International Journal Of Medical Informatics[JOURNAL]
Sun
200 papers
Web-based family-reported outcomes to assess family functioning in multiple sclerosis: an Italian multicenter digital survey.
Lavorgna L, Miele G, Ponzano M … +23 more , Maida E, Abbadessa G, Bile F, Marfia GA, Landi D, Proietti F, Inglese M, Laroni A, Poirè I, Romano G, Signoriello E, Lus G, Ruocco E, Lauro F, Rosa L, Lanzillo R, Perutelli V, Di Tella ML, Streito LM, Sormani MP, Bonavita S, Castelli L, Clerico M
Int J Med Inform · 2026 Sep · PMID 42269253 · Publisher ↗
BACKGROUND AND OBJECTIVES: Multiple sclerosis (MS) affects not only individuals but also family dynamics. Scalable, digitally delivered assessments may support routine screening of psychosocial and relational needs acros...
BACKGROUND AND OBJECTIVES: Multiple sclerosis (MS) affects not only individuals but also family dynamics. Scalable, digitally delivered assessments may support routine screening of psychosocial and relational needs across the household. This study aimed to investigate perceived family functioning in families living with MS, using a web-based, multi-informant approach, focusing on relationship quality and psychosocial factors. METHODS: This Italian multicenter, cross-sectional study recruited families with a member diagnosed with MS. Participants included individuals with MS (PwMS), partners, adolescent children, and other adult relatives. Families without self-reported chronic illness were included as a comparison group. Participants completed anonymous, validated online questionnaires. Standardized instruments included: Hospital Anxiety and Depression Scale (HADS), Family Assessment Measure-III (FAM-III), Multidimensional Scale of Perceived Social Support (MSPSS), Toronto Alexithymia Scale (TAS-20), Dyadic Adjustment Scale (DAS), and Inventory of Parent and Peer Attachment (IPPA). RESULTS: The final sample consisted of 50 families: 50 PwMS (mean age: 42.6 ± 11.4 years), 34 partners (mean age: 45.2 ± 10.6 years), 25 adolescent children (mean age: 16.2 ± 3.6 years), 25 adult relatives (mean age: 47.2 ± 17.7 years). Compared with controls, PwMS reported lower anxiety (median HADS-A: 6.5 vs. 9; p = 0.002), higher dyadic adjustment (median DAS: 109.5 vs. 104.5; p = 0.024) and greater perceived support from significant others (median MSPSS: 27 vs. 24; p = 0.047). Adolescent children showed stronger attachment to parents (IPPA total mother: 94 vs. 83; father: 92.5 vs. 76.5; both p < 0.001) and peers (IPPA total peers: 98 vs. 90; p = 0.016). In PwMS, higher disability correlated with more severe depressive symptoms (ρ = 0.55, p < 0.001) and poorer family functioning (ρ = 0.37, p = 0.009). DISCUSSION: A web-based, standardized, multi-informant assessment can capture family functioning and key psychosocial correlates in MS, supporting the design of digitally enabled screening pathways to identify families who may benefit from early psychological or relational interventions.
Artificial intelligence in health systems of the WHO European region: implementation, applications, opportunities, and barriers.
Adib K, Dunning HE, Letchford N … +6 more , Salama N, Tolias Y, Barros J, Azzopardi-Muscat N, Kluge HHP, Novillo-Ortiz D
Int J Med Inform · 2026 Sep · PMID 42269252 · Publisher ↗
BACKGROUND: Artificial intelligence (AI) is rapidly emerging as a transformative force in healthcare, with the potential to improve patient outcomes, enhance health system efficiency, and support data-driven decision-mak...
BACKGROUND: Artificial intelligence (AI) is rapidly emerging as a transformative force in healthcare, with the potential to improve patient outcomes, enhance health system efficiency, and support data-driven decision-making. This study provides an overview of the current landscape, exploring how Member States are prioritizing AI in response to health system needs, engaging key stakeholders, and addressing critical enablers such as workforce upskilling and preparedness. METHODS: This study is based on a cross-sectional survey developed and administered by the WHO Regional Office for Europe, which was launched in June 2024 and remained open until March 2025. All responses were consolidated into a standardized database, reviewed for internal consistency, and analyzed using an exploratory, descriptive approach. Results are presented as percentages and absolute values, disaggregated at the regional, subregional, and EU27 levels. FINDINGS: The survey response rate was 94%, with 50 out of 53 Member States participating. The top opportunities rated by Member States for the use of AI in health were improving patient care and health outcomes (96%; 48/50), reducing pressure on the healthcare workforce (92%; 46/50), and enhancing health system efficiency (90%; 45/50). The most commonly reported applications of AI included AI-assisted diagnostics (64%; 32/50), conversational platforms or chatbots (50%; 25/50). The highest rated barriers rated to widespread adoption were legal uncertainty (48%; 24/50) and financial affordability (46%; 23/50). In terms of collaboration and engagement, 72% of Member States (36/50) reported stakeholder involvement, primarily through focus groups (46%; 23/50) and informal meetings, seminars, or workshops (44%; 22/50). Of the 36 Member States reporting at least one form of engagement, the stakeholders most frequently engaged were government actors (81%; 29/36), healthcare providers (75%; 27/36), and AI developers (75%; 27/36), whereas patient associations (42%; 15/36) and the general public (22%; 8/36) were less commonly involved. Finally, with regard to health workforce education and AI literacy, the study highlights a significant gap in preparedness, where only 24% of Member States (12/50) reported offering in-service AI training for the health workforce, and 20% (10/50) offered pre-service training. INTERPRETATION: The integration of AI in healthcare across the WHO European Region is advancing rapidly, demonstrating potential to enhance patient outcomes and alleviating workforce pressures. However, its widespread and equitable adoption remains limited by challenges related to affordability, regulatory and legal ambiguities, insufficient patient engagement, and the need for strategic investment in workforce training and upskilling.
Mobile health (mHealth) applications for community health workers in low- and middle-income countries: A scoping review.
Charoensilpchai C, Lkhagvajav Z, Turner AM
Int J Med Inform · 2026 Sep · PMID 42263381 · Publisher ↗
BACKGROUND/PURPOSE: Mobile health (mHealth) applications are increasingly used to support community health workers (CHWs) in delivering primary care services in low- and middle-income countries (LMICs); however, evidence...
BACKGROUND/PURPOSE: Mobile health (mHealth) applications are increasingly used to support community health workers (CHWs) in delivering primary care services in low- and middle-income countries (LMICs); however, evidence remains fragmented regarding how technical configuration intersects with the sociotechnical experience of the workforce. This scoping review aimed to characterize the technological and sociotechnical landscape of smartphone- and tablet-based mHealth applications used by CHWs in LMIC primary care settings. METHODS: Following the PRISMA-ScR framework, systematic searches were conducted in PubMed, CINAHL, Cochrane, and Global Index Medicus to identify peer-reviewed and regionally diverse literature published between 2013 and 2025. Eligible studies reported on smartphone- or tablet-based applications designed to support CHWs working in LMIC primary care settings. Data were synthesized using a three-dimensional architecture framework (functions, connectivity, and interoperability) and thematic synthesis across five neutral sociotechnical domains: the Tool, Workflow, Worker, Client, and Health System. RESULTS: Thirty studies met the inclusion criteria. Evidence was geographically concentrated in Sub-Saharan Africa and South Asia, with notable gaps in upper-middle-income economies, including regions of Latin America and East Asia. While 25 of 30 applications utilized offline-first architecture, 29 lacked full interoperability with national health information systems. Thematic synthesis identified inherent tensions: while digital protocols (WHO function A3) enhanced clinical confidence and professional status, they simultaneously introduced dual-entry workload friction and managerial monitoring. Sustainability was consistently hindered by a digital island effect resulting from a lack of comprehensive back-end interoperability (functions D2, D6). CONCLUSIONS: The potential of mHealth applications for CHWs in LMICs is undermined by an architectural-systemic mismatch. Long-term success requires a shift from pilot-centric hardware deployment toward interoperability-first strategies that prioritize technical interoperability and clinical empowerment over top-down monitoring. Governments and partners should mandate architectural alignment with national health infrastructures to transition CHWs from isolated data-entry clerks to connected digital clinicians.
Cross-cultural drivers of satisfaction in digital public health services: evidence from India and the UK using text mining.
Rai P, Bera S, Raghavendra AH
Int J Med Inform · 2026 Sep · PMID 42263380 · Publisher ↗
BACKGROUND: Digital public health services (DPHS) succeed when three elements align: dependable operations, clear public-value benefits, and credible institutional trust. How these elements jointly shape user satisfactio...
BACKGROUND: Digital public health services (DPHS) succeed when three elements align: dependable operations, clear public-value benefits, and credible institutional trust. How these elements jointly shape user satisfaction across cultures remains unclear. OBJECTIVE: Identify which user-voiced themes relate most to satisfaction, whether legitimacy and civic messages change how complaints affect ratings, and how these relationships vary across cultures. METHODS: We analyzed app-store reviews for two DPHS apps (Indian/UK) using structural topic modeling aligned with Public Value Theory dimensions, public-value outcomes (PVO), operational capacity (OC), and the authorizing environment (AE). Ordered logit models estimated main effects of theme salience on ratings; interaction models tested whether AE cues (institutional legitimacy, prosocial messaging) moderate OC and PVO complaints; a cross-country sensitivity analysis compared India and the UK. RESULTS: OC problems, access, verification, updates, stability, are the most consistent drivers of dissatisfaction. PVO features lift satisfaction only when they deliver end-to-end in real conditions (reliable booking, trustworthy exposure logic, seamless results integration); ambiguous signals or broken data flows quickly erode confidence. Legitimacy and civic cues matter, but not as substitutes for reliability: pairing "trust in government" or solidarity appeals with unresolved core faults typically worsens perceptions; the same cues help when problems are peripheral or visibly being fixed. Cultural differences reshape these effects. In more collectivist, higher power-distance settings, stewardship and shared-purpose framing contribute more to satisfaction once basics work. In more individualist, egalitarian settings, users weigh demonstrable privacy safeguards, transparent evidence, and precise remediation over patriotic messaging. CONCLUSIONS: Make reliability non-negotiable, verify public-value features in live conditions before promotion, and tailor legitimacy strategies to culture. Use civic appeals to amplify, not replace, operational excellence; lead with privacy, transparency, and proof where autonomy expectations are higher. This framework turns Public Value Theory into an actionable playbook for DPHS design, communication, and rollout.
Digitally supported, patient-initiated care: Maintaining control of inflammatory bowel diseases while achieving high patient satisfaction.
Esbjørn M, Okdahl T, Fallingborg J … +5 more , Larsen P, Andersen DV, Drewes AM, Jess T, Larsen L
Int J Med Inform · 2026 Sep · PMID 42259122 · Publisher ↗
BACKGROUND: The increasing prevalence of inflammatory bowel disease (IBD) and constrained clinical resources necessitate a shift from fixed-interval outpatient visits towards patient-initiated, needs-based follow-up. Rea...
BACKGROUND: The increasing prevalence of inflammatory bowel disease (IBD) and constrained clinical resources necessitate a shift from fixed-interval outpatient visits towards patient-initiated, needs-based follow-up. Real-world evidence on whether such transitions maintain disease control and achieve patient acceptance remains limited. In 2024, Aalborg University Hospital (Denmark) transitioned all patients with IBD not receiving biological therapy to a digitally supported, patient-initiated monitoring model. We aimed to evaluate the quality of care and patient satisfaction after the transition. METHODS: This pre-post quality improvement study included 966 patients with IBD (Crohn's disease (CD): n = 279; ulcerative colitis (UC): n = 687). Clinical data, disease activity scores, faecal calprotectin levels, and health-related quality of life were obtained from the GASTROBIO web-based regional database for two years before and up to two years after the transition. Patient satisfaction was assessed via a questionnaire adapted from the validated Danish "LUP survey". RESULTS: Entries of patient reported outcomes (PROs) and faecal calprotectin sampling increased following transition. The proportion of patients in faecal calprotectin-defined remission increased in both CD (63 % to 76 %; p = 0.01) and UC (62 % to 70 %; p = 0.04), while disease activity scores remained stable. Health-related quality of life improved in both CD (p = 0.02) and UC (p < 0.01). Of 966 eligible patients, 371 (38 %) completed the satisfaction survey; the majority preferred the new model over standard of care (CD: 77 %; UC: 84 %) and rated satisfaction as high or very high (CD: 66 %; UC: 71 %). Higher satisfaction correlated with better health-related quality of life in both groups (CD: Spearman's ρ = - 0.22, p = 0.05; UC: ρ = - 0.18, p = 0.01). CONCLUSION: A patient-initiated, digitally supported remote monitoring model maintained disease control, improved health-related quality of life, and achieved high patient satisfaction in a large real-world IBD cohort. These findings support the feasibility and scalability of needs-based IBD follow-up underpinned by an integrated digital infrastructure.
Beyond the checklist: An exploration of informatics needs in emergency mental health triage.
Shu R, Hidellaarachchi D, Low J … +1 more , Haryanto A
Int J Med Inform · 2026 Sep · PMID 42259121 · Publisher ↗
BACKGROUND: Emergency mental health triage faces significant challenges due to protocols primarily designed for physical trauma. These protocols tend to focus heavily on physiological metrics, frequently overlooking the...
BACKGROUND: Emergency mental health triage faces significant challenges due to protocols primarily designed for physical trauma. These protocols tend to focus heavily on physiological metrics, frequently overlooking the complex behavioural cues that are essential for effective psychiatric assessment. While Clinical Decision Support Systems (CDSS), which are digital tools designed to assist clinical decision-making through real-time data integration, offer potential solutions to reduce cognitive load, their adoption is often limited by poor usability and a lack of alignment with the practical realities of clinical practice. OBJECTIVE: This study explores the informatics needs and workflow barriers of frontline clinicians to inform the design of user-centred digital mental health triage systems. METHODS: A qualitative exploratory study was conducted using an interpretivist approach. Semi-structured interviews were held with 14 clinicians involved in emergency mental health triage across emergency department (ED) and related crisis care settings, spanning both standardised protocol-driven and resource-constrained distributed triage environments. Data were analysed using reflexive thematic analysis. RESULTS: Three core themes were identified: (1) systemic constraints from fragmented information systems and resource shortages force clinicians to make high-stakes decisions with incomplete data; (2) clinicians employ experiential heuristics and pattern recognition to navigate psychiatric complexity beyond standardised protocols; (3) future digital tools should prioritise deep interoperability, support non-linear workflows, and incorporate culturally sensitive privacy protections. CONCLUSION: This study reveals that triage challenges are deeply rooted in fragmented information infrastructure and workflow misalignment. Effective digital support systems require interoperability, workflow integration, and user-centred design to transition from administrative burden to real cognitive support for ED clinicians.
Association Between Lactate-to-Albumin Ratio and 28-Day All-Cause Mortality in ICU-Admitted Acute Pancreatitis Patients: Development and Validation of a Machine Learning-Based Predictive Model Integrating Albumin-Corrected Indices.
Wei K, Li H, Huang H … +4 more , Lu M, Li X, Huang X, Qin M
Int J Med Inform · 2026 Sep · PMID 42259120 · Publisher ↗
OBJECTIVE: Patients with severe acute pancreatitis (AP) carry a high mortality risk. The lactate-to-albumin ratio (LAR) has shown prognostic value in conditions including sepsis, heart failure and acute respiratory failu...
OBJECTIVE: Patients with severe acute pancreatitis (AP) carry a high mortality risk. The lactate-to-albumin ratio (LAR) has shown prognostic value in conditions including sepsis, heart failure and acute respiratory failure. We aimed to assess LAR's predictive value for 28-day mortality in ICU-admitted AP patients and to establish a machine learning-based predictive model integrating three albumin-derived indices-LAR, red blood cell distribution width-to-albumin ratio (RAR), and blood urea nitrogen-to-albumin ratio (BAR)-for this outcome. METHODS: We analysed a cohort of 380 ICU-admitted AP patients from the MIMIC-IV database (2008-2022) as the training set to evaluate LAR's prediction of 28-day mortality. Predictors were screened using the Boruta algorithm, incorporating LAR, RAR, BAR and clinical variables (e.g., age, creatinine). Seven machine learning models were developed to predict 28-day all-cause mortality. Internal validation employed a 2:1 random data split; external validation used 298 AP patients from the eICU-CRD database (2014-2015). The SHAP method elucidated prediction mechanisms. RESULTS: Multivariate Cox regression analysis revealed that patients in the highest quartile of LAR (>1.72 mmol/g) exhibited a significantly elevated mortality risk (adjusted HR = 8.25, 95% CI 2.57-26.54). The Boruta feature selection algorithm identified 11 independent predictors, including LAR, BAR, RAR, and age. Following 5-fold internal cross-validation, the Extratrees model demonstrated superior performance in the training set benchmark testing (AUC = 0.8546). During internal validation, the Extratrees model maintained optimal predictive capability (AUC = 0.852). Calibration curve analysis and Decision Curve Analysis (DCA) substantiated its high predictive accuracy and clinical net benefit. External validation further corroborated the robustness of Extratrees model (AUC = 0.82). A web-based computational platform (https://bssrmyy.shinyapps.io/DynNom_Model2/) has been deployed to enable online individualized risk assessment. CONCLUSION: LAR is a strong independent predictor of 28-day mortality in ICU-admitted AP patients. This study pioneers the development and validation of a prognostic prediction model for AP based on three albumin-corrected indices (LAR, BAR, RAR) and machine learning. The Extratrees model performed excellently in both internal and external validations, providing an efficient tool for precise risk stratification and early intervention in ICU-managed AP. TRIAL REGISTRATION: Not applicable (retrospective cohort study).
Corrigendum to "Development and validation of a risk prediction model and prediction tools for post-thrombotic syndrome in patients with lower limb deep vein thrombosis" [Int. J. Med. Inform. 187 (2024) 105468].
Sun XX, Ling H, Zhang L … +6 more , Chen RB, Zhong AQ, Feng LQ, Yu R, Chen Y, Liu JQ
Int J Med Inform · 2026 Sep · PMID 42252249 · Publisher ↗
Abstract loading — click title to view on PubMed.
Artificial intelligence applications using patient-generated health data for pre-care processes in elective healthcare: a systematic review.
Warren T, van der Weegen W, Kool RB … +2 more , Hoogendoorn M, Timmers T
Int J Med Inform · 2026 Sep · PMID 42250455 · Publisher ↗
PURPOSE: Artificial intelligence (AI) can leverage patient-generated health data (PGHD) to support pre-care processes such as triage, symptom assessment, and history-taking. Existing systematic reviews have examined AI c...
PURPOSE: Artificial intelligence (AI) can leverage patient-generated health data (PGHD) to support pre-care processes such as triage, symptom assessment, and history-taking. Existing systematic reviews have examined AI clinical decision support, PGHD use, and AI for specific data modalities as separate domains, but none addresses their intersection for pre-care. We aimed to map AI methods and PGHD modalities, synthesize outcomes across technical, clinical, operational, user experience, and equity domains, and identify barriers to deployment and gaps in reporting. METHODS: This systematic review was conducted in accordance with the PRISMA 2020 statement and prospectively registered with PROSPERO (CRD420251134235). We searched PubMed, MEDLINE, and Web of Science (January 2020-June 2025) for studies evaluating AI applications using PGHD to support pre-care processes in elective care. Risk of bias was assessed using validated tools appropriate to each study design. Narrative synthesis addressed heterogeneity across outcome domains. RESULTS: Twenty-one studies analyzed PGHD from free text (38%), questionnaires (33%), voice recordings (14%), wearables (10%), and images (5%). Most used classical machine learning (67%), with deep learning present in 43% of studies and large language models emerging recently (14%). Model performance appeared promising, with area under the curve values ranging from 0.64 to 0.98 (median 0.78). However, this evidence has serious limitations: risk of bias was high in 95% of studies, external validation occurred in only 6% of evaluations, and clinical outcomes were measured in just one study. Equity was assessed in only 14% of studies. No study demonstrated patient benefit or described routine clinical deployment. CONCLUSION: Current evidence establishes proof-of-concept but not proof-of-benefit. The field requires a methodological shift from algorithm development toward prospective validation, clinical outcome measurement, and equity assessment before deployment can be justified.
Artificial intelligence in predicting anesthetic complications: current techniques, clinical applications, and limitations.
Mohammadi A
Int J Med Inform · 2026 Sep · PMID 42248065 · Publisher ↗
Artificial intelligence (AI) is revolutionizing anesthesiology by enhancing the prediction and management of perioperative complications, including intraoperative hypotension, respiratory failure, postoperative nausea an...
Artificial intelligence (AI) is revolutionizing anesthesiology by enhancing the prediction and management of perioperative complications, including intraoperative hypotension, respiratory failure, postoperative nausea and vomiting (PONV), and pain control challenges. This scoping review synthesizes evidence from 82 studies, identified through a systematic search of PubMed, Scopus, Web of Science, and grey literature from January 2010 to September 2025, to map AI techniques, their clinical applications, and limitations. Techniques include Machine Learning (ML) (e.g., random forests, support vector machines), deep learning, natural language processing (NLP), Computer Vision, Bayesian models, and fuzzy logic, applied across preoperative, intraoperative, and postoperative phases. AI models achieve superior predictive accuracy (AUC 0.85-0.94) compared to traditional methods (AUC 0.76-0.88), enabling early detection of complications and reducing opioid use by 15-35%. Applications include preoperative risk stratification, intraoperative monitoring, and postoperative analgesia optimization. Challenges include algorithmic bias, data reliability, interoperability, and real-time integration barriers. Ethical considerations emphasize transparency, equity, and clinician oversight. This review positions AI as a decision-support tool within the P4 medicine framework (Predictive, Preventive, Personalized, Participatory), advocating for validation, ethical frameworks, and integration with anesthesia information management systems (AIMS) to enhance perioperative safety.
A VR obstetric simulator based on the Novint Falcon haptic device: development and evaluation.
Ordoño López C, Molina Massó JP, García Bravo AB … +2 more , García AS, Fernández-Caballero A
Int J Med Inform · 2026 Sep · PMID 42248064 · Publisher ↗
BACKGROUND: Digital vaginal examination is the technique that midwives use to assess the stage of labour, but its learning process is based on theory, mannequins and mostly practice with pregnant women, which often disco...
BACKGROUND: Digital vaginal examination is the technique that midwives use to assess the stage of labour, but its learning process is based on theory, mannequins and mostly practice with pregnant women, which often discomforts those women and does not assure learning all kinds of situations that midwives can face. Virtual reality can greatly help in learning this technique, but current virtual medical simulators are based on consumer headsets and controllers that do not provide the required haptic feedback. OBJECTIVE: To develop a realistic, but also affordable, virtual reality obstetric simulator for training the palpation technique on different cervix's conditions and foetal head orientations. Specifically, this work is focused on the examination that midwives perform to find the fontanelles in the foetal head and thus determine its orientation, which is key to anticipating problems during labour. METHODS: This development was carried out using the TRES-D methodology, which addresses 3D interfaces and content for virtual reality. The simulator is based on a low-cost haptic device, the Novint Falcon, which allows the user to feel the force produced when touching a virtual object. One midwife was directly involved in this process, and a guided evaluation with four other midwives was later conducted to check its realism and assess its potential. RESULTS: Preliminary tests gathered mostly positive feedback. Participants found the simulator useful and intuitive, and highlighted its tactile realism and detail, as they were able to assess the cervix and foetal configuration, touching the sutures and fontanelles as in a real examination. CONCLUSIONS: The developed obstetric simulator has the potential to become a useful tool in learning this examination technique. The main limitation to overcome in the future, and common to other-even more expensive- haptic devices, is having not only one point of contact but two.
Wearable devices for improving medication adherence in older adults with chronic conditions: A systematic review.
Taborri S, Renzi E, Massimi A … +7 more , Improta A, Di Simone E, Giannetta N, Panattoni N, Dionisi S, Di Muzio M, Amato L
Int J Med Inform · 2026 Sep · PMID 42248063 · Publisher ↗
BACKGROUND: Medication adherence remains a major public health challenge, especially among older adults managing multiple chronic conditions and complex therapeutic regimens. Poor adherence is linked to increased mortali...
BACKGROUND: Medication adherence remains a major public health challenge, especially among older adults managing multiple chronic conditions and complex therapeutic regimens. Poor adherence is linked to increased mortality, complications, hospital readmissions, higher healthcare costs, and reduced quality of life. Wearable devices (WDs), integrated into everyday accessories and equipped with sensors and reminder systems, may offer a practical strategy to support continuous monitoring and improve adherence. Given the still-limited evidence in this area, this review evaluates the impact of wearable devices on medication adherence among adults aged ≥ 65 years with at least one chronic condition, compared with usual care or other educational and technological interventions. METHODS: This systematic review was conducted in accordance with PRISMA 2020 guidelines and registered in PROSPERO (CRD420251045757). A comprehensive search was performed in PubMed, CINAHL, Cochrane Library, Ovid, and EBSCO. Eligible studies included quantitative intervention studies involving community-dwelling adults aged ≥ 65 years with at least one chronic condition. Risk of bias was assessed using RoB 2 and ROBINS-I tools. RESULTS: A total of 1,094 records were identified, and nine studies met the inclusion criteria (seven randomized controlled trials, one quasi-experimental study, one crossover trial), with sample sizes ranging from 10 to 498 participants. WDs were classified into three categories: smartwatches (n = 4), wearable blood pressure (WBP) monitors (n = 4), and ECG monitors (n = 1). Seven of nine studies reported statistically significant improvements in medication adherence in the intervention groups compared with control conditions. Positive effects were observed in 3/4 smartwatch interventions and 3/4 WBP monitor interventions, whereas the ECG-based intervention showed no significant between-group differences. CONCLUSION: Wearable devices may support medication adherence in older adults, particularly when integrated with reminder systems and healthcare professional involvement. These findings highlight the importance of embedding wearable technologies within digital health systems and technology-enabled care models for chronic disease management.
Optimising screening efficiency in evidence synthesis on health Technology: A simulation study using ASReview.
Dias de Oliveira JM, Mello AT, Scandolara DH … +4 more , Celuppi IC, Corrêa Rampinelli VP, Wazlawick RS, Dalmarco EM
Int J Med Inform · 2026 Sep · PMID 42235438 · Publisher ↗
OBJECTIVE: This study evaluated model-configuration and stopping-rule decisions when using active learning-based title-and-abstract screening in health technology evidence syntheses. METHODS: We conducted retrospective s...
OBJECTIVE: This study evaluated model-configuration and stopping-rule decisions when using active learning-based title-and-abstract screening in health technology evidence syntheses. METHODS: We conducted retrospective simulations using seven pre-labelled datasets from systematic, scoping, and overview reviews in health technology. Simulations were implemented with ASReview Makita and compared lightweight configurations based on one-hot encoding or term frequency-inverse document frequency with naive Bayes, logistic regression, random forest, and support vector machine classifiers. Performance was evaluated using normalised recall regret ("loss"), work saved over sampling at 95% (WSS@95) and 100% recall (WSS@100), early recall, and K%-consecutive-irrelevant stopping rules. Repeated simulations and exploratory dataset-level analyses were conducted for the highest-ranked configuration. RESULTS: SVM + TF-IDF (with bigrams) had the most favourable overall performance, with an average loss of 0.08 (95% CI 0.06 to 0.09), WSS@95 of 0.70 (95% CI 0.59 to 0.79), and WSS@100 of 0.50 (95% CI 0.30 to 0.69). At a fixed 7% consecutive-irrelevant stopping rule, all datasets reached at least 95% recall in the main analysis, with mean recall of 98%. In repeated simulations, the fixed 7% rule achieved mean recall of 97%; however, one very low-prevalence dataset did not reach 95% recall until K = 33%. Exploratory analyses suggested that relevant-record prevalence, textual similarity among relevant records, and abstract completeness may help explain variation in model performance and stopping-rule reliability, although these analyses were hypothesis-generating. CONCLUSION: Active learning-based screening reduced workload in these health technology datasets, but its use requires explicit implementation choices. SVM + TF-IDF (with bigrams) was the most pragmatic initial configuration, and a 7% consecutive-irrelevant rule was a useful stopping heuristic. However, stopping decisions should depend on the review's tolerance for missed studies, dataset quality, topic heterogeneity, and available safeguards, rather than on a fixed threshold alone.
Operational taxonomy as a determinant of classification stability in AI-assisted evidence synthesis.
Ball AM
Int J Med Inform · 2026 Sep · PMID 42235437 · Publisher ↗
PURPOSE: Reproducibility in rehabilitation evidence synthesis is influenced not only by search strategy and adjudication architecture but also by the structural clarity of operational taxonomy. This study evaluated wheth...
PURPOSE: Reproducibility in rehabilitation evidence synthesis is influenced not only by search strategy and adjudication architecture but also by the structural clarity of operational taxonomy. This study evaluated whether shared operational definitions support classification stability across AI-assisted adjudication architectures. METHODS: A previously established deduplicated rehabilitation corpus was analyzed using a fixed multi-adjudicator architecture under standardized operational constraints. Inter-architecture concordance (agreement, Cohen's κ, and Gwet's AC1) was assessed. Corpus expansion was modeled through staged database inclusion, and stability bounds were estimated under best- and worst-case perturbation scenarios without re-adjudication of newly identified records. Risk of bias assessment was not performed, as the objective was classification concordance rather than therapeutic effect estimation. RESULTS: High inter-architecture concordance was observed under fixed operational definitions. Sensitivity-envelope modeling identified both stability-preserving and stability-failing boundary conditions. Under worst-case forced-discordance assumptions, κ declined substantially as modeled corpus expansion increased, indicating that robustness was conditional rather than unconditional. CONCLUSIONS: Explicit operational taxonomy may constrain classification variability across AI-assisted adjudication architectures when citation-level metadata are incomplete. AI systems cannot recover procedural specificity that is not encoded within bibliographic records. Because taxonomy was held constant rather than experimentally varied, the relative contribution of taxonomy versus architecture remains an empirical question for future work. Although evaluated within a dry needling corpus, the underlying metadata-signal problem may extend to other intervention domains, with implications for AI-assisted evidence workflows in healthcare decision-making.
Finding safety in the mask: how platform anonymity and the COVID-19 pandemic affect domestic violence disclosures.
Wang K, Aldkheel A, Zhou L
Int J Med Inform · 2026 Sep · PMID 42235436 · Publisher ↗
BACKGROUND: Domestic violence (DV) can result in serious consequences for victims. Social media has emerged as an alternative venue for DV victims to share their experiences while seeking potential support, particularly...
BACKGROUND: Domestic violence (DV) can result in serious consequences for victims. Social media has emerged as an alternative venue for DV victims to share their experiences while seeking potential support, particularly during the COVID-19 pandemic. Despite some studies exploring the characterization of DV disclosure on social media, a systematic investigation of the dimensions of DV characteristics remains lacking. Moreover, empirical evidence for DV disclosure on social media remains limited. In particular, the effects of platform anonymity and the COVID-19 pandemic on DV disclosure are largely overlooked. METHODS: To address the literature gaps, this study examines DV dimensions, including depth, breadth, and severity, through a detailed characterization of victimization disclosure in social media communities. More importantly, it empirically investigates the effects of platform anonymity and the pandemic on DV disclosure by analyzing data collected from different social media platforms, spanning both pre-pandemic and post-pandemic declaration periods. RESULTS: Our results show that the effects of anonymity are consistently strong and statistically significant across depth, breadth, and severity, whereas the effects of the pandemic on social media DV disclosures are less consistent and should be interpreted more cautiously. CONCLUSIONS: The findings not only deepen our understanding of the factors contributing to DV disclosure on social media but also offer valuable guidance for designing platform features and developing effective online interventions to address DV.
Corrigendum to "Management of digital competence sharing in healthcare: Exploring managers' self-assessment and associated factors in a cross-sectional study" [Int. J. Med. Inform. 211 (2026) 106357].
Hammarén M, Pölkki T, Kanste O
Int J Med Inform · 2026 Sep · PMID 42225470 · Publisher ↗
Abstract loading — click title to view on PubMed.
Safety guardrails in patient-facing large language model systems for chronic disease self-management: a realist review.
Zhao Y, Miao Y, Luo Y … +3 more , Guo R, Wang H, Wu Y
Int J Med Inform · 2026 Sep · PMID 42224752 · Publisher ↗
BACKGROUND: Patient-facing large language model (LLM) systems are increasingly proposed as scalable tools for chronic disease self-management support. In this setting, safety depends not only on factual accuracy but also...
BACKGROUND: Patient-facing large language model (LLM) systems are increasingly proposed as scalable tools for chronic disease self-management support. In this setting, safety depends not only on factual accuracy but also on whether outputs are interpretable, trusted, and used safely over time. OBJECTIVE: To explain how safety guardrails shape outcomes in patient-facing LLM-supported self-management across different task-risk, user, and interaction contexts. METHODS: We conducted a realist review of LLM-based or LLM-enabled generative conversational systems used for self-management of long-term physical health conditions. Searches of PubMed, Web of Science Core Collection, IEEE Xplore, ACM Digital Library, and arXiv covered all indexed years to 1 April 2026 and used terms for chronic disease or self-management, patient-facing conversational systems, and LLMs or generative AI, identifying 1,154 records before deduplication. The core evidence base comprised 21 studies and 38 context-mechanism-outcome configurations (CMOCs). The patient-facing self-management task, rather than disease label, was the unit of synthesis. RESULTS: At the study level, the dominant patient-facing task was coded as low risk in 6 studies, moderate risk in 11, and high risk in 4; 17 studies evaluated mainly single-turn interactions and 4 included multi-turn, sequential, or simulated-consultation elements. Most evidence concerned simulated or expert-judged patient-facing tasks rather than sustained real-world deployment. Three patterns recurred. Provenance-related safeguards improved transparency and checkability more consistently than they ensured safe downstream action. Communication-oriented safeguards improved readability or perceived comprehensibility while leaving recurring gaps in completeness or actionability. Boundary-control strategies, including source-bounded retrieval, clinician deferral, and escalation support, became more important as task actionability and interaction complexity increased. CONCLUSIONS: In patient-facing chronic disease self-management, safety cannot be judged adequately by answer plausibility alone. This review develops a refined programme theory and a risk-linked, theory-generating heuristic framework, but many proposed mechanisms remain indirect and require real-world, longitudinal, multi-turn testing before deployment.
Clinical agents fail silently on patient identity.
Klang E, Glicksberg BS, Gorenshtein A … +7 more , Gavin N, Freeman R, Stump L, Charney AW, Wei Ting DS, Omar M, Nadkarni GN
Int J Med Inform · 2026 Sep · PMID 42217263 · Publisher ↗
BACKGROUND: LLMs now power clinical agents that plan, call tools, and write into EHRs in routine clinical workflows today. Before clinical deployment, we must know whether agents detect patient identity faults or write t...
BACKGROUND: LLMs now power clinical agents that plan, call tools, and write into EHRs in routine clinical workflows today. Before clinical deployment, we must know whether agents detect patient identity faults or write through them. METHODS: We built a record environment from publicly available MIMIC-IV emergency department data. Agents copied ICD-10-CM codes from visit headers into patient records using Extract and Store tools, with "UNKNOWN" allowed when uncertain. Missing Store calls were counted as non-writes. Each run presented ten visits from one patient (clean), then one visit was tampered. We tested four conditions: clean baseline, one visit with a fully swapped header from another patient, one visit with a one-digit MRN change, and one visit with age shifted. Six models, spanning closed and open weights, completed 1.2 million tool calls. Tamper detection was defined as withholding a write on the tampered visit by outputting "UNKNOWN" or omitting the Store call. RESULTS: Agents usually copied codes into tampered charts. On the tampered visit, GPT-4.1 withheld writing in 17.4% of header-swap runs, but detection of subtle faults (MRN or age changes) was near zero. GPT-4.1-nano detected 4.4% of header swaps and < 1% of MRN or age changes. GPT-5-chat never identified mismatches but produced non-writes in 12.6% of cases. Other models rarely withheld writing. CONCLUSIONS: Clinical agents often fail to detect patient identity inconsistencies. The central risk is misbinding, not miscoding. Safe deployment requires explicit identity verification, abstention when uncertain, and benchmarks that treat record integrity, not just accuracy, as a primary outcome.
Clinical text embeddings: A systematic review of methods, applications, and future directions.
Jo H, Kim S, Son H … +1 more , Kim J
Int J Med Inform · 2026 Sep · PMID 42214285 · Publisher ↗
BACKGROUND: Clinical text embeddings are a foundational component of modern clinical natural language processing (NLP), which function by mapping high-dimensional, heterogeneous clinical texts into vector spaces. This pr...
BACKGROUND: Clinical text embeddings are a foundational component of modern clinical natural language processing (NLP), which function by mapping high-dimensional, heterogeneous clinical texts into vector spaces. This process transforms diverse clinical texts into numerical vectors that capture underlying meaning, enabling different types of clinical texts to be analyzed and compared in a consistent way. The field has progressed with improvements made in general-domain NLP, moving from word-level embeddings to more advanced transformer-based models, and recently to LLM-based embeddings. OBJECTIVES: This review provides a systematic synthesis of various methods for clinical text embedding, and applications across clinical NLP tasks. For discussion, challenges that need to be resolved to successfully bring embedding-based methods into clinical practice are presented. METHODS: This study was conducted as a systematic review following the PRISMA 2020 guidelines. To comprehensively capture both clinical and computational perspectives of clinical text embeddings, literature from PubMed and DBLP published between January 1, 2017 and December 31, 2025 was considered. Out of an initial 17,644 records, quantitative filtering and eligibility criteria were applied to select 59 articles. Selected studies were synthesized into a taxonomy comprising three technical categories: Static Word/Document Embeddings, Contextual Transformer-Based Embeddings, and Knowledge-Enriched & Multimodal Embeddings, along with a survey of their clinical applications. RESULTS: Static embeddings, while limited in context, remain valuable for their efficiency. Contextual transformer-based models have significantly advanced the field through domain-specific pretraining and instruction tuning. Knowledge-enriched and multimodal approaches have been found to further enhance performance by integrating structured knowledge graphs and EHR data. These techniques are actively applied across diverse clinical domains, powering tasks ranging from semantic search to patient risk prediction. CONCLUSION: The paper concludes by proposing future directions necessary to achieve trustworthy, robust, and interoperable clinical text embeddings.
Quality and safety of large language model generated communication in psychodermatology: a bilingual scenario-based evaluation.
Atılan AU, Çetin N, Çöpür A
Int J Med Inform · 2026 Sep · PMID 42214284 · Publisher ↗
BACKGROUND: Large language models (LLMs) are increasingly used by patients for medical information and reassurance. In psychodermatology, where communication must address psychological distress, stigma, and functional im...
BACKGROUND: Large language models (LLMs) are increasingly used by patients for medical information and reassurance. In psychodermatology, where communication must address psychological distress, stigma, and functional impact, the safety and quality of AI-generated educational content have not been systematically assessed. OBJECTIVE: To evaluate the quality and safety of patient-facing educational responses generated by contemporary LLMs for psychodermatologic conditions in a bilingual setting, comparing English and a non-English language (Turkish). METHODS: This cross-sectional, scenario-based evaluation analyzed responses from five LLMs (GPT-4o, GPT-5, Claude 4 Sonnet, Gemini 2.5 Flash, and LLaMA 3.1 70B) to 16 standardized psychodermatology scenarios. Three blinded clinician-evaluators (two from dermatology and one from psychiatry) independently rated each response across six clinical communication domains. Readability indices and word counts were assessed, and repeated-measures nonparametric analyses were performed. RESULTS: Gemini 2.5 Flash achieved the highest overall quality scores in both English and Turkish, significantly outperforming LLaMA 3.1, Claude 4 Sonnet, and GPT-4o (P < 0.05). Across models, empathy and stigma-free communication scored highest, whereas actionability and risk management scored lowest. English outputs were longer than Turkish (mean 373.9 vs 269.4 words; P < 0.001). LLaMA 3.1 showed significantly lower quality in Turkish (66.3%) compared with English (77.2%; P < 0.001). Interrater agreement was good (ICC = 0.703). CONCLUSIONS: While LLMs demonstrated strong empathic and stigma-sensitive communication in psychodermatology, they consistently lacked actionable guidance and robust risk framing. These findings support cautious, clinician-supervised use of LLMs as adjunctive tools for patient education in psychodermatologic care.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe