Journal of Medical Internet Research

Effectiveness of Telemedicine vs Face-to-Face Consultation in Fighting COVID-19: Retrospective Cohort Study of Adult Patients With COVID-19 in a Primary Care Setting

2026-05-14T17:00:04-04:00

Background: Telemedicine use expanded rapidly during the COVID-19 pandemic. The Hong Kong Hospital Authority (HA) launched both tele-designated clinics (Tele-DCs) and face-to-face physical designated clinics (PDCs) to manage mild cases. However, the comparative effectiveness of these models remains unclear. Objective: This study aimed to compare clinical outcomes, specifically hospitalization and severe complications, between patients with mild COVID-19 managed via Tele-DCs versus PDCs in Hong Kong’s public primary care setting. Methods: We conducted a retrospective cohort study involving all patients with COVID-19, aged 18 years or older, who visited a PDC (n=23,031) or a Tele-DC (n=38,628) at the Kowloon Central Cluster in Hong Kong from July 28, 2022, to January 29, 2023. Patients were matched 1:1 using propensity score matching based on age, sex, smoking status, comprehensive social security assistance (CSSA) status, and the Charlson comorbidity score, resulting in 17,199 patients per group. The primary outcome was the hospital admission rate between day 1 and day 28. Secondary outcomes included severe complications, mortality, accident and emergency department (AED) use, the antiviral prescription rate, and DC revisit. Results: The average age of patients in the Tele-DC and PDC groups was 58.55 (SD 17.53) and 58.53 (SD 17.54) years, respectively (P=.93). In both groups, 9.05% (n=1557) of patients were on CSSA, and 11% (n=1892) were smokers. Compared to the PDC group, the Tele-DC group demonstrated similar hospital admission rates (Tele-DC: n=497, 2.89%; PDC: n=471, 2.74%; between-group difference 0.15%, 95% CI –0.20% to 0.50%, P=.40), lengths of stay (Tele-DC: mean 6.92, SD 0.47 days; PDC: mean 6.61, SD 0.50 days; between-group difference 0.31 days, 95% CI –1.65 to 1.04, P=.66), severe complication rates (Tele-DC: n=46, 0.27%; PDC: n=33, 0.19%; between-group difference 0.08%, 95% CI –0.03% to 0.18%, P=.18), and mortality rates (Tele-DC: n=23, 0.13%; PDC: n=18, 0.10%; between-group difference 0.03%, 95% CI –0.04% to 0.10%, P=.39). However, the Tele-DC group exhibited a higher AED visit rate (n=641, 3.73%, vs n=542, 3.15%; between-group difference 0.58%, 95% CI 0.19%-0.96%, P.003) and DC revisit rate (n=1446, 8.41%, vs n=1287, 7.48%; between-group difference 0.93%, 95% CI 0.09%-1.50%, P.002). In addition, the Tele-DC group had a lower antiviral prescription rate (n=9872, 57.4%, vs n=10,797, 62.78%; between-group difference –5.38%, 95% CI –6.41% to –4.32%, P<.001). Conclusions: The tele-DC demonstrated clinical safety comparable to the PDC regarding hospitalization and severe complications for patients with mild COVID-19. By validating a scalable model without complex home monitoring, these findings challenge the strict necessity of physical examinations for safe triage and support a digital-first strategy for future infectious surges. However, the disparities observed in AED visits and antiviral prescription rates suggest that integrated remote monitoring tools and improved medication logistics are needed to fully replicate the efficacy of conventional care.

Evaluating Encoder and Decoder Models for Extended Clinical Concept Recognition in Japanese Clinical Texts: Comparative Study With Weighted Soft Matching

2026-05-14T16:30:19-04:00

Background: Extracting medical knowledge for secondary purposes, such as diagnostic support, continues to pose a substantial challenge. Conventional named entity recognition has focused on short terms (eg, genes, diseases, and chemicals), whereas extraction and assessment of longer, complex expressions remain underexplored. Clinically vital concepts, such as diseases, pathologies, symptoms, and findings, often appear as long phrases, and accurate extraction is crucial for applications such as constructing causal knowledge from case reports. Consequently, a framework addressing both short terms and clinically meaningful long phrases—termed extended Clinical Concept Recognition (E-CCR)—is essential. Objective: This study, the first comprehensive investigation of E-CCR model selection, aimed to identify optimal strategies by comparing encoder versus decoder models and general-purpose versus domain-specific pretraining. We analyzed variation in effectiveness by target length and proposed a novel E-CCR evaluation metric. Methods: We evaluated 17 encoder and decoder models using J-CaseMap, a database of approximately 20,000 Japanese case reports annotated with clinical concepts. Performance was primarily assessed using the weighted soft matching score, which penalizes fragmentation of long extraction targets and weights scores by target length to account for the greater difficulty of extracting longer expressions. Results: On J-CaseMap, JMedDeBERTa(s)—an encoder model pretrained on domain-specific medical text—achieved the highest mean performance (F1-score=0.758, SD 0.002), with similarly strong results from JMedDeBERTa(c), suggesting comparable performance among the top encoder models. As the fragmentation penalty increased, performance generally declined; however, no consistently severe degradation was observed. On the Medical Report Named Entity Recognition for positive disease dataset, the general-domain DeBERTaV2-base yielded the highest mean F1 score, and differences among the medical-domain JMedDeBERTa(s) and JMedDeBERTa(c) variants were small, suggesting limited benefit of domain-specific pretraining. Overall, under our experimental settings (low-rank adaptation fine-tuning for decoders and full fine-tuning for encoders), encoder models outperformed decoder models, and token classification outperformed our instruction tuning setup. Conclusions: Under our experimental setting, encoder-based token classification achieved the highest mean performance on our internal dataset. Differences among the top encoder models were small and should be interpreted as comparable within the uncertainty implied by our annotation review, whereas decoder-based approaches did not surpass encoder-based models in this setup, suggesting that encoder models can deliver high accuracy with fewer parameters and may offer practical advantages in resource-constrained environments. Token classification outperformed instruction tuning for extracting long expressions, whereas instruction tuning was better suited to short terms. Using the weighted soft matching score, we found that performance did not substantially deteriorate as the fragmentation penalty increased, indicating that extracted spans were rarely fragmented. Similar trends in external validation datasets suggest that findings under our setup may generalize to information extraction tasks on Japanese medical text. Further investigation is needed to determine whether these findings hold across other languages and medical document types.

Machine Learning and Deep Learning Models for Predicting Future Falls in Community-Dwelling Older Adults: Systematic Review and Meta-Analysis of Longitudinal Evidence

2026-05-14T16:00:22-04:00

Background: Machine learning (ML) and deep learning (DL) show promise for fall risk prediction, but prior reviews focused mainly on real-time fall detection, in-hospital falls, or conventional statistical models. The performance of ML-DL–based models for predicting future falls in community-dwelling older adults remains unclear. Objective: This study aimed to review ML-DL studies for predicting future falls among community-dwelling older adults and meta-analyze discrimination where feasible. Methods: Six databases were searched from inception to September 23, 2024, with updates on August 31, 2025, and February 28, 2026. We included longitudinal studies developing or validating ML-DL models to predict future falls in community-dwelling adults aged ≥60 years and excluded real-time detection, simulated or no fall, and inpatient studies. Risk of bias was assessed using the Prediction Model Risk of Bias Assessment Tool (PROBAST). Areas under the curve (AUCs) were meta-analyzed using Hartung-Knapp-Sidik-Jonkman random-effects models with 95% CIs. Heterogeneity, 95% prediction intervals (PIs), sensitivity analyses, and subgroup analyses were conducted. Results: After screening 10,253 records, 28 (0.3%) studies were included; 18 (64.3%) focused on general older adults. Prediction horizons ranged from 3 months to 7 years, and fall incidence ranged from 1.6% to 46.6%. Twenty-three (82.1%) studies applied ML, and 5 (17.9%) studies used DL. Input modalities included text (n=18, 64.3%), sensor (n=5, 17.9%), image (n=1, 3.6%), and multimodal data (n=4, 14.3%). Common predictors included age, sex, fall history, depression, and basic daily activities. Only one model underwent external validation. Calibration reporting was sparse. All models were rated at high risk of bias. Ten models were meta-analyzed, yielding a pooled AUC of 0.79 (95% CI 0.69‐0.87) with extreme heterogeneity (=0.64; =0.80; =99.8%; =4128.99). The confidence-distribution bootstrap PI was 0.20 to 0.99, indicating substantial uncertainty in expected performance across new populations. Subgroup analyses indicated moderation by sample size and population type, with higher discrimination in specific populations than in general samples; however, the specific population subgroup included only 2 studies. Although all participants were community dwelling, some cohorts were recruited through clinically enriched pathways rather than general community sampling. Conclusions: ML-DL models show potential for identifying community-dwelling older adults at elevated future fall risk; however, wide PIs, limited external validation, and high risk of bias suggest real-world performance may be optimistic. The pooled AUC should be interpreted as a summary of reported discrimination under study-specific conditions, predominantly from internally validated, high-risk-of-bias models, rather than as a robust estimate of transportable real-world performance. This review extends prior reviews by focusing on community-dwelling settings and by integrating PROBAST, Hartung-Knapp-Sidik-Jonkman meta-analysis, PIs, and modality-specific synthesis to evaluate both discrimination and uncertainty. Findings support the use of ML-DL models for proactive fall prevention while emphasizing the need for validation and context-specific implementation. Trial Registration: PROSPERO CRD42024580902; https://www.crd.york.ac.uk/PROSPERO/view/CRD42024580902

A Digital Diabetes Self-Management Education and Support Program Integrated With Continuous Glucose Monitoring for Type 2 Diabetes: Randomized Controlled Trial

2026-05-14T15:45:14-04:00

Background: Previous research has demonstrated that the use of continuous glucose monitoring (CGM) can improve glycemic control in people with type 2 diabetes when used regularly alongside a digital diabetes self-management education and support (DSMES) program. However, to date, there is limited evidence showing the benefits of a digitally delivered DSMES program combined with real-time CGM for adults with type 2 diabetes. Objective: The objective of this study is to evaluate the impact of a DSMES program coupled with CGM on hemoglobin A (HbA) and CGM-derived glycemic measures compared to usual care for adults with type 2 diabetes over 6 months. Methods: Participants with type 2 diabetes and HbA of 8% or higher (64 mmol/mol) who were not using mealtime bolus insulin (aged 26‐83 y; mean HbA 9.6%, SD 1.4% [mean 81.2 mmol/mol, SD 15.8 mmol/mol]) were randomly assigned to a digital DSMES+CGM integrated solution (n=51) or usual care (n=49) for 6 months. The primary outcome was HbA. The secondary outcomes were CGM-derived glycemic measures, including glucose management indicator, percent time in range 70 to 180 mg/dL, percent time above range (>180 mg/dL), percent time below range (<70 mg/dL), and mean glucose. Linear mixed effects models were used for intention-to-treat analyses. Results: HbA was lower among the intervention group versus the usual care group at 3 months (difference=−0.7%, 95% CI −1.4% to −0.1% or difference=−8.1 mmol/mol, 95% CI −15.5 to −0.7 mmol/mol; =.03) and at 6 months (difference=−0.6%, 95% CI −1.4% to 0.2% or difference=−6.9 mmol/mol, 95% CI −15.7 to 1.9 mmol/mol; =.12) but only reached statistical significance at 3 months. CGM-derived glycemic measures, including glucose management indicator (difference=−0.9%, 95% CI −1.7% to −0.1%; =.03), time in range (difference=14.6%, 95% CI 1.0% to 28.2%; =.04), time above range (difference=−14.9%, 95% CI −29.0% to −0.9%; =.04), and mean glucose (difference=−36.4 mg/dL, 95% CI −70.0 to −2.9 mg/dL; =.03), also significantly improved for the intervention group versus the usual care group at 6 months. Conclusions: The combination of digital DSMES+CGM is effective for supporting adults with type 2 diabetes in managing their condition and has the potential to reduce the risk of long-term health complications. Trial Registration: ClinicalTrials.gov NCT05368454; https://clinicaltrials.gov/ct2/show/NCT05368454

Mode Effects Between Mobile Web and Telephone Surveys on Patient Experience Scores in South Korea: Secondary Analysis of a Randomized Controlled Trial Under Various Missingness Scenarios

2026-05-14T15:45:14-04:00

Background: Patient experience surveys are essential tools for assessing health care quality, yet the potential influence of survey mode on patient experience scores remains understudied. This study investigates the mode effects between mobile web and telephone surveys within South Korea’s Patient Experience Assessment. Objective: This study aimed to examine the presence and extent of the mode effects of mobile web versus telephone surveys on patient experience scores. The primary outcome was defined as the total score across all 21 survey items, rescaled to 0‐100. Methods: This is a secondary analysis using experimental data from a parallel-group randomized controlled trial involving 3200 patients (adults aged ≥19 years, hospitalized >1 day, discharged 2‐56 days before the survey) from 4 general hospitals between October and November 2022, equally allocated to telephone and mobile web survey modes. An independent survey company generated the random allocation sequence using computer-generated random numbers and assigned participants to the survey modes. Due to the nature of the intervention, blinding of participants, interviewers, and outcome assessors was not feasible after assignment. We calculated unadjusted score differences among respondents and estimated adjusted differences accounting for nonresponse using inverse probability weighting (IPW) and multiple imputation (MI) under the missing-at-random assumption. Sensitivity analyses, using the delta-adjustment method based on the missing-not-at-random assumption, assessed robustness to departures from the missing-at-random assumption. Subgroup analyses by sex, age group, and field of care were also conducted. Results: Of 3200 patients randomized (1600 per mode), 878 completed the survey (520 mobile web and 358 telephone). Analyses included all randomized participants (n=3200), with nonresponse addressed through IPW and MI. No adverse events were reported in this survey-based study. The total patient experience score was significantly lower in the mobile web group (mean 81.5, SD 16.4) than in the telephone group (mean 84.9, SD 14.3; unadjusted difference –3.41 points, 95% CI –5.51 to –1.31; IPW-adjusted –4.11, 95% CI –6.17 to –2.04; MI-adjusted –4.59, 95% CI –7.45 to –1.73). Similar patterns were observed across most subdomains. Subgroup analyses revealed consistent mode effects across different demographic categories. Sensitivity analyses using the delta-adjustment method confirmed the robustness of these findings under various missing data scenarios. Conclusions: Mobile surveys may yield substantially lower patient experience scores than telephone surveys. Unlike previous studies, our study analyzes randomized experimental data under various missingness scenarios and provides evidence that this mode effect is unlikely to be attributable to analytical methods or heterogeneity in respondent characteristics between the 2 survey administration modes. Accordingly, caution is warranted when comparing patient experience scores obtained from traditional telephone surveys with those from mobile surveys. Methodologically, our sensitivity analysis approach provides a robust framework for assessing and addressing potential nonresponse bias in patient experience assessments. Trial Registration: Clinical Research Information Service KCT0011374; https://tinyurl.com/3e3u5mjs

Understanding How a Digital Platform for Chronic Disease Management Can Enable and Limit Patient Self-Care: Qualitative Study

2026-05-14T15:15:13-04:00

Background: A growing segment of the population requires ongoing care and support for managing their chronic diseases. Digital platforms for self-management are rapidly emerging to meet this need, but patients’ experiences with these platforms vary significantly. This may be due to the complexity and flexibility of digital platforms, where the wide array of available features can generate unexpected impacts. Objective: This study aims to explore how a digital platform can both enable and limit patients with a chronic disease in managing their own health. Methods: We conducted semistructured qualitative interviews with patients to better understand their experience of using a digital platform for self-managing their chronic diseases. Patients who had been using a digital platform (the Chronic Care Platform) for at least 1 month were invited to participate. Twenty-four patients were recruited and interviewed in person or by phone. The collected data were analyzed using template analysis, which is a type of thematic analysis that allows inductive identification of themes from data and deductive application of theory-informed themes. We leveraged Self-Care Theory to understand how patients’ motivation to use the platform and their subsequent use of its features generated perceived value and challenges in achieving this value. Results: The platform was shown to support patients’ development of core self-care abilities (cognitive, psychosocial, and sociocultural abilities) and self-care behaviors (maintenance, monitoring, management), but it did not provide any support to the development of physiological abilities. Moreover, results indicate important limitations in the way in which the digital platform supported all self-care abilities and behaviors—in particular, self-care management. Hence, while the platform was viewed as valuable overall, patients reported several challenges in effectively using the Chronic Care Platform for self-care. Conclusions: Digital platforms for chronic disease management can enhance patient self-care by providing valuable resources and support for reinforcing desired behaviors. However, gaps in platform features can limit patients’ ability to comprehensively care for themselves. This study shows that relating platform features to specific dimensions of self-care can help identify missing features, providing a fine-grained understanding of how a given platform is generating positive impacts and how it may be improved to fully support self-care.

Understanding Patient-Reported Offenses in Electronic Health Records: Cross-Sectional Mixed Methods Survey

2026-05-14T14:30:13-04:00

Background: Patients’ access to their electronic health record (EHR) supports their participation and satisfaction with care. Despite the benefits, some patients have been upset after reading their EHR. Additionally, health care professionals are concerned that patients, particularly those with mental health conditions, may be offended, and they have expressed a need for further guidelines on how to write EHRs. Experiences among various patient groups are essential to support the relationship between patients and professionals. However, prior studies have often focused on single patient groups or specific clinical contexts, leaving a limited understanding of differences across multiple patient groups. Objective: This study aimed to determine whether certain patient groups are more likely to feel offended while reading their EHRs and which information is perceived as offensive and to provide a comparison across multiple patient groups using a mixed methods approach. Methods: A cross-sectional survey was conducted via the Finnish national patient portal using a web-based patient survey, adopting a mixed methods approach. The survey included multiple-choice and open-ended questions. The total sample comprised 4681 respondents. The survey respondents were placed into 4 patient groups: those who had received care for mental health, cancer, or other conditions and those who had received no care. Associations between the type of care and patients who felt offended were estimated using multivariate binary logistic regression. Inductive content analysis (n=502) was conducted to identify information perceived as offensive in the EHR. Results: The patients who had received mental health care (166/654, 25.4%) or cancer and mental health care (9/39, 23.1%) were more likely to be offended by information in their EHR compared to the other groups (cancer care: 37/375, 9.9%; other conditions care: 383/3316, 11.6%; no care: 22/206, 10.7%; other conditions care: odds ratio 0.37, 95% CI 0.29‐0.46; <.001; model A). Additionally, female patients, those with bad or very bad health conditions, and patients with bachelor’s or master’s degrees were significantly more likely to feel offended. Errors, the health care professionals’ disrespectful language, and perceived unnecessary information were the most frequently mentioned reasons for being offended. Patients with mental health care reported more often that unnecessary information and professionals’ opinions and word choices were experienced as offensive compared to other patients. Conclusions: This study contributes new knowledge by identifying differences across multiple patient groups. Although a minority of patients felt offended by their EHR, health care professionals should consider that some patients, particularly those who have received mental health care or cancer and mental health care, may be offended by specific information or word choices in their EHRs. To address this, health care professionals should receive education on how to write their notes in a neutral tone and avoid potentially offensive topics. Improving the quality of EHRs could strengthen the relationship between patients and professionals.

Digital Health Literacy, Technology Acceptance, and Competence Among Older Adults Aged ≥65 Years: Cross-Sectional Study Investigating Differences Between Women and Men

2026-05-14T13:30:14-04:00

Background: Digital health literacy (DHL) has the potential to improve health among older adults by enhancing access to health-related information and health care services. Objective: The aim of this study was to analyze the relationship between DHL and technology commitment in adults aged 65 years and older, while also investigating possible gender differences. Methods: The analytical sample consisted of 1824 individuals. The analysis included descriptive comparisons in terms of DHL, technology acceptance, competency, support, and internet use. Multivariate regression models (generalized linear models) were applied in order to test the association between DHL and technology commitment, controlling for internet use as well as health-related and sociodemographic characteristics. Results: Male and female participants did not differ in terms of DHL (mean score: 3.5, SD 1.2 [men] and 3.5, SD 1.3 [women]; =.70); however, male participants reported significantly higher technology acceptance (<.001) and higher technology competencies (<.001), but less support with regard to technology use (<.001). Within regression models, only higher technology acceptance (coefficient=0.023, 95% CI 0.006‐0.041; =.01) and support (coefficient=0.027, 95% CI 0.014‐0.040; <.001) were significantly linked to greater DHL. The subgroup analysis revealed that DHL was significantly associated with technology acceptance among men (coefficient=0.036, 95% CI 0.012‐0.060; =.003) but not women (coefficient=0.024, 95% CI 0.008‐0.040; =.44). Conclusions: According to the current results, DHL is highly related to technology commitment. Gender differences should be taken into account when developing and evaluating appropriate interventions to improve DHL by addressing the acceptance of technologies and optimizing support infrastructures.

Maturity, Safety, and Equity of AI-Enabled Systems and Triage in Integrated Primary Care

2026-05-14T13:15:09-04:00

Artificial intelligence (AI)–enabled systems must simultaneously improve the Quintuple Aim and digital health maturity, including equitable access to and quality and interoperability of data, tools, agents, and services. This requires a comprehensive sociotechnical and global approach to cocreation, management, and governance for individuals and organizations in the ecosystem.

EPOCA Tele-Monitoring System for Older Adults at High Risk of Hospitalization: Budget Impact and Cost-Effectiveness Analysis

2026-05-14T13:15:09-04:00

Background: France’s aging population faces high rates of chronic illness, multimorbidity, and avoidable hospitalizations, placing pressure on an already strained health care system. Remote monitoring systems have shown promise in improving care coordination and reducing acute care use. Objective: The objective of this study was to assess the cost-effectiveness of the EPOCA remote monitoring system, implemented within the Vigie-Age framework, compared to the standard of care for older adults with multiple chronic conditions. Methods: Using data from the Vigie-Age Article 51 pilot study (with 722 participants, including 408 participants with long-term follow-up), a cost-utility analysis was conducted over a 10-year lifetime horizon. A Markov model with daily cycles simulated transitions across health states: at home, emergency department visits, hospitalization, and death. Analyses were conducted from both the French National Health Insurance (NHI) and collective perspectives. Direct medical costs, including hospital, outpatient, and intervention costs, were included. Health outcomes were measured in quality-adjusted life years (QALYs). Deterministic and probabilistic sensitivity analyses assessed model robustness. Results: EPOCA was associated with a reduction in emergency department visits by 54% and in hospitalizations by 46%, cutting the average hospital stay from 55.6 (SD 51.7) to 30.6 (SD 27.8) days. Total costs per patient were €29,165 (EUR €1=US $1.13) with EPOCA and €39,929 for standard of care, representing a €10,764 saving from the collective perspective and a €7,430 saving from the NHI perspective. EPOCA yielded 0.04 additional QALYs and remained cost-saving even at higher program costs. Sensitivity analyses confirmed the robustness of the results. EPOCA had a 90% probability of being dominant and a 95% probability of being cost-effective at a €30,000 per QALY threshold. Conclusions: On the basis of currently available evidence, EPOCA may be a cost-effective strategy for older patients at high risk of hospitalization. It could reduce health care use while improving outcomes, supporting its integration into national older adult care pathways and reimbursement by the French NHI.