PLOS digital health最新文献

{"title":"Socioeconomic determinants of malaria in Ugandan children: An interpretable machine learning approach for public health policy.","authors":"Cleiane Gonçalves Oliveira, Marcos Flávio S V D'Angelo, Matheus P Libório, Marcelo Perim Baldo, Hasheem Mannan","doi":"10.1371/journal.pdig.0001152","DOIUrl":"10.1371/journal.pdig.0001152","url":null,"abstract":"<p><p>Malaria remains a critical global health crisis, placing a disproportionate burden on children under five in Uganda. To transition from broad surveillance to targeted intervention, this study applies interpretable machine learning to identify key socioeconomic predictors of malaria using the 2018-2019 Uganda Malaria Indicator Survey. By employing Random Forests for feature selection and Decision Trees for classification, we addressed the inherent class imbalance using robust metrics such as the F2-score, Matthews Correlation Coefficient, and Precision-Recall Curve. Specifically, the Random Under-Sampling technique enabled the model to achieve a Recall of 77%, prioritizing the reliable detection of true positives over simple accuracy. The analysis highlights the hierarchical importance of determinants such as household size, mosquito net ownership, and maternal education. The study's defining contribution is the extraction of explicit \"if-then\" rules that visualize how these factors combine to create risk profiles, particularly revealing distinct disparities across regions such as Busoga and West Nile. These interpretable findings empower policymakers with actionable, evidence-based insights, moving beyond simple prediction to facilitate the design of structural and region-specific public health strategies.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001152"},"PeriodicalIF":7.7,"publicationDate":"2026-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13541157/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148889734","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deceptive bias measurement in deep learning: Assessing shortcut reliance in TCGA cancer models.","authors":"Farnaz Kheiri, Shahryar Rahnamayan, Masoud Makrehchi","doi":"10.1371/journal.pdig.0001165","DOIUrl":"10.1371/journal.pdig.0001165","url":null,"abstract":"<p><p>Machine learning bias is a persistent challenge because it can create unfair outcomes, limit generalization, and reduce trust in real-world applications. A key source of this problem is shortcut learning, where models exploit signals linked to sensitive attributes, such as the data source or collection site, instead of relying on task-relevant features. To address this, we propose the Deceptive Signal metric, a novel quantitative measure designed to assess the extent of a model's reliance on hidden shortcuts during the learning process. This metric is derived via the Deceptive Bias Detection pipeline, which isolates shortcut dependence by contrasting the model's behavior under two controlled conditions: (1) Full Exclusion, where a sensitive subgroup is completely removed from training; and (2) Partial Exclusion, where the model has limited access to specific classes within that subgroup. By calculating the behavioral shift between these settings, the Deceptive Signal metric provides a quantitative estimate of the model's susceptibility to learning task-irrelevant patterns. In experiments with the TCGA histopathology dataset, our metric successfully quantified substantial dependencies on center-specific artifacts in models trained for cancer classification.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001165"},"PeriodicalIF":7.7,"publicationDate":"2026-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13541255/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148889683","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Community-based malaria surveillance in Cambodia using a mobile reporting app for village malaria workers: An implementation case study.","authors":"Pengby Ngor, Huy Rekol, Siv Sovannaroth, Po Ly, Eng Thyda, Sok Kimleng, Yoem Rattana, Ou Vunsokserey, Hem Vanna, Ran Sany, Chang Pumeoung, Steven Mellor, Rady Try, Pascal Ringwald, Richard J Maude, Lisa J White","doi":"10.1371/journal.pdig.0001626","DOIUrl":"10.1371/journal.pdig.0001626","url":null,"abstract":"<p><p>To support malaria elimination, Cambodia implemented a locally developed, nationally integrated, community-based mobile digital surveillance system enabling real-time, case-based detection, reporting and response. This implementation case study describes the development, national deployment and operational performance of an Android-based mobile app used by Village Malaria Workers (VMWs) and Mobile Malaria Workers (MMWs) integrated within the national Malaria Information System (MIS). Developed collaboratively with VMWs, the system enables real-time, geolocated case reporting with offline functionality, built-in validation checks, centralized device management and automated analytics. Evaluation employed three data sources: routine surveillance data (2018-2024), a contextual survey of VMW/MMWs in three high-risk provinces (2024), and a national technical usability survey (2025). Outcomes included reporting timeliness, completeness, surveillance performance indicators aligned with the national malaria elimination framework and user experience. Following national implementation in 2017- 2018, data completeness and timeliness improved. Between 2018 and 2024 reporting coverage increased from 19% to 99% and data completeness has remained ≥99% since 2019. Case notification within 24 hours increased from 2% to 99% and foci investigation within 7 days from 0% to 98%. Malaria testing rates expanded from 135,664 to 715,277 per year. Intensified surveillance was concurrent with a decline in active malaria foci from 129 in 2020 to 0 in 2024. A contextual evaluation including 88 VMW/MMWs showed strong support for surveillance activities, including reporting and data analysis at the community level. The technical evaluation survey found that 94% (904/966) of VMWs were very satisfied/satisfied with the app, with technical problems experienced never/rarely by 82% (795/966). Key operational challenges included intermittent internet connectivity, limited electricity supply, and transport constraints, although offline functionality and data synchronization prevented data loss. Cambodia's experience demonstrates the feasibility and sustainability of large-scale digital surveillance in resource-constrained settings and highlights the importance of integration, training and sustainability through local ownership.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001626"},"PeriodicalIF":7.7,"publicationDate":"2026-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13541183/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148889677","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Implementation of an opioid use disorder (OUD) machine-learning phenotype in real-time for the ADAPT clinical trial.","authors":"Huan Li, Mark Iscoe, John Lutz, Carolina Diniz Hopper, Sabrina Fried, Josue Minaya, Caroline Raymond King, Olga Reykhart, Hyung Paek, Daniella Meeker, Edward R Melnick, R Andrew Taylor","doi":"10.1371/journal.pdig.0001140","DOIUrl":"10.1371/journal.pdig.0001140","url":null,"abstract":"<p><p>We developed and deploy a real‑time, electronic health record‑integrated machine learning phenotype to identify emergency department patients with opioid use disorder for prospective clinical trial screening and buprenorphine initiation. We conducted a multi‑phase study across three emergency departments in a single United States health system from 2014 to 2025. Using visit‑level data available at or before triage, we trained a random‑forest classifier to estimate opioid use disorder risk and embedded scoring in the electronic health record to trigger point‑of‑care alerts for trial eligibility review. A high-specificity computable training label supported retrospective development, and a clinician gold-standard reference was established through structured chart review. Performance was summarized using discrimination, calibration, and threshold-based classification metrics; prospective validation used a stratified random sample of flagged and unflagged encounters. Retrospective discrimination compared to the high-specificity computable training label was high (area under the receiver operating characteristic curve, 0.99; 95% CI, 0.98-0.99; area under the precision-recall curve, 0.92; 95% CI, 0.90-0.94), and calibration plots informed operating‑point selection for real‑time use. In prospective gold‑standard validation (n = 217), 89.3% of model-positive encounters and 95.7% of model-negative encounters agreed with physician adjudication at the prespecified threshold. Because the sample was stratified on model prediction, these are within-stratum agreement estimates; weighting them by the source-population flag rate (28,284 of 866,569 eligible encounters, 3.26%) yields design-weighted estimates of population-level prevalence (7.1%), sensitivity (0.40, 95% CI 0.24-0.64), and specificity (0.996). An electronic health record‑embedded, machine learning phenotype can accurately and feasibly identify emergency department patients with opioid use disorder in real time, streamlining clinical trial enrollment and treatment initiation. Ongoing work will report operational metrics (e.g., alert volume and latency), monitor performance drift and equity across subgroups, and evaluate downstream clinical and trial outcomes.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001140"},"PeriodicalIF":7.7,"publicationDate":"2026-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13541308/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148889732","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Data distribution impacts the performance and generalisability of contrastive learning-based foundation models of electrocardiograms.","authors":"Gul Rukh Khattak, Konstantinos Patlatzoglou, Joseph Barker, Libor Pastika, Boroumand Zeidaabadi, Aidan R Birdi, Jiayu Huo, Ahmed El-Medany, Hesham Aggour, Yixiu Liang, Antonio H Ribeiro, Jeffrey Annis, Antonio Luiz Pinho Ribeiro, Junbo Ge, Daniel B Kramer, Jonathan W Waks, Evan Brittain, Nicholas Peters, Fu Siong Ng, Arunashis Sau","doi":"10.1371/journal.pdig.0001623","DOIUrl":"10.1371/journal.pdig.0001623","url":null,"abstract":"<p><p>Contrastive learning is a widely adopted self-supervised pretraining strategy, yet its dependence on cohort composition remains underexplored. We present Contrasting by Augmented Patient Electrocardiograms (CAPE) foundation model and pretrain on four cohorts (n = 5,203,269), from diverse populations across three continents (North America, South America, Asia). We systematically assess how cohort demographics, health status, and population diversity influence the downstream performance for prediction tasks also including two additional cohorts from another continent (Europe). We find that downstream performance depends on the distributional properties of the pretraining cohort, including demographics and health status. Moreover, while pretraining with a multi-centre, demographically diverse cohort improves in-distribution accuracy, it reduces out-of-distribution (OOD) generalisation of our contrastive approach by encoding cohort-specific artifacts. To address this, we propose the In-Distribution Batch (IDB) strategy, which preserves intra-cohort consistency during pretraining, discourages learning of spurious cohort-specific features, and instead promotes clinically meaningful variability within cohorts. This leads to improved out-of-distribution robustness, with gains of 9-40% in downstream label prediction performance. This work provides insights into pretraining strategies for more clinically deployable and generalisable foundation models.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001623"},"PeriodicalIF":7.7,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13537565/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148882644","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessing body composition via a smartphone computer vision application: High repeatability but method-dependent agreement compared with BODPOD and Inbody.","authors":"Jahliya Sisohor, Jatin P Ambegaonkar, Bryndan Lindsey, Qi Wei, Yosef Shaul, Joel Martin","doi":"10.1371/journal.pdig.0000983","DOIUrl":"10.1371/journal.pdig.0000983","url":null,"abstract":"<p><p>Accurate assessment of body composition is essential for monitoring health status, fitness progress, and disease risk. Traditional methods such as air displacement plethysmography (BODPOD) and bioelectrical impedance analysis (BIA) are widely used to estimate body fat percentage (BF%), fat mass (FM), and fat-free mass (FFM), but can be costly or inaccessible. Smartphone applications utilizing computer vision (CV) offer a promising alternative. This study evaluated the relative agreement and same session repeatability under standardized conditions of a smartphone-based CV application (CVapp) compared to BODPOD and the InBody BIA device. Forty-nine adults (ages 18-70; 29 females, 27 racial and ethnic minority participants) completed two consecutive measurements using BODPOD, InBody, and CVapp in a single session. Differences in BF%, FM, and FFM estimates were analyzed using repeated-measures ANOVA. Agreement metrics included mean absolute error (MAE), root mean square error (RMSE), concordance correlation coefficients (CCC), and Bland-Altman analysis. Subgroup analyses examined differences by sex and minority status. The CVapp yielded higher BF% (mean=+2.2%, P = 0.004) and FM (+1.5 kg, p = 0.015) compared to BODPOD, and lower FFM than InBody (mean=-1.84 kg, p = 0.043). The CVapp agreement with BODPOD (MAE = 4.0, RMSE = 5.0, CCC = 0.86) was weaker than with the InBody (MAE = 3.3, RMSE = 4.3, CCC = 0.89), with wider limits of agreement. All methods showed excellent within session repeatability under standardized conditions (ICC > 0.99). A significant Sex×Method interaction was observed for BF% (p < 0.001), FM (p < 0.001), and FFM (p = 0.013), with females showing greater overestimation of BF% (mean=+3.9%) and FM (mean=+2.7 kg, p < 0.001), and underestimation of FFM (mean=-2.5 kg, p < 0.001) by the CVapp. Among Minority participants, BF% and FM estimates from the CVapp were significantly higher than BODPOD (mean=+2.7%, p = 0.009; mean=+1.8 kg, p = 0.031), and FFM was lower (mean=-1.8 kg, p = 0.035), with significant Method×Minority Group interactions for FM and FFM (p < 0.03). These findings support cautious use of CVapp-derived values for within-person monitoring under similar conditions but suggest they should not be treated as interchangeable with established comparator methods or used as standalone diagnostic estimates of body composition. Future efforts should address the need for more accurate, transparent, and equitable algorithms that are validated across diverse populations and testing contexts.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0000983"},"PeriodicalIF":7.7,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13537566/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148882587","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From assistant to collaborator: A systematic review of the evolution of artificial intelligence in end-stage renal disease care and management.","authors":"Mohan Wang, Caogen Hong, Zhengxing Huang, Fengmin Shao, Yue Gu","doi":"10.1371/journal.pdig.0001635","DOIUrl":"10.1371/journal.pdig.0001635","url":null,"abstract":"<p><p>We aimed to systematically analyze the historical evolution of artificial intelligence (AI) in end-stage renal disease (ESRD) management and propose a developmental framework to map its progression from assistive tools to cognitive collaborators. A systematic review was conducted following PRISMA 2020 guidelines, identifying 100 eligible studies from PubMed, IEEE Xplore, and Web of Science. A three-stage analytical framework was applied to chart the technological evolution of AI in ESRD care: (1) rule-based assistive tools, (2) data-driven learning systems, and (3) emerging large language model- and agent-based cognitive systems. A total of 100 studies were included, all focusing on patients with end-stage renal disease managed through hemodialysis, peritoneal dialysis, or kidney transplantation. Three primary application domains were identified: risk prediction (49.0%), diagnostic support (25.0%), and monitoring and management (26.0%). The analytical framework revealed a developmental progression from interpretable rule-based systems (Stage 1) to high-performing data-driven models (Stage 2), which achieve clinically relevant metrics (e.g., area under the receiver operating characteristic curve [AUC] 0.80-0.90) but often lack external validation. Emerging large language model- and agent-based systems (Stage 3) demonstrate notable versatility but introduce new challenges related to reliability, factual accuracy, and safety alignment. The proposed three-stage framework clarifies AI's technological and functional evolution in ESRD care. This perspective highlights a critical need to bridge the gap between high-performance modeling and validated clinical utility. Future work should focus on robust external validation and the development of frameworks for the safe, reliable, and ethical deployment of next-generation cognitive AI agents.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001635"},"PeriodicalIF":7.7,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13537592/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148882834","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Advancing foundational models in digital health technology adoption: A systematic literature review of multidisciplinary factors.","authors":"Huey Yu Ng, Shamala Ramasamy, Saravanan Muthaiyah, Asnina Anandan, Seong Ting Chen, Thulasi Munohsamy, Gunasekar Thangarasu","doi":"10.1371/journal.pdig.0001690","DOIUrl":"10.1371/journal.pdig.0001690","url":null,"abstract":"<p><p>Digital health technologies (DHTs) such as patient portals, mobile applications, and electronic health records can improve access to healthcare, self-management and care coordination. However, their adoption remains inconsistent. This study systematically reviews the technological, psychological, social, cultural, health-related and environmental factors influencing DHT adoption. This protocol-registered review (PROSPERO: CRD420251056883) was conducted following the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 guidelines. A search of multidisciplinary databases was conducted through the EBSCO Discovery Service (EDS) to identify peer-reviewed English primary studies published between 2015 and 18 June 2025. Although studies published from 2015 onward were searched, only studies published from 2020 onward were retained for synthesis. Two researchers independently applied the Sample, Phenomenon of Interest, Design, Evaluation, Research type (SPIDER) framework to screen articles and extract data. As Covidence operationalises screening using the Population, Intervention, Comparison, Outcome, Study type (PICOS) framework, SPIDER elements were mapped to PICOS to ensure consistency across screening and extraction. Methodological quality was appraised using the Mixed Methods Appraisal Tool (MMAT) to inform interpretation. Data were synthesised using a structured thematic analysis workflow informed by the Thematic Analysis Matrix proposed by Zairul, which operationalises the thematic analysis principles described by Braun and Clarke. Coding, category development and theme generation were managed using ATLAS.ti (Version 24). Eighty-two studies published between 2020 and 2025 met the inclusion criteria. Five themes were identified: (1) access, equity and affordability; (2) usability, engagement and user empowerment; (3) trust, privacy and governance; (4) integration, workforce and sustainability; and (5) clinical effectiveness and quality of care. The findings showed that adoption of DHTs is a multi-layered process shaped by multidisciplinary factors. This review provides researchers, policymakers and healthcare providers with theoretical and practical insights to support sustainable, effective and equitable DHT adoption and to guide the development of future strategies.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001690"},"PeriodicalIF":7.7,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13537594/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148882627","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An interpretable data-driven approach to optimizing clinical fall risk assessment.","authors":"Fardin Ganjkhanloo, Emmett Springer, Erik H Hoyer, Daniel L Young, Holley Farley, Kimia Ghobadi","doi":"10.1371/journal.pdig.0001631","DOIUrl":"10.1371/journal.pdig.0001631","url":null,"abstract":"<p><p>In this study, we aim to better align fall risk prediction from the Johns Hopkins Fall Risk Assessment Tool (JHFRAT) with clinical expert fall risk perception via a data-driven modelling approach. We conducted a retrospective cohort analysis of 54,209 inpatient admissions from three Johns Hopkins Health System hospitals between March 2022 and October 2023. In the absence of a true fall risk ground truth, we apply proxy labels based on clinician choices for the application of targeted preventative interventions, resulting in a total of 20,208 high-risk encounters and 13,941 low-risk encounters. We employed constrained score optimization (CSO) models to recalibrate the JHFRAT scoring weights, while preserving its additive structure and clinical thresholds. Recalibration refers to adjusting item weights so that the resulting score can order encounters more consistently by the study's risk labels, and without changing the tool's form factor or deployment workflow. The CSO model demonstrated significant improvements over the current JHFRAT in classification alignment with the proxy labels (CSO AUC-ROC = 0.91, JHFRAT AUC-ROC = 0.86). This model performance translates to a weekly average of an additional 35 Johns Hopkins Health System patients who are perceived as high risk (per our proxy labels) being classified by JHFRAT as high risk. The ablation analyses also suggest that the CSO model, though outperformed in prediction metrics by the benchmark black-box XGBoost model, is more robust than XGBoost to variations in risk labeling. Our evidence-based approach provides a robust foundation for understanding risk factor contributions to various indicators of clinician-perceived fall risk. Future research can build upon this foundation to improve risk assessment utility as a decision-support tool in clinical practice.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001631"},"PeriodicalIF":7.7,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13533424/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148876186","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The promise and peril of AI for disability health equity: A justice-centered narrative review.","authors":"Jyoti Dalal, Douglas Teodoro, Minerva Rivas Velarde","doi":"10.1371/journal.pdig.0001663","DOIUrl":"10.1371/journal.pdig.0001663","url":null,"abstract":"<p><p>Despite the promising potential of Artificial Intelligence (AI) models to enhance health equities for persons with disabilities, poorly designed applications risk exacerbating inequities. To address this, we conducted a narrative review using disability justice frameworks as an analytical lens to evaluate how AI applications operationalize disability and health equity across various domains. AI models demonstrate potential in early intervention, personalized care, equitable resource allocation, assistive technologies, and health surveillance. However, most AI models rely primarily on biomedical and functional data, often trained on biased datasets that neglect social and structural determinants of disability. As a result, these applications insufficiently capture broader dimensions of wellbeing, including capabilities, recognition, and structural justice, limiting their effectiveness in addressing health inequities. To advance health equity for persons with disabilities, AI applications must incorporate disability-inclusive datasets, participatory co-design with persons with disabilities, auditing for fairness, and real-world validation. Integrating clinical, social, and structural dimensions in a justice-oriented framework can help AI models move beyond narrow biomedical models toward more inclusive, context-sensitive, and equitable health systems.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"5 9","pages":"e0001663"},"PeriodicalIF":7.7,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13533375/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148876156","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}