMed   Google Scholar  

Brown, T. et al. Language models are few-shot learners. In Advances in Neural Information Processing Systems Vol. 33 1877–1901 (Curran Associates, 2020).

Moor, M. et al. Foundation models for generalist medical artificial intelligence. Nature 616 , 259–265 (2023).

Kaplan, J. et al. Scaling laws for neural language models. Preprint at arXiv https://doi.org/10.48550/arXiv.2001.08361 (2020).

Shoeybi, M. et al. Megatron-LM: training multi-billion parameter language models using model parallelism. Preprint at arXiv https://doi.org/10.48550/arXiv.1909.08053 (2020).

Thoppilan, R. et al. LaMDA: language models for dialog applications. Preprint at arXiv https://doi.org/10.48550/arXiv.2201.08239 (2022).

Zeng, A. et al. GLM-130B: an open bilingual pre-trained model. Preprint at arXiv https://doi.org/10.48550/arXiv.2210.02414 (2022).

Amatriain, X. Transformer models: an introduction and catalog. Preprint at arXiv https://doi.org/10.48550/arXiv.2302.07730 (2023).

Introducing ChatGPT. https://openai.com/blog/chatgpt

Ouyang, L. et al. Training language models to follow instructions with human feedback. Preprint at arXiv https://doi.org/10.48550/arXiv.2203.02155 (2022).

OpenAI. GPT-4 technical report. Preprint at arXiv https://doi.org/10.48550/arXiv.2303.08774 (2023).

Kung, T. H. et al. Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models. PLoS Digit. Health 2 , e0000198 (2023).

Thirunavukarasu, A. J. et al. Trialling a large language model (ChatGPT) in general practice with the applied knowledge test: observational study demonstrating opportunities and limitations in primary care. JMIR Med. Educ. 9 , e46599 (2023).

Ayers, J. W. et al. Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum. JAMA Intern. Med. 183 , 589–596 (2023).

Rajpurkar, P., Chen, E., Banerjee, O. & Topol, E. J. AI in health and medicine. Nat. Med. 28 , 31–38 (2022).

Radford, A., Narasimhan, K., Salimans, T. & Sutskever, I. Improving language understanding by generative pre-training. https://openai.com/research/language-unsupervised (2018).

Radford, A. et al. Language models are unsupervised multitask learners. Preprint at Semantic Scholar https://www.semanticscholar.org/paper/Language-Models-are-Unsupervised-Multitask-Learners-Radford-Wu/9405cc0d6169988371b2755e573cc28650d14dfe (2018).

Qiu, X. et al. Pre-trained models for natural language processing: a survey. Sci. China Technol. Sci. 63 , 1872–1897 (2020).

Google Scholar  

Touvron, H. et al. LLaMA: open and efficient foundation language models. Preprint at arXiv https://doi.org/10.48550/arXiv.2302.13971 (2023).

Dennean, K., Gantori, S., Limas, D. K., Pu, A. & Gilligan, R. Let’s chat about ChatGPT. https://www.ubs.com/global/en/wealth-management/our-approach/marketnews/article.1585717.html (2023).

Dai, D. et al. Why can GPT learn in-context? Language models secretly perform gradient descent as meta-optimizers. Preprint at arXiv https://doi.org/10.48550/arXiv.2212.10559 (2022).

Confirmed: the new Bing runs on OpenAI’s GPT-4. https://blogs.bing.com/search/march_2023/Confirmed-the-new-Bing-runs-on-OpenAI’s-GPT-4/ (2023).

Glaese, A. et al. Improving alignment of dialogue agents via targeted human judgements. Preprint at arXiv https://doi.org/10.48550/arXiv.2209.14375 (2022).

Shuster, K. et al. BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage. Preprint at arXiv https://doi.org/10.48550/arXiv.2208.03188 (2022).

Shuster, K. et al. Language models that seek for knowledge: modular search & generation for dialogue and prompt completion. Preprint at arXiv https://doi.org/10.48550/arXiv.2203.13224 (2022).

Anil, R. et al. PaLM 2 technical report. Preprint at arXiv https://doi.org/10.48550/arXiv.2305.10403 (2023).

HuggingChat. https://hf.co/chat

Taori, R. et al. Alpaca: a strong, replicable instruction-following model. Preprint at https://crfm.stanford.edu/2023/03/13/alpaca.html (2023).

OpenAI. GPT-4 system card. https://cdn.openai.com/papers/gpt-4-system-card.pdf (2023).

Lacoste, A., Luccioni, A., Schmidt, V. & Dandres, T. Quantifying the carbon emissions of machine learning. Preprint at arXiv https://doi.org/10.48550/arXiv.1910.09700 (2019).

Patterson, D. et al. The carbon footprint of machine learning training will plateau, then shrink. Preprint at arXiv https://doi.org/10.48550/arXiv.2204.05149 (2022).

Strubell, E., Ganesh, A. & McCallum, A. Energy and policy considerations for deep learning in NLP. Preprint at arXiv https://doi.org/10.48550/arXiv.1906.02243 (2019).

Bender, E. M., Gebru, T., McMillan-Major, A. & Shmitchell, S. On the dangers of stochastic parrots: can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency 610–623 https://doi.org/10.1145/3442188.3445922 (Association for Computing Machinery, 2021).

ARK Investment Management LLC. Big Ideas 2023. https://ark-invest.com/home-thank-you-big-ideas-2023/?submissionGuid=d741a6f9-1a47-43d4-ac82-901cd909ff96 (2023).

Nori, H., King, N., McKinney, S. M., Carignan, D. & Horvitz, E. Capabilities of GPT-4 on medical challenge problems. Preprint at arXiv https://doi.org/10.48550/arXiv.2303.13375 (2023).

Singhal, K. et al. Towards expert-level medical question answering with large language models. Preprint at arXiv https://doi.org/10.48550/arXiv.2305.09617 (2023).

Looi, M.-K. Sixty seconds on… ChatGPT. BMJ 380 , p205 (2023).

Pause giant AI experiments: an open letter. Future of Life Institute. https://futureoflife.org/open-letter/pause-giant-ai-experiments/ (2023).

Lee, P., Bubeck, S. & Petro, J. Benefits, limits, and risks of GPT-4 as an AI chatbot for medicine. N. Engl. J. Med. 388 , 1233–1239 (2023).

Singhal, K. et al. Large language models encode clinical knowledge. Preprint at arXiv https://doi.org/10.48550/arXiv.2212.13138 (2022).

Gilson, A. et al. How does ChatGPT perform on the United States Medical Licensing Examination? The implications of large language models for medical education and knowledge assessment. JMIR Med. Educ. 9 , e45312 (2023).

Sarraju, A. et al. Appropriateness of cardiovascular disease prevention recommendations obtained from a popular online chat-based artificial intelligence model. JAMA 329 , 842–844 (2023).

Nastasi, A. J., Courtright, K. R., Halpern, S. D. & Weissman, G. E. Does ChatGPT provide appropriate and equitable medical advice?: a vignette-based, clinical evaluation across care contexts. Preprint at medRxiv https://doi.org/10.1101/2023.02.25.23286451 (2023).

Rao, A. et al. Assessing the utility of ChatGPT throughout the entire clinical workflow. Preprint at medRxiv https://doi.org/10.1101/2023.02.21.23285886 (2023).

Levine, D. M. et al. The diagnostic and triage accuracy of the GPT-3 artificial intelligence model. Preprint at medRxiv https://doi.org/10.1101/2023.01.30.23285067 (2023).

Nov, O., Singh, N. & Mann, D. M. Putting ChatGPT’s medical advice to the (Turing) test. Preprint at medRxiv https://doi.org/10.1101/2023.01.23.23284735 (2023).

Thirunavukarasu, A. J. Large language models will not replace healthcare professionals: curbing popular fears and hype. J. R. Soc. Med. 116 , 181–182 (2023).

Kraljevic, Z. et al. Foresight—Generative Pretrained Transformer (GPT) for modelling of patient timelines using EHRs. Preprint at arXiv https://doi.org/10.48550/arXiv.2212.08072 (2023).

Shao, Y. et al. Hybrid value-aware transformer architecture for joint learning from longitudinal and non-longitudinal clinical data. Preprint at medRxiv https://doi.org/10.1101/2023.03.09.23287046 (2023).

Adams, L. C. et al. Leveraging GPT-4 for post hoc transformation of free-text radiology reports into structured reporting: a multilingual feasibility study. Radiology 307 , e230725 (2023).

Arora, A. & Arora, A. The promise of large language models in health care. Lancet 401 , 641 (2023).

Spataro, J. Introducing Microsoft 365 Copilot—your copilot for work. The Official Microsoft Blog. https://blogs.microsoft.com/blog/2023/03/16/introducing-microsoft-365-copilot-your-copilot-for-work/ (2023).

Ghahramani, Z. Introducing PaLM 2. Google. https://blog.google/technology/ai/google-palm-2-ai-large-language-model/ (2023).

Patel, S. B. & Lam, K. ChatGPT: the future of discharge summaries? Lancet Digit. Health 5 , e107–e108 (2023).

Will ChatGPT transform healthcare? Nat. Med. 29 , 505–506 (2023).

Our latest health AI research updates. Google. https://blog.google/technology/health/ai-llm-medpalm-research-thecheckup/ (2023).

Khan, S. Harnessing GPT-4 so that all students benefit. A nonprofit approach for equal access! Khan Academy Blog. https://blog.khanacademy.org/harnessing-ai-so-that-all-students-benefit-a-nonprofit-approach-for-equal-access/ (2023).

Duolingo Team. Introducing Duolingo Max, a learning experience powered by GPT-4. Duolingo Blog. https://blog.duolingo.com/duolingo-max/ (2023).

Han, Z., Battaglia, F., Udaiyar, A., Fooks, A. & Terlecky, S. R. An explorative assessment of ChatGPT as an aid in medical education: use it with caution. Preprint at medRxiv https://doi.org/10.1101/2023.02.13.23285879 (2023).

Benoit, J. R. A. ChatGPT for clinical vignette generation, revision, and evaluation. Preprint at medRxiv https://doi.org/10.1101/2023.02.04.23285478 (2023).

Lee, J. et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics 36 , 1234–1240 (2020).

Gu, Y. et al. Domain-specific language model pretraining for biomedical natural language processing. ACM Trans. Comput. Health. 3 , 1–23 (2022).

Salganik, M. Can ChatGPT—and its successors—go from cool to tool? Freedom to Tinker. https://freedom-to-tinker.com/2023/03/08/can-chatgpt-and-its-successors-go-from-cool-to-tool/ (2023).

Zhavoronkov, A. Caution with AI-generated content in biomedicine. Nat. Med 29 , 532 (2023).

Yang, X. et al. A large language model for electronic health records. NPJ Digit. Med. 5 , 194 (2022).

Agrawal, M., Hegselmann, S., Lang, H., Kim, Y. & Sontag, D. Large language models are few-shot clinical information extractors. Preprint at arXiv https://doi.org/10.48550/arXiv.2205.12689 (2022).

Huang, K., Altosaar, J. & Ranganath, R. ClinicalBERT: modeling clinical notes and predicting hospital readmission. Preprint at arXiv https://doi.org/10.48550/arXiv.1904.05342 (2020).

Madani, A. et al. Large language models generate functional protein sequences across diverse families. Nat. Biotechnol. https://doi.org/10.1038/s41587-022-01618-2 (2023).

Mai, D. H. A., Nguyen, L. T. & Lee, E. Y. TSSNote-CyaPromBERT: development of an integrated platform for highly accurate promoter prediction and visualization of Synechococcus sp. and Synechocystis sp. through a state-of-the-art natural language processing model BERT. Front. Genet. 13 , 1067562 (2022).

CAS   PubMed   PubMed Central   Google Scholar  

Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596 , 583–589 (2021).

Yan, C. et al. A multifaceted benchmarking of synthetic electronic health record generation models. Nat. Commun. 13 , 7609 (2022).

OpenAI. Model index for researchers. https://platform.openai.com/docs/model-index-for-researchers

Ball, P. The lightning-fast quest for COVID vaccines—and what it means for other diseases. Nature 589 , 16–18 (2021).

Hallin, J. et al. Anti-tumor efficacy of a potent and selective non-covalent KRASG12D inhibitor. Nat. Med. 28 , 2171–2182 (2022).

Babbage, C. Passages from the Life of a Philosopher (Longman, Green, Longman, Roberts, & Green, 1864).

Total data volume worldwide 2010-2025. Statista . https://www.statista.com/statistics/871513/worldwide-data-created/

Villalobos, P. et al. Will we run out of data? An analysis of the limits of scaling datasets in machine learning. Preprint at arXiv https://doi.org/10.48550/arXiv.2211.04325 (2022).

Ji, Z. et al. Survey of hallucination in natural language generation. ACM Comput. Surv. 55 , 1–38 (2023).

Alkaissi, H. & McFarlane, S. I. Artificial hallucinations in ChatGPT: implications in scientific writing. Cureus 15 , e35179 (2023).

Huang, J. et al. Large language models can self-improve. Preprint at arXiv https://doi.org/10.48550/arXiv.2210.11610 (2022).

Wang, X. et al. Self-consistency improves chain of thought reasoning in language models. Preprint at arXiv https://doi.org/10.48550/arXiv.2203.11171 (2023).

Bommasani, R. et al. On the opportunities and risks of foundation models. Preprint at arXiv https://doi.org/10.48550/arXiv.2108.07258 (2022).

Ramesh, A., Dhariwal, P., Nichol, A., Chu, C. & Chen, M. Hierarchical text-conditional image generation with CLIP latents. Preprint at arXiv https://doi.org/10.48550/arXiv.2204.06125 (2022).

Zini, J. E. & Awad, M. On the explainability of natural language processing deep models. ACM Comput. Surv. 55 , 1–103 (2022).

Barredo Arrieta, A. et al. Explainable Artificial Intelligence (XAI): concepts, taxonomies, opportunities and challenges toward responsible AI. Inf. Fusion 58 , 82–115 (2020).

Else, H. Abstracts written by ChatGPT fool scientists. Nature 613 , 423–423 (2023).

Taylor, J. ChatGPT’s alter ego, Dan: users jailbreak AI program to get around ethical safeguards. The Guardian https://www.theguardian.com/technology/2023/mar/08/chatgpt-alter-ego-dan-users-jailbreak-ai-program-to-get-around-ethical-safeguards (2023).

Perez, F. & Ribeiro, I. Ignore previous prompt: attack techniques for language models. Preprint at arXiv https://doi.org/10.48550/arXiv.2211.09527 (2022).

Li, X. & Zhang, T. An exploration on artificial intelligence application: from security, privacy and ethic perspective. In 2017 IEEE 2nd International Conference on Cloud Computing and Big Data Analysis (ICCCBDA) 416–420 https://doi.org/10.1109/ICCCBDA.2017.7951949 (Curran Associates, 2017).

Wolford, B. What is GDPR, the EU’s new data protection law? https://gdpr.eu/what-is-gdpr/ (2018).

Thorp, H. H. ChatGPT is fun, but not an author. Science 379 , 313 (2023).

Yeo-Teh, N. S. L. & Tang, B. L. NLP systems such as ChatGPT cannot be listed as an author because these cannot fulfill widely adopted authorship criteria. Account Res. https://doi.org/10.1080/08989621.2023.2185776 (2023).

Stokel-Walker, C. ChatGPT listed as author on research papers: many scientists disapprove. Nature 613 , 620–621 (2023).

Lehman, E. et al. Do we still need clinical language models? Preprint at arXiv https://doi.org/10.48550/arXiv.2302.08091 (2023).

Yang, X. et al. GatorTron: a large clinical language model to unlock patient information from unstructured electronic health records. Preprint at arXiv https://doi.org/10.48550/arXiv.2203.03540 (2022).

Weiner, S. J., Wang, S., Kelly, B., Sharma, G. & Schwartz, A. How accurate is the medical record? A comparison of the physician’s note with a concealed audio recording in unannounced standardized patient encounters. J. Am. Med. Inf. Assoc. 27 , 770–775 (2020).

Ioannidis, J. P. A. Why most published research findings are false. PLoS Med. 2 , e124 (2005).

Liebrenz, M., Schleifer, R., Buadze, A., Bhugra, D. & Smith, A. Generating scholarly content with ChatGPT: ethical challenges for medical publishing. Lancet Digit. Health 5 , e105–e106 (2023).

Stokel-Walker C. AI bot ChatGPT writes smart essays—should academics worry? Nature https://doi.org/10.1038/d41586-022-04397-7 (2022).

Elali, F. R. & Rachid, L. N. AI-generated research paper fabrication and plagiarism in the scientific community. Patterns 4 , 100706 (2023).

Tools such as ChatGPT threaten transparent science; here are our ground rules for their use. Nature 613 , 612–612 (2023).

Sample, I. Science journals ban listing of ChatGPT as co-author on papers. The Guardian https://www.theguardian.com/science/2023/jan/26/science-journals-ban-listing-of-chatgpt-as-co-author-on-papers (2023).

Flanagin, A., Bibbins-Domingo, K., Berkwits, M. & Christiansen, S. L. Nonhuman ‘authors’ and implications for the integrity of scientific publication and medical knowledge. JAMA 329 , 637–639 (2023).

Authorship and contributorship. Cambridge Core. https://www.cambridge.org/core/services/authors/publishing-ethics/research-publishing-ethics-guidelines-for-journals/authorship-and-contributorship

New AI classifier for indicating AI-written text. https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text

Kirchenbauer, J. et al. A watermark for large language models. Preprint at arXiv http://arxiv.org/abs/2301.10226 (2023).

The Lancet Digital Health. ChatGPT: friend or foe? Lancet Digit. Health 5 , e102 (2023).

Mbakwe, A. B., Lourentzou, I., Celi, L. A., Mechanic, O. J. & Dagan, A. ChatGPT passing USMLE shines a spotlight on the flaws of medical education. PLoS Digit. Health 2 , e0000205 (2023).

Abid, A., Farooqi, M. & Zou, J. Large language models associate Muslims with violence. Nat. Mach. Intell. 3 , 461–463 (2021).

Nangia, N., Vania, C., Bhalerao, R. & Bowman, S. R. CrowS-Pairs: a challenge dataset for measuring social biases in masked language models. In Proc. of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) 1953–1967 https://doi.org/10.18653/v1/2020.emnlp-main.154 (Association for Computational Linguistics, 2020).

Bender, E. M. & Friedman, B. Data statements for natural language processing: toward mitigating system bias and enabling better science. In Transactions of the Association for Computational Linguistics 6 , 587–604 (2018).

Li, H. et al. Ethics of large language models in medicine and medical research. Lancet Digit. Health 5 , e333–e335 (2023).

Aggarwal, A., Tam, C. C., Wu, D., Li, X. & Qiao, S. Artificial intelligence–based chatbots for promoting health behavioral changes: systematic review. J. Med. Internet Res. 25 , e40789 (2023).

Vasey, B. et al. Reporting guideline for the early-stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI. Nat. Med. 28 , 924–933 (2022).

Friedberg, M. W. et al. Factors affecting physician professional satisfaction and their implications for patient care, health systems, and health policy. RAND Health Q 3 , 1 (2014).

Kwee, A., Teo, Z. L. & Ting, D. S. W. Digital health in medicine: important considerations in evaluating health economic analysis. Lancet Reg. Health West Pac. 23 , 100476 (2022).

Littmann, M. et al. Validity of machine learning in biology and medicine increased through collaborations across fields of expertise. Nat. Mach. Intell. 2 , 18–24 (2020).

Download references

Acknowledgements

D.S.W.T. is supported by the National Medical Research Council, Singapore (NMCR/HSRG/0087/2018, MOH-000655-00 and MOH-001014-00), the Duke-NUS Medical School (Duke-NUS/RSF/2021/0018 and 05/FY2020/EX/15-A58) and the Agency for Science, Technology and Research (A20H4g2141 and H20C6a0032). These funders were not involved in the conception, execution or reporting of this review.

Author information

Authors and affiliations.

University of Cambridge School of Clinical Medicine, Cambridge, UK

Arun James Thirunavukarasu

Corpus Christi College, University of Cambridge, Cambridge, UK

Academic Unit of Ophthalmology, Institute of Inflammation and Ageing, University of Birmingham, Birmingham, UK

Darren Shu Jeng Ting

Birmingham and Midland Eye Centre, Birmingham, UK

Academic Ophthalmology, School of Medicine, University of Nottingham, Nottingham, UK

Artificial Intelligence and Digital Innovation Research Group, Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore

Kabilan Elangovan, Laura Gutierrez, Ting Fang Tan & Daniel Shu Wei Ting

Department of Ophthalmology and Visual Sciences, Duke-National University of Singapore Medical School, Singapore, Singapore

Ting Fang Tan & Daniel Shu Wei Ting

Byers Eye Institute, Stanford University, Palo Alto, CA, USA

Daniel Shu Wei Ting

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Daniel Shu Wei Ting .

Ethics declarations

Competing interests.

D.S.W.T. holds a patent on a deep learning system for the detection of retinal diseases. The other authors declare no conflicts of interest.

Peer review

Peer review information.

Nature Medicine thanks Melissa McCradden, Pranav Rajpurkar and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary handling editor: Karen O’Leary, in collaboration with the Nature Medicine team.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Cite this article.

Thirunavukarasu, A.J., Ting, D.S.J., Elangovan, K. et al. Large language models in medicine. Nat Med 29 , 1930–1940 (2023). https://doi.org/10.1038/s41591-023-02448-8

Download citation

Received : 24 March 2023

Accepted : 08 June 2023

Published : 17 July 2023

Issue Date : August 2023

DOI : https://doi.org/10.1038/s41591-023-02448-8

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

An evaluation of gpt models for phenotype concept recognition.

  • Tudor Groza
  • Harry Caufield
  • Justin T. Reese

BMC Medical Informatics and Decision Making (2024)

The validity of electronic health data for measuring smoking status: a systematic review and meta-analysis

  • Md Ashiqul Haque
  • Muditha Lakmali Bodawatte Gedara
  • Lisa M. Lix

Feasibility of combining spatial computing and AI for mental health support in anxiety and depression

  • Brennan M. R. Spiegel
  • Muskaan Mehra

npj Digital Medicine (2024)

Diagnostic reasoning prompts reveal the potential for large language model interpretability in medicine

  • Thomas Savage
  • Ashwin Nayak
  • Jonathan H. Chen

Performance evaluation of ChatGPT, GPT-4, and Bard on the official board examination of the Japan Radiology Society

  • Yoshitaka Toyama
  • Ayaka Harigai

Japanese Journal of Radiology (2024)

Quick links

  • Explore articles by subject
  • Guide to authors
  • Editorial policies

Sign up for the Nature Briefing: Translational Research newsletter — top stories in biotechnology, drug discovery and pharma.

research papers in medical domain

IMAGES

  1. (PDF) The medical research paper: Structure and functions

    research papers in medical domain

  2. Research Paper Example Of Scientific Paper / Publications Special Issue

    research papers in medical domain

  3. Reflection essay: Example of subject of the study in research paper

    research papers in medical domain

  4. Scientific Research Paper Sample

    research papers in medical domain

  5. 31+ Research Paper Templates in PDF

    research papers in medical domain

  6. 😎 Medical paper. Medical Health Research Papers Custom Written. 2019-03-02

    research papers in medical domain

VIDEO

  1. Introduction to research methodology for health sciences-Second day 2023

  2. how to write university exam papers for good marks

  3. Mastering Research Methodology: An Introduction to Medical Research

  4. How to critically review research papers in 4 steps #academicwriting #studytips #thesis

  5. Different Types of Research in Nursing

  6. Struggling finding research papers for your literature review? #researchtips #studytips #chatgpt

COMMENTS

  1. A study of generative large language model for medical research and

    This study develops a generative clinical LLM, GatorTronGPT, using 277 billion words of text including (1) 82 billion words of clinical text from 126 clinical departments and approximately 2...

  2. Significance of machine learning in healthcare ...

    This paper identifies the need for ML in healthcare. Paper identifies and discusses the significant applications of ML for Healthcare. Paper explores how ML-based tools are used to provide various treatment alternatives and individualised treatments and improve the overall efficiency of healthcare systems.

  3. Deep learning for healthcare: review, opportunities and challenges

    Various types of data have been emerging in modern biomedical research, including electronic health records, imaging, -omics, sensor data and text, which are complex, heterogeneous, poorly annotated and generally unstructured.

  4. Medical deep learning—A systematic meta-review

    Deep learning is also showing vast potential in the medical domain. With the collection of large quantities of patient records and data, and a trend towards personalized treatments, there is a great need for automated and reliable processing and analysis of health information.

  5. Clinical Predictive Modeling of Heart Failure: Domain ...

    This study attempts to identify and briefly describe the current directions in applied and theoretical clinical prediction research. Context-rich chronic heart failure syndrome (CHFS) telemedicine provides the medical foundation for this effort. In the chronic stage of heart failure, there are sudden exacerbations of syndromes with subsequent hospitalizations, which are called acute ...

  6. Recommender systems in the healthcare domain: state-of-the-art and

    1 Introduction In the past decades, a considerable amount of clinical data representing patients' health status (e.g., medical reports, laboratory results, and disease treatment plans) have been collected. This has remarkably increased digital information available for patient-oriented decision making.

  7. Large language models encode clinical knowledge

    Open access Published: 12 July 2023 Large language models encode clinical knowledge Karan Singhal, Shekoofeh Azizi, Tao Tu, S. Sara Mahdavi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tanwani,...

  8. Deep learning for unsupervised domain adaptation in medical imaging

    This comprehensive survey includes more than 140 papers published between 2018 and 2023, as depicted in . These papers cover a diverse range of applications for employing deep unsupervised domain adaptation learning methods in the analysis of medical images.

  9. Self-supervised learning for medical image classification: a ...

    Some papers leveraged medical domain priors to create specialized strategies for creating positive pairs. ... The research papers had to be original research in the form of journal articles ...

  10. Data Preprocessing for Decision Making in Medical Informatics

    This paper presents a systematic mapping study of the current studies using the data preprocessing in medical domain. 110 primary studies published in the period between January 2000 and March 2017 were selected and classified according to three classification criteria: publication year and channel, research type and preprocessing tasks used ...

  11. (PDF) Explainable Artificial Intelligence in the Medical Domain: A

    Current research on explainable artificial intelligence (XAI) focuses on the interpretability of opaque ML models (Arrieta et al., 2020). The primary goal is to explain the logic behind a model's ...

  12. The Research Domain of Family Medicine

    This article characterizes the large research domain of family medicine. It is a domain that can be productively explored from different perspectives, including: (1) the ecology of medical care and its focus on the environments of health care and interactions among them; (2) the realm of causation and important opportunities to discover how people lose and regain their health; (3) knowing ...

  13. [2303.14070] ChatDoctor: A Medical Chat Model Fine-Tuned on a Large

    [2303.14070] ChatDoctor: A Medical Chat Model Fine-Tuned on a Large Language Model Meta-AI (LLaMA) Using Medical Domain Knowledge Computer Science > Computation and Language [Submitted on 24 Mar 2023 ( v1 ), last revised 24 Jun 2023 (this version, v5)]

  14. A Survey on Medical Explainable AI (XAI): Recent Progress ...

    The emerging field of eXplainable AI (XAI) in the medical domain is considered to be of utmost importance. Meanwhile, incorporating explanations in the medical domain with respect to legal and ethical AI is necessary to understand detailed decisions, results, and current status of the patient's conditions. Successively, we will be presenting a detailed survey for the medical XAI with the ...

  15. PDF Prompt Engineering for Healthcare: Methodologies and Applications

    Figure 2: A graphical representation is utilized to depict the number of research papers on prompt engineering for NLP in the medical domain, published from 2019 to April 6, 2023, revealing the trend and growth of this eld over time. 1.2 Literature Search Process The importance of NLP tasks in the medical eld cannot be overstated. With

  16. Internet of Medical Things (IoMT)-Based Smart Healthcare System: Trends

    Internet of Medical Thing (IoMT) is the most emerging era of the Internet of Thing (IoT), which is exponentially gaining researchers' attention with every passing day because of its wide applicability in Smart Healthcare systems (SHS).

  17. Diagnostic accuracy of deep learning in medical imaging: a ...

    Five hundred three papers fulfilled inclusion criteria for the systematic review and contained data required for sensitivity, specificity or AUC. ... For the flow and timing domain, a high or ...

  18. An exploratory study of automatic text summarization in biomedical and

    The information related to biomedical and healthcare domain comes in different formats and from various documents like scientific journals, research papers, clinical documents, online reports, EHRs, unstructured medical texts and many more [7]. So, characteristics of each type of documents should be considered before developing or applying any ...

  19. Systematic Mapping Study of AI/Machine Learning in Healthcare and

    This study attempts to categorise research conducted in the area of: use of machine learning in healthcare, using a systematic mapping study methodology.In our attempt, we reviewed literature from top journals, articles, and conference papers by using the keywords use of machine learning in healthcare.We queried Google Scholar, resulted in 1400 papers, and then categorised the results on the ...

  20. The domain expert perspective: A qualitative study into the views

    This paper investigates what matters to medical domain experts when setting standards on a language for specific purposes (LSP) English proficiency test: the Occupational English Test's ... This paper has sought answers to two research questions. In response to the first (i.e., what criteria do standard-setting participants in the healthcare ...

  21. Med-PaLM: A Medical Large Language Model

    A large language model from Google Research, designed for the medical domain. Watch Video Introduction Med-PaLM is a large language model (LLM) designed to provide high quality answers to medical questions.

  22. A Literature Review On Chatbots In Healthcare Domain

    (PDF) A Literature Review On Chatbots In Healthcare Domain A Literature Review On Chatbots In Healthcare Domain Authors: Nivedita Bhirud Vishwakarma Institute Of Information Technology, Pune...

  23. A Survey on Domain Generalization for Medical Image Analysis

    Medical Image Analysis (MedIA) has emerged as a crucial tool in computer-aided diagnosis systems, particularly with the advancement of deep learning (DL) in recent years. However, well-trained deep models often experience significant performance degradation when deployed in different medical sites, modalities, and sequences, known as a domain shift issue. In light of this, Domain ...

  24. Journal of Medical Internet Research

    Interactive Journal of Medical Research 350 articles JMIRx Med 334 articles ... This paper is in the following e-collection/theme issue: Telehealth and Telemonitoring (1433) Diagnostic Tools in Mental Health (205) Autism ...

  25. The future landscape of large language models in medicine

    However, in general, recently updated versions and models designed specifically for medical applications and trained on medical data show promising progress in this domain 2,5,74. Nevertheless ...

  26. Journal of Medical Internet Research

    Background: With the advent of a new era for health and medical treatment, characterized by the integration of mobile technology, a significant digital divide has surfaced, particularly in the engagement of older individuals with mobile health (mHealth). The health of a family is intricately connected to the well-being of its members, and the use of media plays a crucial role in facilitating ...

  27. A Columbia Surgeon's Study Was Pulled. He Kept Publishing Flawed Data

    That led the paper's publisher, the medical journal Oncogene, to add corrections in 2016. But the journal did not catch all of the duplicated data , Dr. David said.

  28. Large language models in medicine

    Models fine-tuned with domain-specific information may ... L. N. AI-generated research paper fabrication and plagiarism in the scientific community. ... D.S.W.T. is supported by the National ...