Journal of Medical Internet Research

Why People Conceal Mental Health Problems: Qualitative Analysis of TikTok Posts

2026-05-19T15:15:17-04:00

Background: Concealment of psychiatric symptoms is a barrier to effective mental health treatment, particularly among patients with suicidal thoughts and behaviors. Prior research on concealment has relied on retrospective self-report or laboratory-based interviews, which may not capture real-world decision-making about disclosure. Social media platforms such as TikTok provide a context in which individuals publicly narrate their experiences about concealing psychiatric symptoms, offering insight into motivations for concealment uninfluenced by experimenter demand characteristics. Objective: To understand patient decision-making about when to conceal and when to disclose psychiatric symptoms, this study examined social media content about patient experiences of concealing mental health symptoms. TikTok was chosen because it is the fastest-growing social media platform, and social media platforms provide an open-ended format for people to express their thoughts and feelings on various topics. Methods: Using a newly created TikTok account to minimize algorithmic bias, we identified and downloaded the 25 most-viewed English-language videos from 4 search terms about concealment in clinical contexts (“lying to therapist,” “lying to my therapist,” “lying to doctor about mental health,” and “lying to doctors about mental health”). After exclusions, 98 videos were included in the analysis. Videos were analyzed using reflexive thematic analysis. Four coders collaboratively developed a codebook through iterative review, triangulation, and consensus discussions. Engagement metrics (views, likes, comments, shares, saves) were recorded and summarized. Results: The 98 videos had 73,252,531 views, 14,356,874 likes, 74,954 comments, 770,027 shares, and 1,204,006 saves. Four themes were constructed among the 90 videos that explicitly discussed motivations for concealment: (1) disclosure perceived as punitive (31/90, 34.4% of videos), including desire to avoid hospitalization (17/90, 18.8%); (2) managing others’ feelings and impressions (28/90, 31.1%), including fear of upsetting therapists (5/90, 5.5%) and maintaining a façade of wellness (7/90, 7.7%); (3) negative emotions or inability to identify feelings (21/90, 23.3%), including fear of vulnerability (6/90, 6.6%); and (4) negative opinions of psychiatric treatment (17/90, 18.8%), including concerns about confidentiality (3/90, 3.3%). An exploratory theme captured ambivalence and guilt surrounding nondisclosure. Conclusions: Results provide insight into patient motivations for concealing their suicidal thoughts and behaviors and offer potential avenues for improving rates of disclosure, which is critical to reducing death by suicide. TikTok creators frequently described concealment as a strategy to avoid perceived punitive consequences, manage interpersonal dynamics, or cope with emotional distress. Findings suggest that current risk management practices and stigma surrounding psychiatric care may unintentionally reinforce concealment behaviors. These insights may inform interventions aimed at improving the therapeutic alliance, enhancing transparency around hospitalization criteria, and reducing barriers to honest reporting of suicide risk.

Efficacy of Digital Speech Therapy for Poststroke Dysarthria: Randomized Noninferiority Trial

2026-05-18T13:45:17-04:00

Background: Poststroke dysarthria, a common speech impairment, affects up to half of all stroke survivors, often reducing their ability to communicate, and adversely affecting their quality of life. Although conventional speech therapy for poststroke dysarthria is effective, access is often limited by time and geographical constraints. Here, digital speech therapy may serve as a remotely deliverable alternative for selected patients. However, few trials have assessed its efficacy, safety, and usability. Objective: This study aimed to evaluate whether a smartphone-based speech therapy app is noninferior to conventional workbook-based therapy in improving speech intelligibility among individuals with poststroke dysarthria. Methods: This single-blind, randomized controlled, noninferiority trial was performed at 3 hospitals in South Korea. Adults (≥19 y) with poststroke dysarthria who were cognitively intact, without aphasia, and able to use a smartphone were eligible. Participants were enrolled between July 20, 2023, and April 15, 2024. Participants were randomly assigned (1:1), stratified by stroke phase, using a block randomization method, to receive either a smartphone-based digital therapy app or a conventional workbook-based therapy for 4 weeks. The primary outcome was speech intelligibility (0‐100 perceptual rating) after the intervention. Primary analysis was intention-to-treat using analysis of covariance. A noninferiority margin of 19 points was pre-defined. Results: A total of 73 participants were enrolled (median age 62.00 years). Among them, 38 were assigned to the digital speech therapy group and 35 to the control group. Intelligibility scores improved from 80.48 (SD 18.92) to 92.08 (SD 12.38) in the intervention group, and from 80.94 (SD 16.74) to 88.11 (SD 18.06) in the control group. The adjusted between-group difference was 4.49 (95% CI 0.61-8.37), and the lower bound of the 95% CI was above the prespecified noninferiority margin (–19), which supported noninferiority. No significant between-group differences were observed in the secondary outcomes related to speech function or psychological status. The system usability score was 89.6, and adherence in the digital speech therapy group was 64.6% based on app logs, with no treatment-related adverse events. Conclusions: Digital speech therapy was noninferior to conventional workbook-based therapy in improving speech intelligibility and was feasible across acute to early subacute and chronic stroke phases in cognitively intact stroke survivors with predominantly mild-to-moderate dysarthria. However, feasibility and efficacy in older stroke survivors with cognitive deficits or co-occurring aphasia, or in those unable to use smartphones, remain to be established. Trial Registration: : ClinicalTrials.gov NCT05865106; https://clinicaltrials.gov/study/NCT05865106 International Registered Report Identifier (IRRID): RR2-10.3389/fneur.2024.1305297

Digital Patient Decision Aid for Antiobesity Medications: Mixed Methods Study of Human-Centered Design and Usability Evaluation

2026-05-15T18:00:05-04:00

Background: The global burden of obesity continues to rise, highlighting the need for patient-centered approaches to weight management. Shared decision-making is particularly important in the selection of antiobesity medications (AOMs), as treatment options differ in mechanism, effectiveness, side effects, routes of administration, and cost. Despite this preference-sensitive context, only a few patient decision aids (PDAs) have been culturally and clinically adapted for use in Asian populations. Objective: This study aims to design, develop, and evaluate a digital PDA, OptiWeight, to support shared decision-making for AOM selection, incorporating perspectives from health care professionals and patients. Methods: This mixed methods, multicenter study, conducted between August 2022 and November 2025, applied a 4-stage human-centered design process. An evidence-informed prototype was developed based on clinical guidelines, followed by 2 rounds of usability testing using think-aloud protocols to assess navigation structures, perceived usability (System Usability Scale [SUS]), and cognitive workload (NASA Task Load Index [NASA-TLX]). Semistructured interviews with health care professionals specializing in weight management, guided by the Consolidated Framework for Implementation Research, informed clinical implementation and workflow integration. Finally, patients with overweight or obesity evaluated usability, cognitive workload, and overall user experience in outpatient settings. Qualitative data were analyzed using content analysis, and 1-way analysis of variance examined changes in usability and workload across stages. Results: A total of 174 individuals were included across all study stages (usability testing among adults: n=78; health care professional interviews: n=18; and clinical evaluation among patients: n=78). Iterative usability testing comparing system- and user-controlled navigation structures revealed complementary strengths and limitations, leading to the adoption of a hybrid navigation structure supporting both sequential guidance and flexible comparison. Additional design requirements included the use of icon arrays to enhance risk comprehension and localization features such as treatment cost displays and clarification of socially impactful side effects. Perceived usability increased from initial testing to clinical evaluation (SUS: 60.53-73.65, P<.001), meeting good usability thresholds, while cognitive workload decreased (NASA-TLX: 40.35-16.69, P<.001). Conclusions: Through a systematic human-centered design process integrating health care professional and patient perspectives, OptiWeight addresses the lack of culturally adapted PDAs for AOM decision-making in Mandarin-speaking populations while capturing user needs—particularly regarding navigation flexibility and risk visualization. The final tool demonstrated good usability and feasibility, and workflow considerations suggest potential for integration into routine weight-management care. Further research is needed to evaluate its impact on decision quality and real-world implementation outcomes.

Assessing the Use of Wearable Mobile-Monitoring Devices Among Individuals With Serious Mental Illness: Qualitative Acceptability and Feasibility Study

2026-05-15T17:30:17-04:00

Background: Serious mental illness (SMI) is difficult to treat for various reasons, such as rapid changes in symptoms, comorbid health conditions, long gaps between provider visits, and additional societal barriers experienced by this population. Wearable mobile-sensing devices can be used to passively collect valuable patient-generated health data, such as daily step count, heart rate variability, sleep information, and other health-related behaviors, which could inform and improve treatment for individuals with SMI. Wearable health devices have become more economically accessible, providing promise for the possibility of their implementation in health care. However, more information regarding how individuals with SMI perceive and interact with these devices is needed. Objective: This study aimed to assess the acceptability and feasibility of using wearable mobile-sensing devices to improve treatment outcomes for Veterans with SMI. In addition, we were also interested in learning if privacy concerns would influence acceptability of devices, specifically surrounding location tracking and health information sharing, as well as assessing other barriers to device use. Methods: Qualitative interviews were conducted with participants who had been using a wearable health and fitness tracker for at least 2 weeks to explore their thoughts and perceptions of these devices. A total of 15 Veterans diagnosed with a SMI participated in interviews. Both thematic analysis and rapid qualitative analysis approaches were used to uncover findings in key domains and emergent themes. Results: Wearable fitness trackers allowed participants to conveniently monitor various aspects of their physical and mental health, provided a greater understanding of their overall well-being, and motivated them to reach personal health goals. Individuals were open to sharing their personal health information collected from the devices with providers to improve their health care treatment and expressed no privacy concerns surrounding data tracking or the device’s global positioning system that monitors physical location. Participants experienced some technological challenges with using the fitness trackers, as well as the device’s accompanying cell phone app. Furthermore, participants expressed difficulties in understanding and interpreting the health data that was collected from the health and fitness trackers. Greater ongoing technological support, in addition to physical device adjustments to enhance comfort and usability, were suggested ways of improving overall user experience. Conclusions: Participants with SMI in this sample were accepting of wearable mobile-monitoring devices and believe it is feasible to incorporate these fitness trackers into their daily lives. Furthermore, participants in this sample expressed no privacy concerns regarding location tracking or the sharing of health information collected from these devices with providers. Patient-generated health data collected from these devices may offer valuable information that could be used to inform health care treatment for this population.

Concerns of Using Large Language Models in Health Care Research and Practice: Umbrella Review

2026-05-15T16:15:15-04:00

Background: Large language models (LLMs), such as ChatGPT (OpenAI), are rapidly evolving, and their applications in health care are increasing. There is a growing demand for automation of routine tasks and a drive to use LLMs or similar to support research. Objective: This umbrella review examines concerns of health care professionals and researchers related to the use of LLMs in health care research and practice. We aimed to identify common issues raised and the implications for patient care, policy, and practice. Methods: A protocol was registered on PROSPERO (CRD420250640997). Searches were conducted in 7 databases (Ovid MEDLINE, Ovid Embase, Scopus, Web of Science, JBI Database of Systematic Reviews and Implementation Reports, Cochrane Database of Systematic Reviews, and Epistemonikos) in February 2025 and updated in February 2026. Screening was conducted in 2 stages, with independent screening by 2 reviewers. Studies published in the English language after January 2017 with at least one outcome expressing concerns of LLM or generative artificial intelligence use in health care research were included. The included studies were quality appraised for risk of bias and certainty of the evidence using AMSTAR-2 (A Measurement Tool to Assess Systematic Reviews) and GRADE (Grading of Recommendations Assessment, Development, and Evaluation), respectively. Data was extracted using a piloted form and narratively synthesized following SWiM guidelines and the PRIOR (Preferred Reporting Items for Overviews of Reviews) checklist. Results: The search retrieved 448 systematic reviews, of which 42 met the inclusion criteria. Further, 12 distinct populations were identified, including researchers and clinicians in various medical specialties. The included reviews were assessed to be of very poor quality, and the level of overlap between primary studies could not be determined. Additionally, 15 reviews focused on ChatGPT, a further 15 on two or more LLMs, and 12 on generic artificial intelligence. Thus, 3 main themes emerged from the narrative synthesis. In order of most to least frequently discussed: (1) technical capability; (2) ethical, legal, and societal; and (3) costs. Conclusions: To our knowledge, this is the first umbrella review to address the concerns of LLMs in health care research and practice. Thematic analyses provided insight into the complexity of different perspectives, and by using a whole population approach, it demonstrates common narratives. However, the poor quality of the included studies and potential overlap of results are substantial limitations. Data quality is at the heart of these concerns, and combative action must ensure health care professionals and researchers have the resources required to overcome these apprehensions. Ethical, legal, and societal implications of artificial intelligence use were also commonly raised. As technology accelerates and demands on health care increase, we must adapt and embrace change with equity, diversity, inclusion, and safety at the core. Trial Registration: PROSPERO CRD420250640997; https://www.crd.york.ac.uk/PROSPERO/view/CRD420250640997

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