Pub Date : 2026-03-17DOI: 10.1038/s41746-026-02546-0
Sophie Christine Eicher, Friederike Fenski, Solveig Behr, Leona Hammelrath, Johanna Boettcher, Carmen Schaeuffele, Christine Knaevelsrud
Evidence suggests that blended therapy combining face-to-face psychotherapy with digital components may reduce treatment dropout, yet definitions of dropout vary widely. This variability is particularly pronounced in blended therapy, where dropout may involve discontinuation of in-person sessions, disengagement from digital components, or both. This study aimed to identify operational definitions of treatment dropout in blended therapy and to examine how different definitions influence dropout rates, treatment outcomes, and usage patterns. A scoping review identified 14 studies reporting operational definitions of dropout. Five synthesized definitions were applied to data from a large blended therapy trial, revealing variation in dropout rates and their associations with depressive symptoms, anxiety, and life satisfaction. Cluster analysis further identified distinct digital usage patterns. These findings highlight the need for transparent and differentiated reporting of dropout definitions in blended therapy research to improve comparability and interpretation across studies.
{"title":"Defining and reporting treatment dropout in blended therapy for mental health: scoping review and analysis","authors":"Sophie Christine Eicher, Friederike Fenski, Solveig Behr, Leona Hammelrath, Johanna Boettcher, Carmen Schaeuffele, Christine Knaevelsrud","doi":"10.1038/s41746-026-02546-0","DOIUrl":"https://doi.org/10.1038/s41746-026-02546-0","url":null,"abstract":"Evidence suggests that blended therapy combining face-to-face psychotherapy with digital components may reduce treatment dropout, yet definitions of dropout vary widely. This variability is particularly pronounced in blended therapy, where dropout may involve discontinuation of in-person sessions, disengagement from digital components, or both. This study aimed to identify operational definitions of treatment dropout in blended therapy and to examine how different definitions influence dropout rates, treatment outcomes, and usage patterns. A scoping review identified 14 studies reporting operational definitions of dropout. Five synthesized definitions were applied to data from a large blended therapy trial, revealing variation in dropout rates and their associations with depressive symptoms, anxiety, and life satisfaction. Cluster analysis further identified distinct digital usage patterns. These findings highlight the need for transparent and differentiated reporting of dropout definitions in blended therapy research to improve comparability and interpretation across studies.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"17 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465221","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-17DOI: 10.1038/s41746-026-02343-9
Alfredo Cesario,Federico Chinni
{"title":"Toward global standards for SaMD: introducing a proposal for Good Digital Medicine Practices (GDMP).","authors":"Alfredo Cesario,Federico Chinni","doi":"10.1038/s41746-026-02343-9","DOIUrl":"https://doi.org/10.1038/s41746-026-02343-9","url":null,"abstract":"","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"14 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147471801","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-17DOI: 10.1038/s41746-026-02389-9
Yiqun Lin, Haoran Sun, Yongqing Li, Rabia Aslam, Lung Fung Tse, Tiange Cheng, Chun Sing Chui, Wing Fung Yau, Victorine R. Le Meur, Meruyert Amangeldy, Kiho Cho, Yinyu Ye, James Zou, Wei Zhao, Xiaomeng Li
Patient-specific bone models are essential for designing surgical guides and preoperative planning, as they enable the visualization of intricate anatomical structures. However, traditional CT-based approaches for creating bone models are limited to preoperative use due to the low flexibility and high radiation exposure of CT and time-consuming manual delineation. Here, we introduce Semi-Supervised Reconstruction with Knowledge Distillation (SSR-KD), a fast and accurate AI framework to reconstruct high-quality bone models from biplanar X-rays in 30 seconds, with an average error under 1.0 mm, eliminating the dependence on CT and manual work. Additionally, high tibial osteotomy simulation was performed by experts on reconstructed bone models, demonstrating that bone models reconstructed from biplanar X-rays have comparable clinical applicability to those annotated from CT. Overall, our approach accelerates the process, reduces radiation exposure, enables intraoperative guidance, and significantly improves the practicality of bone models, offering transformative applications in orthopedics.
{"title":"Real-time reconstruction of 3D bone models via very-low-dose protocols","authors":"Yiqun Lin, Haoran Sun, Yongqing Li, Rabia Aslam, Lung Fung Tse, Tiange Cheng, Chun Sing Chui, Wing Fung Yau, Victorine R. Le Meur, Meruyert Amangeldy, Kiho Cho, Yinyu Ye, James Zou, Wei Zhao, Xiaomeng Li","doi":"10.1038/s41746-026-02389-9","DOIUrl":"https://doi.org/10.1038/s41746-026-02389-9","url":null,"abstract":"Patient-specific bone models are essential for designing surgical guides and preoperative planning, as they enable the visualization of intricate anatomical structures. However, traditional CT-based approaches for creating bone models are limited to preoperative use due to the low flexibility and high radiation exposure of CT and time-consuming manual delineation. Here, we introduce Semi-Supervised Reconstruction with Knowledge Distillation (SSR-KD), a fast and accurate AI framework to reconstruct high-quality bone models from biplanar X-rays in 30 seconds, with an average error under 1.0 mm, eliminating the dependence on CT and manual work. Additionally, high tibial osteotomy simulation was performed by experts on reconstructed bone models, demonstrating that bone models reconstructed from biplanar X-rays have comparable clinical applicability to those annotated from CT. Overall, our approach accelerates the process, reduces radiation exposure, enables intraoperative guidance, and significantly improves the practicality of bone models, offering transformative applications in orthopedics.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"12 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465223","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-17DOI: 10.1038/s41746-026-02553-1
Cindy Welzel,Gökhan Ertaylan,Irina S Babina,Stephen Gilbert
{"title":"Regulating complexity in AI-enabled omics and multi-omics technologies for precision medicine.","authors":"Cindy Welzel,Gökhan Ertaylan,Irina S Babina,Stephen Gilbert","doi":"10.1038/s41746-026-02553-1","DOIUrl":"https://doi.org/10.1038/s41746-026-02553-1","url":null,"abstract":"","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"44 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147471793","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-17DOI: 10.1038/s41746-026-02536-2
Michele Serra, Daniela Alceste, Nicole Jucker, Lotta Haupt, Sebastian Elben, Samuel Müller, Paul J. M. Hulshof, Harro A. J. Meijer, Andreas Thalheimer, Robert E. Steinert, Philipp A. Gerber, Alan C. Spector, Daniel Gero, Marco Bueter
Artificial intelligence (AI) is transforming dietary assessment, yet few tools have been clinically validated against physiological reference methods. This cross-sectional observational validation study conducted under free-living conditions evaluated the validity of SNAQ, an AI-powered image-based dietary assessment app, against doubly labelled water (DLW) in females with obesity. Twenty participants completed a 7-day protocol, including DLW-based measurement of total daily energy expenditure (TDEE) and estimation of total daily energy intake using SNAQ and 24-h dietary recall (24HR). Compared with DLW-derived TDEE (3004 ± 481 kcal/day), SNAQ underestimated energy intake by 25% (bias −817 kcal/day; limits of agreement −3707 to 2073 kcal/day), while 24HR underestimated intake by 50%. Individual-level agreement had negligible within-subject reliability (ICC = 0.00). Despite advanced AI architecture, SNAQ showed systematic group-level underestimation and poor individual-level agreement, underscoring the translational gap between algorithmic performance and clinical feasibility and the need for standardised clinical validation before implementation.
{"title":"Limited validity of an AI-powered app for dietary assessment in females with obesity","authors":"Michele Serra, Daniela Alceste, Nicole Jucker, Lotta Haupt, Sebastian Elben, Samuel Müller, Paul J. M. Hulshof, Harro A. J. Meijer, Andreas Thalheimer, Robert E. Steinert, Philipp A. Gerber, Alan C. Spector, Daniel Gero, Marco Bueter","doi":"10.1038/s41746-026-02536-2","DOIUrl":"https://doi.org/10.1038/s41746-026-02536-2","url":null,"abstract":"Artificial intelligence (AI) is transforming dietary assessment, yet few tools have been clinically validated against physiological reference methods. This cross-sectional observational validation study conducted under free-living conditions evaluated the validity of SNAQ, an AI-powered image-based dietary assessment app, against doubly labelled water (DLW) in females with obesity. Twenty participants completed a 7-day protocol, including DLW-based measurement of total daily energy expenditure (TDEE) and estimation of total daily energy intake using SNAQ and 24-h dietary recall (24HR). Compared with DLW-derived TDEE (3004 ± 481 kcal/day), SNAQ underestimated energy intake by 25% (bias −817 kcal/day; limits of agreement −3707 to 2073 kcal/day), while 24HR underestimated intake by 50%. Individual-level agreement had negligible within-subject reliability (ICC = 0.00). Despite advanced AI architecture, SNAQ showed systematic group-level underestimation and poor individual-level agreement, underscoring the translational gap between algorithmic performance and clinical feasibility and the need for standardised clinical validation before implementation.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"130 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465217","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-17DOI: 10.1038/s41746-026-02547-z
Da Teng, Lihua Tan, Qiyuan Cao, Yanwei Xia, Na Zhang, Jiantao Li, Dan Zhao
For novice medical learners, do the benefits of correct AI explanations outweigh the risks of plausible misinformation? In a randomized trial with 111 students, we found they do not. Our results reveal a significant and problematic asymmetry: misleading AI explanations significantly degraded diagnostic accuracy, while correct explanations offered no significant improvement over a no-explanation control. Misleading explanations reduced diagnostic accuracy and showed no evidence of confidence calibration, such that confidence did not reliably distinguish correct from incorrect responses. This study provides crucial empirical evidence that, without proper safeguards, the harm caused by AI-generated falsehoods in this population and task is more potent and robust than the benefit derived from correct guidance. This finding highlights a fundamental safety challenge for AI in medical education, demanding a strategic pivot towards building learners’ critical appraisal skills. Trial registration: Chinese Clinical Trial Registry (ChiCTR), ChiCTR2500111932, registered on 7 November 2025.
{"title":"Impact of AI misinformation on diagnostic accuracy and confidence calibration in novice medical students","authors":"Da Teng, Lihua Tan, Qiyuan Cao, Yanwei Xia, Na Zhang, Jiantao Li, Dan Zhao","doi":"10.1038/s41746-026-02547-z","DOIUrl":"https://doi.org/10.1038/s41746-026-02547-z","url":null,"abstract":"For novice medical learners, do the benefits of correct AI explanations outweigh the risks of plausible misinformation? In a randomized trial with 111 students, we found they do not. Our results reveal a significant and problematic asymmetry: misleading AI explanations significantly degraded diagnostic accuracy, while correct explanations offered no significant improvement over a no-explanation control. Misleading explanations reduced diagnostic accuracy and showed no evidence of confidence calibration, such that confidence did not reliably distinguish correct from incorrect responses. This study provides crucial empirical evidence that, without proper safeguards, the harm caused by AI-generated falsehoods in this population and task is more potent and robust than the benefit derived from correct guidance. This finding highlights a fundamental safety challenge for AI in medical education, demanding a strategic pivot towards building learners’ critical appraisal skills. Trial registration: Chinese Clinical Trial Registry (ChiCTR), ChiCTR2500111932, registered on 7 November 2025.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"11 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465219","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-17DOI: 10.1038/s41746-026-02542-4
Maddison L. Mellow, Tyman E. Stanford, Timothy Olds, Aaron Miatke, Ashleigh E. Smith, Dorothea Dumuid
Personalised interventions which optimise the balance of physical activity (PA), sleep and sedentary behaviour (i.e., time use) in the 24-h day may be more effective than one-size-fits-all approaches. We present an interactive app to personalise 24-h time use based on individuals’ health and sociodemographic characteristics. Analyses used cross-sectional data from 53,057 UK Biobank participants. Average daily time use was measured using 7-day accelerometry data and expressed as a 24-h composition using isometric log-ratio transformation. Five cognitive composites were derived from web-based tests. Regularized linear regression examined the relationship between 24-h time-use composition and cognition, with sociodemographic and health characteristics as additional predictors. Model estimates were used to estimate optimized cognition based on the interaction of 24-h time-use composition and personal characteristics. Our ‘ideal day’ app delivers personalised 24-h time-use recommendations tailored to individual characteristics. We demonstrate that personalisation of time-use interventions can be achieved in real time using open-source software.
{"title":"An interactive tool to personalise 24-hour activity, sitting and sleep prescription for optimal health outcomes","authors":"Maddison L. Mellow, Tyman E. Stanford, Timothy Olds, Aaron Miatke, Ashleigh E. Smith, Dorothea Dumuid","doi":"10.1038/s41746-026-02542-4","DOIUrl":"https://doi.org/10.1038/s41746-026-02542-4","url":null,"abstract":"Personalised interventions which optimise the balance of physical activity (PA), sleep and sedentary behaviour (i.e., time use) in the 24-h day may be more effective than one-size-fits-all approaches. We present an interactive app to personalise 24-h time use based on individuals’ health and sociodemographic characteristics. Analyses used cross-sectional data from 53,057 UK Biobank participants. Average daily time use was measured using 7-day accelerometry data and expressed as a 24-h composition using isometric log-ratio transformation. Five cognitive composites were derived from web-based tests. Regularized linear regression examined the relationship between 24-h time-use composition and cognition, with sociodemographic and health characteristics as additional predictors. Model estimates were used to estimate optimized cognition based on the interaction of 24-h time-use composition and personal characteristics. Our ‘ideal day’ app delivers personalised 24-h time-use recommendations tailored to individual characteristics. We demonstrate that personalisation of time-use interventions can be achieved in real time using open-source software.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"35 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465220","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-17DOI: 10.1038/s41746-025-02279-6
David Reinecke, Nina Müller, Anna-Katharina Meissner, Gina Fürtjes, Lili Leyer, Claire Wang, Adrian Ion-Margineanu, Nader Maarouf, Andrew Smith, Todd C. Hollon, Cheng Jiang, Xinhai Hou, Abdulkader Al-Shughri, Lisa I. Körner, Georg Widhalm, Thomas Roetzer-Pejrimovsky, Matija Snuderl, Sandra Camelo-Piragua, John G. Golfinos, Roland Goldbrunner, Daniel A. Orringer, Niklas von Spreckelsen, Volker Neuschmelting
Spinal tumor surgery requires rapid tissue diagnosis to guide surgical decisions and further treatment strategies, yet current intraoperative methods are time-intensive and require specialized expertise. No AI systems exist for real-time spinal tumor classification during surgery. We developed SpineXtract, the first AI-powered system for rapid intraoperative spinal tumor diagnosis using stimulated Raman histology (SRH) — a label-free Raman spectromics imaging technique without tissue processing available during surgery. We created a transformer-based classifier optimized for spinal tissue characteristics to identify common tumor types: meningioma, schwannoma, ependymoma, and metastasis. The system was tested in an international, multicenter, simulated, single-arm study using existing SRH datasets (44 patients, 142 slide-images) from three international institutions, with final pathological diagnosis as reference standard. SpineXtract achieved a 92.9% macro-average balanced accuracy (95% CI: 85.5–98.2) within 5 minutes (tumor-specific accuracy range, 84.2–98.6%), while providing quantitative microscopic feedback for granular tissue analysis. Performance remained consistent across institutions (macro balanced accuracy 91.4–92.0%) and outperformed existing brain tumor classifiers by 15.6%. Our results demonstrate clinical applicability, enabling rapid intraoperative diagnosis with performance exceeding current methods, potentially transforming intraoperative diagnostic workflows in spinal tumor surgery.
{"title":"AI-driven label-free Raman spectromics for intraoperative spinal tumor assessment","authors":"David Reinecke, Nina Müller, Anna-Katharina Meissner, Gina Fürtjes, Lili Leyer, Claire Wang, Adrian Ion-Margineanu, Nader Maarouf, Andrew Smith, Todd C. Hollon, Cheng Jiang, Xinhai Hou, Abdulkader Al-Shughri, Lisa I. Körner, Georg Widhalm, Thomas Roetzer-Pejrimovsky, Matija Snuderl, Sandra Camelo-Piragua, John G. Golfinos, Roland Goldbrunner, Daniel A. Orringer, Niklas von Spreckelsen, Volker Neuschmelting","doi":"10.1038/s41746-025-02279-6","DOIUrl":"https://doi.org/10.1038/s41746-025-02279-6","url":null,"abstract":"Spinal tumor surgery requires rapid tissue diagnosis to guide surgical decisions and further treatment strategies, yet current intraoperative methods are time-intensive and require specialized expertise. No AI systems exist for real-time spinal tumor classification during surgery. We developed SpineXtract, the first AI-powered system for rapid intraoperative spinal tumor diagnosis using stimulated Raman histology (SRH) — a label-free Raman spectromics imaging technique without tissue processing available during surgery. We created a transformer-based classifier optimized for spinal tissue characteristics to identify common tumor types: meningioma, schwannoma, ependymoma, and metastasis. The system was tested in an international, multicenter, simulated, single-arm study using existing SRH datasets (44 patients, 142 slide-images) from three international institutions, with final pathological diagnosis as reference standard. SpineXtract achieved a 92.9% macro-average balanced accuracy (95% CI: 85.5–98.2) within 5 minutes (tumor-specific accuracy range, 84.2–98.6%), while providing quantitative microscopic feedback for granular tissue analysis. Performance remained consistent across institutions (macro balanced accuracy 91.4–92.0%) and outperformed existing brain tumor classifiers by 15.6%. Our results demonstrate clinical applicability, enabling rapid intraoperative diagnosis with performance exceeding current methods, potentially transforming intraoperative diagnostic workflows in spinal tumor surgery.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"94 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465222","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Cardiovascular disease remains the leading cause of death and disability worldwide. The convergence of big data and artificial intelligence (AI) is reshaping precision cardiovascular medicine through multimodal integration of electronic health records (EHRs), imaging, omics, and wearable data across the care continuum, enabling predictive, diagnostic, therapeutic, and system-level optimization. However, translation into durable clinical benefit remains constrained by evidentiary gaps, implementation complexity, and fragmented governance architectures.
{"title":"Precision cardiovascular medicine with big data and AI.","authors":"Qian Xu,Yiwen Li,MengMeng Zhu,Yajie Cai,Xi Cheng,Wenting Wang,Jianqing Ju,Yanwu Xu,Yanfei Liu,Yue Liu","doi":"10.1038/s41746-026-02538-0","DOIUrl":"https://doi.org/10.1038/s41746-026-02538-0","url":null,"abstract":"Cardiovascular disease remains the leading cause of death and disability worldwide. The convergence of big data and artificial intelligence (AI) is reshaping precision cardiovascular medicine through multimodal integration of electronic health records (EHRs), imaging, omics, and wearable data across the care continuum, enabling predictive, diagnostic, therapeutic, and system-level optimization. However, translation into durable clinical benefit remains constrained by evidentiary gaps, implementation complexity, and fragmented governance architectures.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"8 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147471798","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-16DOI: 10.1038/s41746-026-02529-1
Andrew Melehy, Jeffrey Feng, Dominic Amara, Vatche G. Agopian, Alex A. T. Bui
Despite advances in liver transplantation (LT), deciding when to transplant a patient within the context of high waitlist mortality, organ scarcity, and risk of graft failure, remains an ongoing challenge. Existing approaches focus on the static prediction of successful LT donor-recipient pairs without weighing the competing interests such as the risk of graft failure against the risk of waitlist mortality and how these risks change over time. Instead, we used an offline reinforcement learning (RL) approach to represent the problem as the optimization of the series of decisions to wait, delist, or transplant a candidate at different timepoints. Using waitlist trajectories for LT candidates from the national Scientific Registry of Transplant Recipients (SRTR) database, we trained a model resulting in the avoidance of 73% of donor-recipient pairs that led to graft failure or death, preservation of 93% of successful transplants, and potentially suitable donors were found for 47% of those patients that died on the waitlist. Notably, the analysis of decisions and post-transplant survival revealed that our model learned features suggestive of successful donor-recipient pairs. Overall, we demonstrate how RL-based approaches better portray real-world LT donor-recipient matching decisions, illustrating their potential as useful clinical tools.
{"title":"Liver transplant donor-recipient matching with offline reinforcement learning","authors":"Andrew Melehy, Jeffrey Feng, Dominic Amara, Vatche G. Agopian, Alex A. T. Bui","doi":"10.1038/s41746-026-02529-1","DOIUrl":"https://doi.org/10.1038/s41746-026-02529-1","url":null,"abstract":"Despite advances in liver transplantation (LT), deciding when to transplant a patient within the context of high waitlist mortality, organ scarcity, and risk of graft failure, remains an ongoing challenge. Existing approaches focus on the static prediction of successful LT donor-recipient pairs without weighing the competing interests such as the risk of graft failure against the risk of waitlist mortality and how these risks change over time. Instead, we used an offline reinforcement learning (RL) approach to represent the problem as the optimization of the series of decisions to wait, delist, or transplant a candidate at different timepoints. Using waitlist trajectories for LT candidates from the national Scientific Registry of Transplant Recipients (SRTR) database, we trained a model resulting in the avoidance of 73% of donor-recipient pairs that led to graft failure or death, preservation of 93% of successful transplants, and potentially suitable donors were found for 47% of those patients that died on the waitlist. Notably, the analysis of decisions and post-transplant survival revealed that our model learned features suggestive of successful donor-recipient pairs. Overall, we demonstrate how RL-based approaches better portray real-world LT donor-recipient matching decisions, illustrating their potential as useful clinical tools.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"11 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147465086","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：