Smart healthcare: A prospective future medical approach for COVID-19 : Journal of the Chinese Medical Association

Secondary Logo

Journal Logo

Review Articles

Smart healthcare: A prospective future medical approach for COVID-19

Yang, De-Minga,b,c,*; Chang, Tai-Jaya,d,e; Hung, Kai-Fenga; Wang, Mong-Liena; Cheng, Yen-Fua; Chiang, Su-Huaa; Chen, Mei-Fanga; Liao, Yi-Tinga,f,g; Lai, Wei-Quna,b,c; Liang, Kung-Haoa,f,g,*

Author Information
Journal of the Chinese Medical Association 86(2):p 138-146, February 2023. | DOI: 10.1097/JCMA.0000000000000824
  • Open



1.1. COVID-19

The COVID-191 pandemic, caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2),2 has lasted for almost 3 years, and at present, this pandemic still continues. The virus that causes COVID-19, SARS-CoV-2, is a betacoronavirus with a single-strand RNA genome containing roughly 29903 base pairs (NC_045512.2). It is related to SARS-CoV (the virus that caused the 2003 pandemic) and the Middle East respiratory syndrome coronavirus.3 They share a common spike protein (S) as a bridge to attach to the surface of human host cells through membrane receptor binding to angiotensin-converting enzyme 2.4 Because the S variant protein mutates frequently,5,6 the virus is able to escape the host immune system and thus cause new waves of transmission. Since the outbreak in 2019, the virus has mutated many times and produced many variants of concern (VOC). Recently, the Delta7 and Omicron strains (from BA.2 to BA.5)8,9 have dominated the viral species sequentially, with Omicron, in particular, having a relatively low death rate.10 Fever and coughing are no longer sufficient to identify all COVID-19 cases.11 The loss of taste and olfactory function (ageusia and anosmia)12 and a sharp sore throat have been identified as two new symptoms of SARS-CoV-2 infection. High numbers of asymptomatic COVID-19 cases have been reported.13 Confirmation of infection is increasingly difficult.

To deter the spread of COVID-19, pandemic monitoring (including rapid viral detection and surveillance of COVID-19 diagnoses),14,15 effective treatment (antiviral drugs and adequate healthcare provision),16 and large-scale vaccination (with continual boosters from next-generation vaccines)17 are all essential (Fig. 1). VOC whole genome sequencing18 is crucial for monitoring the evolution of SARS-CoV-2 variants; however, the economic burden of medical examination is high. Although quarantine measures and contact tracing are effective methods to reduce transmission, space is a limited resource. Moreover, the widespread adoption of public health measures for an extended time period has had a global economic impact. For citizens, the avoidance of unnecessary contact with the hands, nose, and eyes remains an effective method to prevent SARS-CoV-2 infection. As mask use decreases in numerous Western countries, new methods to contain the impacts of COVID-19 are being developed, such as consumer technology combined with cloud computing and big data.19

Fig. 1:
Areas of smart healthcare for the management of COVID-19.

1.2. Medical artificial intelligence

Ever since the emergence of artificial intelligence (AI), researchers have continued to investigate the numerous applications of big data across various fields, such as biomedical science and medical care.20–23 Big data in medicine utilizes four types of information: (1) coded data (diagnostic codes, drug codes, and disposition codes), (2) free text (descriptions by patients and pathological reports by physicians), (3) biophotonic data (including X-ray imaging, angiography, computed tomography, and magnetic resonance imaging), and (4) bioelectronic data (eg, heartbeat, blood pressure, breathing, and brain wave). For coded data, the International Statistical Classification of Diseases and Related Health Problems, 10th Revision recommends that bar code medication administration (BCMA) be replaced by a radiofrequency identification (RFID) or near-field communication (NFC). Free text can sometimes be collected through the recording of patient interactions with inquiry robots, such as Pepper, which was developed by Professor Ohe Kazuhiko of the University of Tokyo as part of a Japanese research group committed to the standardization of structured medical record exchange. These texts can be analyzed using natural language processing (eg, Cortana, Siri, Alexa, Google Home, and Apple HomePod). For biosignal (biophotonic and bioelectronic) data, wearable devices or the Internet of Medical Things (IoMT) can be used for data collection.

Medical AI can prevent ethical problems, which often arise in healthcare applications of big data and smart healthcare.24 Medical AI is not susceptible to errors related to human nature (biased or incomplete dataset recordings or incorrectly obtained consent for certain clinical treatments) or to mishandling of private data (insecure or intentionally leaked data).25 Medical AI uses machine learning (ML) or deep learning (DL) to conduct large-scale big data analysis of unstructured data and narrative texts from various individuals; however, medical AI often produces incorrect or inappropriate predictions.26 This problem can be overcome through the integration of medicine, computer science, and other disciplines.27 A well-designed medical AI can provide fast and valuable suggestions on disease prevention and effective therapeutic treatments by using data analysis methods including pattern recognition, analysis of various cloud computing platforms (eg, Amazon, Google, Microsoft), and the aforementioned biosignal data.28 For cloud calculations using certain algorithms, patient records not only can be stored and recorded but also can be applied to the analysis of clinical data to provide valuable suggestions for physician decision-making.29 Medical AI can play a promising role in healthcare provision and thus may threaten to displace human doctors.30 To date, cardiology, internal medicine, and radiology are three medical fields that make extensive use of medical AI.31,32

Despite the promising applications of medical AI to the management of various diseases,21–23 patient privacy is a serious concern.33 Electronic healthcare data (including medical big data, coded data, text data, and biosignals) used in picture archiving and communications systems should be collected and handled with care whether such data originates from hospitals and national institutes (eg, Taiwan’s National Health Insurance Research Database and Taiwanese biobanks) or from international data sources (eg, PubMed). The digital footprint of individual data contains personal information not only relating to health parameters but also to details, such as credit card information or even party affiliation. With regard to portable or wearable medical AI devices (eg, heart rate monitors or sleep trackers), such data may be recorded, tracked, and even stored through commercially available software for social media such as Facebook and other apps. The potential leak of personal information presents a substantial challenge to the retrieval of medical data.33 A human-centered approach to medical AI should be adopted; for example, in federated learning (FL), all personal data should be stored in a system unique to each individual.34 Taiwan AI Labs presents a model example of patient privacy management. The company, founded by Yi-Chin Tu in 2017, is the first Asian human-centered open AI laboratory. It aims to protect the privacy of individuals while gathering essential information through the precise filtering of individual information.35 The goal of this innovative lab is to develop models to help humans to live a good life. This is an ideal goal for precision medicine. Following cross-trial and validation, the cross-hospital network and central platform can share valuable data and benefit from sharing large-scale clinical research.


To identify the SARS-CoV-2 virus either at the genetic or protein level (detection of viral substances, Section 2.1; 2.2) and to confirm COVID-19 cases through the evaluation of illness status (evaluated with clinical images and related symptoms, Section 2.3; 2.4) are both key tasks to quantify confirmed cases for pandemic monitoring.36 Through this combined approach, the majority of SARS-CoV-2 cases can be identified (Fig. 2).

Fig. 2:
Smart healthcare for mild and severe COVID-19 diseases. For viral detection, RT-qPCR is the gold standard (hospital-only). For home testing, the FDA-approved fast flow kit (through antibody detection) allows for rapid detection of SARS-CoV-2. Although the accuracy of fast flow home testing is low, RT-LAMP and CRISPR have acceptable accuracy and are alternatives. For the symptom-derived signs of COVID-19 illness, wearable devices that can measure some physiological signatures including heartbeat rate, respiratory status, and even oxygen saturation can be connected with smartphone apps to supply prompt warning. These are more readily available than standard medical imaging confirmed by CXR or CT scan (hospital-only). Transferring such data to medical AI with cloud calculation ability can further facilitate prognosis prediction and provide medical suggestions. AI = artificial intelligence; CRISPR = clusters of regularly interspaced short palindromic repeats; CT = computed tomography; CXR = chest X-ray; FDA = food and drug administration; RT-LAMP = real-time loop-mediated isothermal amplification; RT-qPCR = reverse-transcription polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2.

2.1. Detection of SARS-CoV-2

The prevailing view is that detection of SARS-CoV-2 within the human body is the ultimate diagnostic definition of a confirmed case, regardless of patient symptomology. The most direct method of identifying SARS-CoV-2 infection is to detect virus-derived substances (either nucleic acids or proteins) through the analysis of nasopharyngeal swabs, saliva, or other sources.37 The gold standard of viral detection is real-time reverse-transcription polymerase chain reaction (real-time RT-PCR). This is a nucleic acid amplification-based test (NAAT).38 The virus can be detected through the design of specific primer sequences at the conserved viral genome regions of certain gene locations, namely (1) the RNA-dependent RNA polymerase (RdRp, within ORF1ab), (2) the envelope protein (E), and (3) the nucleocapsid protein (N) within SARS-CoV-2.39 In addition, the cycle threshold (Ct) values generated through RT-qPCR testing can provide information on the viral infection status of patients and thus reveal chronological and geographic information (Table 1).

Table 1 - Viral detections for severe acute respiratory syndrome coronavirus 2 and the combination with medical artificial intelligence
Nucleic acid-amplification detection
Methods RT-qPCR
Place to exam Hospital
Sampling Naso swabs (some saliva)
LOD (sensitivity) 97% (reference 42) Ct < 40
Accuracy (specificity) gold standard
Time to get the data 3.5-4 hours (reference 42)
Commercial availability Roche, LabCorp, PerkinElmer (RUO), Mesa Biotech, Cepheid, Qiagen, Thermo Fisher, BioRAD (Research Use Only, RUO)
Commercial with medical AI Cobas® SARS-CoV-2 6,800 (96 results in 3h; 384 in 8h) and 8,800 systems (1056 in 8h) (Roche Molecular Diagnostics, Pleasanton, CA, USA)Abbott Molecular (Des Plaines, IL, USA) the m2000 system 96 samples simultaneously, 470 test results in ~24 h; sensitivity (93%); specificity (100%)
Nucleic acid-isothermal amplification detection
Methods RT-LAMP CRISPR-Cas 12 or Cas13
Place to exam Home (hospital for mega usage)
Sampling Naso swabs (some saliva)
LOD (sensitivity) 10 (iLACO, reference 40), 30 (PD-LAMP, reference 47) copies/µL 97% (reference 48); 1 copies/µL (Cas13a), 10 copies/µL (Cas12a/b) (reference 41)
Accuracy (specificity) 89.9% (reference 40) iLACO 95% (reference 48)
Time to get the data 10 min (NEAR, FDA-EUA), 20 min (iLACO), 30 min (PD-LAMP) 30-90 min
Commercial availability Atila Biosystems, Abbott Mammoth Biosciences (DETECTR), SHERLOCK Biosciences (FDA-EUA)
Protein-based rapid home testing Indoor air quality monitoring Exhaled breath detection
Methods Fast flow detection (antibody) Space Monitoring Device Wearable
Place to exam Home only on-site Home, work, hospital, on-site Home, work, hospital, on-site
Sampling Naso swabs (some saliva) Air Air (collected from mask)
Time to get the data 15 min 30 mins N.A.
Commercial availability Fast flow kit (FDA-approved) N.A. N.A. (reference 53)
AI = artificial intelligence; CRISPR = clusters of regularly interspaced short palindromic repeats; FDA = food and drug administration; LOD = limit of detection; PD = particle diffusometry; RT-LAMP = real-time
loop-mediated isothermal amplification; RT-qPCR = reverse-transcription polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2.
The cited references are: 40, 41, 42, 47, 48, 53

The RT-qPCR assay is time-consuming (from <8 hours to 1 or 2 days), and the performance of RT-qPCR depends on the hospital examination capacity. Therefore, reliance on this gold standard alone does not produce rapid results. Alternatively, loop-mediated isothermal amplification (LAMP) and related methods, such as real-time LAMP (RT-LAMP) and LAMP sequencing, and advanced biotechniques such as clusters of regularly interspaced short palindromic repeats (CRISPR)-Cas (Cas3, Cas9, Cas12, and Cas13)-based biosensors can also perform NAAT-like SARS-CoV-2 detection (the primer design can be the same or different to RT-qPCR) within 40 minutes or less.40,41 Because NAATs require a special platform to read and process the data of clinical samples, the popularity of LAMP-related and CRISPR-related techniques is limited despite their testing speeds being as low as 15 minutes.42

The use of antibodies against viral proteins such as spike constitute protein-based tests for the rapid detection of SARS-CoV-2. This method requires even less time than NAATs.43 This method may exceed the popularity of RT-qPCR. Such tests can be adapted from existing US Food and Drug Administration (FDA)-approved lateral-flu test kits for rapid home testing (eg, the SARS-CoV-2 Antigen Self-Test Nasal manufactured by F. Hoffmann-La Roche, Abbott, or other companies).44 However, a drawback of these protein-based tests is insufficient accuracy when compared with RT-qPCR.45,46 Asymptomatic false-negative cases may lead to the undetected transmission of VOCs. Mutated spike proteins can evade vaccination, the human immune system, and also protein-based rapid-testing antibodies. Despite these disadvantages, home testing as a rapid diagnostic tool remains a useful method of pandemic monitoring.

2.2. Smartphone-based viral detection

Medical AI used with smartphones and advanced biotechnology has great potential for viral infection monitoring. New optoelectronic devices integrated with a test chamber or with home testing papers can serve as fast and powerful viral detectors. These devices can be used in conjunction with cloud computing for both the retrieval and storage of detected signals. The information provided through this method is valuable for monitoring SARS-CoV-2 transmission.

LAMP-related rapid home testing has been approved for emergency use authorization by the FDA for COVID-19 (eg, Abbott ID NOW and Cue COVID-19 Diagnostic Test). RT-LAMP used in conjunction with the novel technique of particle diffusometry (PD) displayed excellent SARS-CoV-2 detection capability; a limit of detection (LOD) of 30 virus particles per μL or less (35 × 104 viral particles per mL) within 35 minutes in saliva was reported. In addition, PD-LAMP is compatible with digital connectivity for point-of-care testing.47 For advanced NAAT, CRISPR-Cas12a integrated with smartphone-based fluorescence devices (with low-cost three-dimensional-printed housing and optics) under a DL/ML-driven environment has been applied for the detection of SARS-CoV-2.48 The LOD of this device is approximately 6.25 RNA copies/μL (based on laboratory samples). It exhibited a test accuracy of 95% and a sensitivity of 97% on 96 nasopharyngeal swab samples with transmissible viral loads (Table 1). Furthermore, without nucleic acid amplification and reverse transcription (not NAAT, but nucleic acids are still detected), a novel portable electrochemical smartphone biosensor was developed to achieve the ultrasensitive detection of SARS-CoV-2 RNA.49 This method is based on the super-sandwich recognition strategy that utilizes calixarene-functionalized graphene oxide as the key biosensor. The LOD of this method to a clinical specimen is 200 copies/mL. This LOD is reported to be the lowest among the published RNA measurements to date.

For the detection of viral protein level, a smartphone-assisted immunoassay-based method integrated with magnetic beads and colorimetry was established for testing saliva and nasopharyngeal swab samples.50 The LOD of this 96-well wax-printed paper plate assay is 0.1 μg/mL. A Spotxel free-charge app for use with this specific assay was also developed. The primary advantage of this cost-effective method is accurate detection within 30 minutes, even when used with samples with low viral load (Ct numbers as high as 30 have been reported from RT-PCR results). It is also crucial to evaluate the quantity and quality of SARS-CoV-2 neutralizing antibodies in blood samples of patients with COVID-19 and vaccinated individuals. A smartphone can serve as track-etched membrane microplate readers for high-throughput quantitative measurement of antibodies.51 With regard to the application of novel photoelectric technology, Kawasaki et al52 described a label-free smartphone-based optical sensor for SARS-CoV-2 testing. The imprinted photonic crystal film inside this optical sensor is functionalized with an anti-SARS-CoV-2 spike protein antibody and thus can conduct the corresponding immunoassay. This novel device can detect SARS-CoV-2 at a low LOD of 429 fg/mL from saliva samples.

Knowledge of the viral genome or proteins from tested samples enables the physical identification of the virus within the human body. However, the results of viral detection (Sections 2.1, 2.2) can be influenced by many factors, such as sampling error and infection status (see the following Sections 2.3, 2.4). For example, the viral load of each infected person may vary depending on the day of infection and vary according to individual immune ability (regardless of vaccination status). In reality, many patients report negative home rapid-testing results (when no reason to undergo RT-qPCR at their own expense exists) while reporting novel COVID-19 symptoms such as ageusia, anosmia, and sharp sore throat without fever and cough (Fig. 2). Whether such cases (rapid test negative but symptom positive) should be viewed as confirmed cases is a matter of debate. False viral detection results statistically influence confirmed case data.

For sample collection, which generally uses nasopharyngeal swabs or saliva, a unique, disposable, and inexpensive wearable collection device was designed for the on-site airborne viral collection.53 It can adhere to the inside of various masks (textile, surgical, KN95, or N95) to collect the virus exhaled in the breath. The efficacy of this collection method was confirmed with the following detection methods: RT-qPCR, RT-LAMP, and antibody-based dot blot assays. Such a device will be highly useful when smartphone-based detection methods (eg, smart RT-LAMP and smart CRISPR) or rapid paper home testing have been developed for widespread application.

2.3. Specific illness of COVID-19: medical imaging and medical AI

Although the majority of suspected cases that are asymptomatic or exhibit few symptoms seem to be mild cases (or may not be recognized as confirmed cases), such cases may not necessarily have good prognoses. The oversight of such possible infections might lead to the underground transmission of SARS-CoV-2. In the absence of notification, such cases can constitute unknown sources of transmission. Conversely, severe cases exhibiting serious symptoms cause major concern because the virus threatens the lives of patients with comorbid conditions, such as those with chronic diseases, metabolic syndrome (including diabetes), or cancers.54 Aggressive treatment is required for such patients, either in the hospital or at home, to achieve the goal of zero severe cases and to reduce the death rate among severe cases.

Confirmation of a COVID-19 diagnosis generally takes place in hospitals but not exclusively so. The clinical diagnosis of COVID-19 can be confirmed using various physical examinations and medical imaging techniques—for example, chest X-ray (CXR) imaging and computed tomography (CT) scanning.55 CXR imaging is a simple method for diagnosing pneumonia. CT scanning allows for rapid diagnosis.56 Similar to RT-qPCR, medical imaging requires medical resources and personnel. The continuation of the pandemic places a burden on such resources.

AI-based classification of pneumonia during the COVID-19 pandemic can be performed using ML-based automated systems with CXR or CT scan images acquired from hospitals and analyzed using medical AI such as NHIA-TAIMedimg, DeepCheX, GoogleNet, Choquet Fuzzy with Xception, and others.57,58 The preliminary results are then returned to the Centers for Disease Control for the final decision.

2.4. Specific illness of COVID-19: wearable devices and medical AI

The use of physical checks similar to CXR and CT imaging, for example—wearable devices that can detect COVID-19 outside the hospital, is another potential monitoring technique. Through the redefinition of the specific criteria of COVID-19, medical AI can use DL and ML to recalculate acceptable physical parameters for further applications. Medical AI used in conjunction with big data analysis allows for the large-scale continuous retrieval of data for simulation and prediction.57 Because CXR and CT can only be performed in clinical settings and only specialist radiologists may perceive subtle anomalies, other tools such as wearable medical devices or sensors have the potential for the rapid and accurate confirmation of SARS-CoV-2 infection at home.59,60 Through the recording of subtle physiological changes such as heart rate, respiratory rate, and skin temperature and through comparison with confirmed cases, early digital biomarkers of infection can be determined (as discernible symptoms of suspected COVID-19 cases). However, a review article analyzed numerous published papers (MEDLINE, Web of Science, etc.), protocols, and data (Embase, Cochrane Central Register of Controlled Trials, International Clinical Trials Registry Platform,, etc.) investigating the use of wearable devices for this purpose. They unexpectedly reported that the majority of this research exhibited a moderate risk of bias, and the accuracy of their results is doubtful.60 It is still not possible to conclude that the aforementioned three physiological parameters can be used as new gold standard parameters for the diagnosis of COVID-19. Medical AI prediction can test the efficacy of such parameters.

Conversely, fingertip pulse oximeters are rapid point-of-care and home care medical devices for the measurement of oxygen saturation values (proportions of oxygenated and deoxygenated hemoglobin) to evaluate the condition of the respiratory system. Oxygen saturation measurement is particularly crucial during the SARS-CoV-2 pandemic because many infected individuals are suspected to be in a silent hypoxic state without overt dyspnea. In the UK, pulse oximeters are provided as home care devices to high-risk patients with COVID. Fingertip pulse oximeters can be an integral component of IoMT against COVID. One such cloud-based monitoring system has been implemented by National Yang Ming Chiao Tung University in Taiwan. The fingertip pulse oximeters are connected wirelessly to smartphones using Bluetooth technology. The app then communicates with the cloud-based monitoring system to send out warning messages to patients when abnormal patterns of measurements are detected.

Further, several smartphone apps have been developed, such as the Zoe COVID study app, the CoroNotes app (University of Tübingen), and the COVID Control App (John Hopkins University). These are either designed for the detection of age-based and sex-based early symptom discrepancies, to enhance user well-being, or to record the daily body temperature of patients. Lovey et al61 recently demonstrated the possibility of remotely monitoring SARS-CoV-2 confirmed cases by conducting a prospective cohort study using a smartphone app called Illness Tracking in Tested Persons. Through this tracking system, significant high odds ratios of difficulty breathing (3.35), a reduced sense of taste (ageusia, 5.45), and a reduced sense of smell (anosmia, 18.24) can be detected. These top three symptoms display COVID-19-specific signs (compared with the SARS-CoV-2 negative set). Fatigue was the only single symptom to considerably influence the daily activities of confirmed cases. This type of tracking system or another platform such as the Blue Dot AI system23 may be applied in various settings including schools, hospitals, companies, sport clubs, or music groups to detect signs of developing infection (once the alternative three can be confirmed as new gold standards of COVID-19 diagnosis). Such systems may be used to control transmission and predict epidemiological trends.


To date, SARS-CoV-2 has infected more than 550 million people globally. The vast spread of the disease over 2–3 years has drastically increased the burden on medical service providers who lack the capacity to cope with a pandemic of this scale. In some areas, medical systems have been overburdened to the point of breakdown. COVID-19 presents diverse symptoms with varying severity among different people, ranging from no symptoms to mild flu-like symptoms such as coughing and sneezing and to severe diseases such as pneumonia, thrombosis, and systemic inflammatory syndrome (Fig. 2). For children, symptoms include respiratory symptoms, gastrointestinal conditions such as abdominal pain, Kawasaki-like conditions, encephalitis, cardiovascular conditions such as tachycardia and hemodynamic shock, and pediatric inflammatory multisystem syndrome temporally associated with SARS-CoV-2 (PIMS-TS; i.e., a multisystem inflammatory syndrome in children).17 Facilities such as negative pressure rooms and intensive care units are in limited supply, and hence, medical resources must be used wisely, and patients should be triaged according to the severity of their symptoms.

The reliable prediction of a patient’s clinical courses can enable personally optimized patient care and the efficient use of medical resources (Fig. 3). Information from the time of infection may be useful for predictions. First, the viral strain (Alpha, Beta, Delta, or Omicron) and viral load (determined by RT-PCR) can affect the clinical course. Second, the exacerbation of preexisting comorbidities also affects the clinical course. Third, age, sex, genetics, ethnicity, serum biochemistry profiles, immunological profiles, and environmental factors may also affect the clinical course. The utilization of data related to viral, host, and environmental factors may effectively predict whether an infected person will exhibit mild symptoms or severe symptoms requiring hospitalization or even intensive care (Fig. 2). The two types of outcomes are not proportional. Mild cases account for >90% of infections, and severe cases account for <10%.62 Hence, the sensitivity and specificity of prediction are both crucial. Big data gathered from the information hub can be integrated and calculated through DL to provide valuable suggestions for the treatment of COVID-19 cases.

Fig. 3:
Successive predictors for clinical courses of patients with COVID-19. A roadmap of the COVID-19 clinical course with three predictors incorporated to facilitate optimized patient care and medical resource allocation. The predictors can be implemented as apps in mobile devices or integrated into the computer systems of hospitals.

A laboratory study conducted in South Korea with 561 adult patients with COVID-19 demonstrated that age, sex, and serum concentrations of C reactive protein, lactate dehydrogenase, and hemoglobin are major independent factors associated with the progression to severe pneumonia, which is defined as resting oxygen saturation <93%, PaO2/FiO2 ≤300 mmHg, or the need for mechanical ventilation.63 The patients were treated at Keimyung University Daegu Dongsan Hospital (KDDH), and the constructed KDDH score using five independent factors achieved an area under the receiver operating characteristic curve of 88.4% in the training dataset and 82.8% in the validation dataset. Apart from age and sex, the independent factors are derived from common laboratory examinations of peripheral blood. The COVID-19 care apps in intelligent devices, such as Zoe, CoroNotes, and COVID Control (as described in Section 2.4), or other similar apps should incorporate a severe disease risk predictor that allows users to input the values of the five factors and receive a prediction (Fig. 3).

3.1. Mild cases with prognosis prediction

Mild symptoms usually resolve within a few days. After the acute symptoms resolve, some symptoms may appear or reappear several weeks after the infection, including palpitations, tachycardia, shortness of breath or breathing difficulties, tightness of the chest, muscle aches, loss of appetite, diarrhea, fatigue, postexertional malaise, brain fog, and memory issues. These symptoms are collectively known as long COVID.64 Approximately one in five patients with COVID develop long COVID.65 In Britain, the Zoe (as described in Section 2.4) smartphone-based COVID Symptom Study app was developed to collect self-reported long COVID conditions.66 Status at the time of infection may be used to predict the long-term prognosis with or without long COVID conditions.

A multiomics study was conducted to reveal factors associated with long COVID. It was found that type 2 diabetes, SARS-CoV-2 viral load, and the detection of Epstein–Barr virus and specific autoantibodies are correlated with long COVID.67 An app for mobile devices should be developed for users to input this information and calculate their risk using an established algorithm (Fig. 3).

3.2. Severe cases with prognosis prediction

The clinical course of severe COVID-19 can be categorized into three phases: early infection phase, pulmonary stage, and hyperinflammatory phase.68 Different treatments are required in different phases. For those with severe disease, whether the patient requires intensive care is a major concern.

A predictive model for 30-day mortality was recently constructed based on cohorts of 4035 and 2126 hospitalized and laboratory-confirmed patients, and these were used for the model construction and validation stages, respectively.69 The patients were recruited from 127 Spanish hospitals. Factors employed in the model include age, age-adjusted oxygen saturation, neutrophil-to-lymphocyte ratio, estimated glomerular filtration rate according to the chronic kidney disease epidemiology collaboration equation, dyspnea, and sex. The constructed model achieved 82.2% and 84.5% area under the receiver operating characteristic curves in the model construction and validation cohorts, respectively. The predictive model could be implemented in a hospital software system or as an app for smartphone devices. Currently, oxygen saturation measurements are mostly measured by electronic point-of-care devices (also mentioned in Section 2.4). These devices can be readily designed to send measurements wirelessly to hospital systems or mobile devices where the apps are installed (Fig. 3).


When transmission remains high and quarantine measures no longer slow viral transmission, it may be necessary to diagnose cases through either of the aforementioned techniques (Section 2; Fig. 2). The combined monitoring of the medical information and health activities of individuals meet the FL requirement of respecting patient privacy. AI-ML with advanced software as medical device (SaMD) can be developed to provide accurate predictions of COVID-19 status on a large-scale (also discussed in Section 3).70 The Taiwan Social Distancing App is a practical example of SaMD for COVID-19.71 Each datapoint collected through Bluetooth from individuals is decentralized before being reported as a confirmed case with geographic distance between the patients with COVID-19 and the users of concern. However, it seems that insufficient data are provided for the users of this app. This is primarily because of obedience to primary protection; for example, it accounts for the lack of data concerning when and where the user has been in short-distance “contact” with unknown numbers of confirmed cases. Further improvements of such apps would provide valuable information on the transmission of COVID-19 for the users while still maintaining individual privacy.

In the period following the COVID-19 pandemic, it is likely that SARS-CoV-2 will be a permanent presence in our lives, similar to the pandemic of influenza A. Thus, we should focus on prognosis prediction (Section 3, Fig. 3) and effective medications. Generally, the medication process contains four steps after the confirmation of illness: (1) prescription (smart healthcare systems can help to confirm the correct prescription), (2) dispensation of medication (a robot can assist in the provision of correct medication), (3) administration (confirm correct drug given, BCMA, RFID, NFC), (4) compliance (a smartphone app can remind patients to medicate, and the learning health system can prevent adverse drug reactions). Medical AI can assist in all of these steps to lower the risk of medical and follow-up errors. The future AI hospital service will integrate disease prevention (including precision health assistance and prehospital consultation plus prediction), AI clinics (assisted diagnosis and treatments), AI nursing (assisted care), and a smart workflow for the entire medical care and telemedicine system (when the burden of a hospital is considerable enough that some mild cases should not present at the hospital but rather receive remote healthcare).

Telemedicine combines mature medical AI with versatile IoMT (eg, AI-assisted medical documentation, voice recognition and decentralization, and wearable monitoring system)71,72 and the analysis of big data through AI-DL/ML cloud computing to develop strategies and algorithms. This may combat the shortage in human resources of hospitals and clinics and has great potential to help patients with COVID-19 at home.73 The IoMT being linked to the available data from hospitals or personal smartphones will increase the speed of case confirmation and of the following steps, such as quarantine or discharge, prognosis prediction, and suggestions for individualized treatments. With the help of telemedicine systems, physicians can first confirm the results of FDA-approved lateral flow home testing and the reported symptoms of suspected cases through smartphone apps online with clinics. Doctors will be responsible for each confirmed case. Adherence to medication can be assisted by medical AI and IoMT.

Big data containing the identification information of individuals will be stored and represented through an Internet interface and can be used for personal medications, treatments, and possible prognosis prediction and tracing (Figs. 2 and 3). Doctors responsible for case confirmations can make decisions regarding the prescription of essential medications and, with the help of IoMT, delivery of anti-SARS-CoV-2 drugs to individual patients. Such measures will allow for long-term control of COVID-19. In the future, medical AI can also contribute to new vaccine design to combat new variants beyond Delta and Omicron BA.2–BA.5 and also to the development of antiviral drugs.

Table 2 - Diagnostic and monitoring methods for COVID-19 and the commercial (personal) availability
Medical imaging Chest X-ray (CXR) CT scanning
Place to exam Hospital Hospital
Sampling In vivo on-site In vivo on-site
Time to get the data N.A. N.A.
Commercial (personal) availability N.A. N.A.
Wearable electronics device Respiratory rate Heart rate Skin temperature Oxygen saturation
Place to exam Home, work, hospital Home, work, hospital Home, work, hospital Home, work, hospital
Sampling In vivo on-site In vivo on-site In vivo on-site In vivo on-site
Time to get the data real-time real-time real-time real-time
Commercial (personal) availability WHOOP Strap; Oura Ring Apple WatchE; 4 wristband; Fitbit;WHOOP Strap; Oura Ring Fitbit; Oura Ring Apple Watch; Fitbit; Oura Ring
The cited references are: 74,75,76,77,78.


This study was supported by Ministry of Science and Technology of Taiwan, MOST (105-2320-B-075-002, 108-2745-8-075-001, 110-2320-B-075-005) and Taipei Veterans General Hospital (V-110-018, V111C-088).


1. Wu F, Zhao S, Yu B, Chen YM, Wang W, Song ZG, et al. A new coronavirus associated with human respiratory disease in China. Nature 2020;579:265–9.
2. Gordon DE, Jang GM, Bouhaddou M, Xu J, Obernier K, White KM, et al. A SARS-CoV-2 protein interaction map reveals targets for drug repurposing. Nature 2020;583:459–68.
3. Rahimi A, Mirzazadeh A, Tavakolpour S. Genetics and genomics of SARS-CoV-2: a review of the literature with the special focus on genetic diversity and SARS-CoV-2 genome detection. Genomics 2020;113:1221–32.
4. Shang J, Wan Y, Luo C, Ye G, Geng Q, Auerbach A, et al. Cell entry mechanisms of SARS-CoV-2. Proc Natl Acad Sci USA 2020;117:11727–34.
5. Chang TJ, Yang DM, Wang ML, Liang KH, Tsai PH, Chiou SH, et al. Genomic analysis and comparative multiple sequences of SARS-CoV2. J Chin Med Assoc 2020;83:537–43.
6. Tsai PH, Wang ML, Yang DM, Liang KH, Chou SJ, Chiou SH, et al. Genomic variance of open reading frames (ORFs) and spike protein in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). J Chin Med Assoc 2020;83:725–32.
7. Mlcochova P, Kemp SA, Dhar MS, Papa G, Meng B, Ferreira IATM, et al.; Indian SARS-CoV-2 Genomics Consortium (INSACOG). SARS-CoV-2 B.1.617.2 Delta variant replication and immune evasion. Nature 2021;599:114–9.
8. Hachmann NP, Miller J, Collier AY, Ventura JD, Yu J, Rowe M, et al. Neutralization escape by SARS-CoV-2 omicron subvariants BA.2.12.1, BA.4, and BA.5. N Engl J Med 2022;387:86–8.
9. Gobeil SM, Henderson R, Stalls V, Janowska K, Huang X, May A, et al. Structural diversity of the SARS-CoV-2 Omicron spike. Mol Cell 2022;82:2050–2068.e6.
10. Araf Y, Akter F, Tang YD, Fatemi R, Parvez MSA, Zheng C, et al. Omicron variant of SARS-CoV-2: genomics, transmissibility, and responses to current COVID-19 vaccines. J Med Virol 2022;94:1825–32.
11. Menni C, Valdes A, Polidori L, Antonelli M, Penamakuri S, Nogal A, et al. Symptom prevalence, duration, and risk of hospital admission in individuals infected with sARS-CoV-2 during periods of omicron and delta variant dominance: a prospective observational study from the ZOE COVID Study. Lancet 2022;399:1618–24.
12. Vaira LA, Salzano G, Deiana G, De Riu G. Anosmia and ageusia: common findings in COVID-19 patients. Laryngosc 2020;130:1787.
13. Ma Q, Liu J, Liu Q, Kang L, Liu R, Jing W, et al. Global percentage of asymptomatic SARS-CoV-2 infections among the tested population and individuals with confirmed COVID-19 diagnosis: a systematic review and meta-analysis. JAMA Netw Open 2021;4:e2137257.
14. Yang DM, Chang TJ, Wang ML, Tsai PH, Lin TH, Wang CT, et al. Hunting severe acute respiratory syndrome coronavirus 2 (2019 novel coronavirus): from laboratory testing back to basic research. J Chin Med Assoc 2020;83:524–6.
15. Liang KH, Chang TJ, Wang ML, Tsai PH, Lin TH, Wang CT, et al. Novel biosensor platforms for the detection of coronavirus infection and severe acute respiratory syndrome coronavirus 2. J Chin Med Assoc 2020;83:701–3.
16. Tu YF, Chien CS, Yarmishyn AA, Lin YY, Luo YH, Lin YT, et al. A review of SARS-CoV-2 and the ongoing clinical trials. Int J Mol Sci 2020;21:2657.
17. Liang KH, Hung KF, Wang ML, Chang TJ, Cheng YF, Chiang SH, et al. SARS-CoV-2 vaccines in children and adolescents: can immunization prevent hospitalization? J Chin Med Assoc 2022;85:891–5.
18. Wu Q, Suo C, Brown T, Wang T, Teichmann SA, Bassett AR. INSIGHT: a population-scale COVID-19 testing strategy combining point-of-care diagnosis with centralized high-throughput sequencing. Sci Adv 2021;7:eabe5054.
19. Singh P, Kaur R. An integrated fog and artificial intelligence smart health framework to predict and prevent COVID-19. Glob Transit 2020;2:283–92.
20. Dwivedi R, Mehrotra D, Chandra S. Potential of internet of medical things (IoMT) applications in building a smart healthcare system: a systematic review. J Oral Biol Craniofac Res 2022;12:302–18.
21. Rani SS, Selvakumar S, Pradeep Mohan Kumar K, Thanh Tai D, Dhiravida Chelvi E. Internet of medical things (IoMT) with machine learning–based COVID-19 diagnosis model using chest X-ray images. Data Sci for COVID-19 2021:627–41.
22. Roy PK, Kumar A. Early prediction of COVID-19 using ensemble of transfer learning. Comput Electr Eng 2022;101:108018.
23. Kollu PK, Kumar K, Kshirsagar PR, Islam S, Naveed QN, Hussain MR, et al. Development of advanced artificial intelligence and IoT automation in the crisis of COVID-19 detection. J Healthc Eng 2022;2022:1987917.
24. Naik N, Hameed BMZ, Shetty DK, Swain D, Shah M, Paul R, et al. Legal and ethical consideration in artificial intelligence in healthcare: who takes responsibility? Front Surg 2022;9:862322.
25. Ewuoso C, Hall S, Dierickx K. How do healthcare professionals respond to ethical challenges regarding information management? A review of empirical studies. Glob Bioeth 2021;32:67–84.
26. Kelly CJ, Karthikesalingam A, Suleyman M, Corrado G, King D. Key challenges for delivering clinical impact with artificial intelligence. BMC Med 2019;17:195.
27. London AJ. Artificial intelligence in medicine: overcoming or recapitulating structural challenges to improving patient care? Cell Rep Med 2022;3:100622.
28. Davenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J 2019;6:94–8.
29. Yadav S, Kaushik A, Sharma S. Simplify the difficult: artificial intelligence and cloud computing in healthcare. In: Verma JK, Saxena D, González-Prida V, editors. IoT and Cloud Computing for Societal Good. EAI/Springer Innovations in Communication and Computing. Switzerland: Springer, Cham; 2022, p. 101–124. Available at Accessed July 21, 2022.
30. Ahuja AS. The impact of artificial intelligence in medicine on the future role of the physician. PeerJ 2019;7:e7702.
31. Dey D, Slomka PJ, Leeson P, Comaniciu D, Shrestha S, Sengupta PP, et al. Artificial intelligence in cardiovascular imaging: JACC state-of-the-art review. J Am Coll Cardiol 2019;73:1317–35.
32. Pesapane F, Codari M, Sardanelli F. Artificial intelligence in medical imaging: threat or opportunity? radiologists again at the forefront of innovation in medicine. Eur Radiol Exp 2018;2:35.
33. Murdoch B. Privacy and artificial intelligence: challenges for protecting health information in a new era. BMC Med Ethics 2021;22:122.
34. Shneiderman Ben. Human-centered artificial intelligence: reliable, safe and trustworthy. (February 23, 2020). Available at Accessed July 28, 2022.
35. Huang KC, Huang CS, Su MY, Hung CL, Ethan Tu YC, Lin LC, et al. Artificial intelligence aids cardiac image quality assessment for improving precision in strain measurements. JACC Cardio Imag 2021;14:335–45.
36. Suthar AB, Schubert S, Garon J, Couture A, Brown AM, Charania S. Coronavirus disease case definitions, diagnostic testing criteria, and surveillance in 25 countries with highest reported case counts. Emerg Infect Dis 2022;28:148–56.
37. Iwasaki S, Fujisawa S, Nakakubo S, Kamada K, Yamashita Y, Fukumoto T, et al. Comparison of SARS-CoV-2 detection in nasopharyngeal swab and saliva. J Infect 2020;81:e145–7.
38. Esbin MN, Whitney ON, Chong S, Maurer A, Darzacq X, Tjian R. Overcoming the bottleneck to widespread testing: a rapid review of nucleic acid testing approaches for COVID-19 detection. RNA 2020;26:771–83.
39. Corman VM, Landt O, Kaiser M, Molenkamp R, Meijer A, Chu DK, et al. Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR. Euro Surveill 2020;25:2000045.
40. Yu L, Wu S, Hao X, Dong X, Mao L, Pelechano V, et al. Rapid detection of COVID-19 coronavirus using a reverse transcriptional loop-mediated isothermal amplification (RT-LAMP) diagnostic platform. Clin Chem 2020;66:975–7.
41. Nouri R, Tang Z, Dong M, Liu T, Kshirsagar A, Guan W. CRISPR-based detection of SARS-CoV-2: a review from sample to result. Biosens Bioelectron 2021;178:113012.
42. Mardian Y, Kosasih H, Karyana M, Neal A, Lau CY. Review of current COVID-19 diagnostics and opportunities for further development. Front Med 2021;8:615099.
43. Habli Z, Saleh S, Zaraket H, Khraiche ML. COVID-19 in-vitro diagnostics: state-of-the-art and challenges for rapid, scalable, and high-accuracy screening. Front Bioeng Biotechnol 2021;8:605702.
44. Rahimi H, Salehiabar M, Barsbay M, Ghaffarlou M, Kavetskyy T, Sharafi A, et al. CRISPR systems for COVID-19 diagnosis. ACS Sens 2021;6:1430–45.
45. Vandenberg O, Martiny D, Rochas O, van Belkum A, Kozlakidis Z. Considerations for diagnostic COVID-19 tests. Nat Rev Microbiol 2021;19:171–83.
46. Diao B, Wen K, Zhang J, Chen J, Han C, Chen Y, et al. Accuracy of a nucleocapsid protein antigen rapid test in the diagnosis of SARS-CoV-2 infection. Clin Microbiol Infect 2021;27:289.e1–4.
47. Colbert AJ, Lee DH, Clayton KN, Wereley ST, Linnes JC, Kinzer-Ursem TL. PD-LAMP smartphone detection of SARS-CoV-2 on chip. Anal Chim Acta 2022;1203:339702.
48. Samacoits A, Nimsamer P, Mayuramart O, Chantaravisoot N, Sitthi-amorn P, Nakhakes C, et al. Machine learning-driven and smartphone-based fluorescence detection for CRISPR diagnostic of SARS-CoV-2. ACS Omega 2021;6:2727–33.
49. Zhao H, Liu F, Xie W, Zhou TC, OuYang J, Jin L, et al. Ultrasensitive supersandwich-type electrochemical sensor for SARS-CoV-2 from the infected COVID-19 patients using a smartphone. Sens Actuators B Chem 2021;327:128899.
50. Fabiani L, Mazzaracchio V, Moscone D, Fillo S, De Santis R, Monte A, et al. Paper-based immunoassay based on 96-well wax-printed paper plate combined with magnetic beads and colorimetric smartphone-assisted measure for reliable detection of SARS-CoV-2 in saliva. Biosens Bioelectron 2022;200:113909.
51. Wang C, Wu Z, Liu B, Zhang P, Lu J, Li J, et al. Track-etched membrane microplate and smartphone immunosensing for SARS-CoV-2 neutralizing antibody. Biosens Bioelectron 2021;192:113550.
52. Kawasaki D, Yamada H, Sueyoshi K, Hisamoto H, Endo T. Imprinted photonic crystal-film-based smartphone-compatible label-free optical sensor for SARS-CoV-2 testing. Biosen 2022;12:200.
53. Soto F, Ozen MO, Guimarães CF, Wang J, Hokanson K, Ahmed R, et al. Wearable collector for noninvasive sampling of SARS-CoV-2 from exhaled breath for rapid detection. ACS Appl Mater Interf 2021;13:41445–53.
54. Hu B, Guo H, Zhou P, Shi ZL. Characteristics of SARS-CoV-2 and COVID-19. Nat Rev Microbiol 2021;19:141–54.
55. Martín J, Tena N, Asuero AG. Current state of diagnostic, screening and surveillance testing methods for COVID-19 from an analytical chemistry point of view. Microchem J 2021;167:106305.
56. Dai WC, Zhang HW, Yu J, Xu HJ, Chen H, Luo SP, et al. CT imaging and differential diagnosis of COVID-19. Can Assoc Radiol J 2020;71:195–200.
57. Nasser N, Emad-Ul-Haq Q, Imran M, Ali A, Razzak I, Al-Helali A. A smart healthcare framework for detection and monitoring of COVID-19 using IoT and cloud computing. Neural Comput Appl 2021;1–15.
58. Mahanty C, Kumar R, Asteris PG, Gandomi AH. COVID-19 patient detection based on fusion of transfer learning and fuzzy ensemble models using CXR images. Appl Sci 2021;11:11423.
59. Hassantabar S, Stefano N, Ghanakota V, Ferrari A, Nicola GN, Bruno R, et al. CovidDeep: SARS-CoV-2/COVID-19 test based on wearable medical sensors and efficient neural networks. IEEE Trans Cons Electr 2021;67:244–56.
60. Mitratza M, Goodale BM, Shagadatova A, Kovacevic V, van de Wijgert J, Brakenhoff TB, et al. The performance of wearable sensors in the detection of SARS-CoV-2 infection: a systematic review. Lancet Digit Health 2022;4:e370–83.
61. Lovey T, Bielecki M, Gültekin N, Stettbacher A, Muggli F, Stanga Z, et al. Illness tracking in SARS-CoV-2 tested persons using a smartphone app: a non-interventional, prospective, cohort study. New Micro New Infect 2022;46:100967.
62. Wang M, Wu D, Liu C-H, Li Y, Hu J, Wang W, et al. Predicting progression to severe COVID-19 using the PAINT score. BMC Infect Dis 2022;22:498.
63. Lee JY, Nam B-H, Kim M, Hwang J, Kim JY, Hyun M, et al. A risk scoring system to predict progression to severe pneumonia in patients with Covid-19. Sci Rep 2022;12:5390.
64. Davis HE, Assaf GS, McCorkell L, Wei H, Low RJ, Re’em Y, et al. Characterizing long COVID in an international cohort: 7 months of symptoms and their impact. EClinicalMedicine 2021;38:101019.
65. Bull-Otterson L, Baca S, Saydah S, Boehmer TK, Adjei S, Gray S, et al. Post–COVID conditions among adult COVID-19 survivors aged 18–64 and ≥65 years — United States, March 2020–November 2021. MMWR Morb Mortal Wkly Rep 2022;71:713–7.
66. Sudre CH, Murray B, Varsavsky T, Graham MS, Penfold RS, Bowyer RC, et al. Attributes and predictors of long COVID. Nat Med 2021;27:626–31.
67. Su Y, Yuan D, Chen DG, Ng RH, Wang K, Choi J, et al.; ISB-Swedish COVID-19 Biobanking Unit. Multiple early factors anticipate post-acute COVID-19 sequelae. Cell 2022;185:881–95.e20.
68. Dos Santos WG. Natural history of COVID-19 and current knowledge on treatment therapeutic options. Biomed Pharmacother 2020;129:110493.
69. Berenguer J, Borobia AM, Ryan P, Rodríguez-Baño J, Bellón JM, Jarrín I, et al.; [email protected] and [email protected] Study Groups. Development and validation of a prediction model for 30-day mortality in hospitalized patients with COVID-19: the COVID-19 SEIMC score. Thorax 2021;76:920–9.
70. Ting DSW, Carin L, Dzau V, Wong TY. Digital technology and COVID-19. Nat Med 2020;26:459–61.
71. Garrett PM, Wang YW, White JP, Kashima Y, Dennis S, Yang CT. High acceptance of COVID-19 tracing technologies in Taiwan: a nationally representative survey analysis. Int J Environ Res Public Health 2022;19:3323.
72. Nguyen PQ, Soenksen LR, Donghia NM, Angenent-Mari NM, de Puig H, Huang A, et al. Wearable materials with embedded synthetic biology sensors for biomolecule detection. Nat Biotechnol 2021;39:1366–74.
73. Shen YT, Chen L, Yue WW, Xu HX. Digital technology-based telemedicine for the COVID-19 pandemic. Front Med 2021;8:646506.
74. Mishra T, Wang M, Metwally AA, Bogu GK, Brooks AW, Bahmani A, et al. Pre-symptomatic detection of COVID-19 from smartwatch data. Nat Biomed Eng 2020;4:1208–20.
75. Quer G, Radin JM, Gadaleta M, Baca-Motes K, Ariniello L, Ramos E, et al. Wearable sensor data and self-reported symptoms for COVID-19 detection. Nat Med 2021;27:73–7.
76. Natarajan A, Su HW, Heneghan C. Assessment of physiological signs associated with COVID-19 measured using wearable devices. NPJ Digit Med 2020;3:156.
77. Smarr BL, Aschbacher K, Fisher SM, Chowdhary A, Dilchert S, Puldon K, et al. Feasibility of continuous fever monitoring using wearable devices. Sci Rep 2020;10:21640.
78. Miller DJ, Capodilupo JV, Lastella M, Sargent C, Roach GD, Lee VH, et al. Analyzing changes in respiratory rate to predict the risk of COVID-19 infection. PLoS One 2020;15:e0243683.

Artificial intelligence; Big data; COVID-19; precision medicine; SARS-CoV-2; Smart healthcare

Copyright © 2022, the Chinese Medical Association.