{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,21]],"date-time":"2026-09-21T18:45:48Z","timestamp":1790016348451,"version":"4.0.1"},"reference-count":28,"publisher":"JMIR Publications Inc.","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["JMIR Med Inform"],"abstract":"Abstract<\/jats:title>\n \n Background<\/jats:title>\n The Cox proportional hazards (CPH) model is a common choice for analyzing time-to-treatment interruptions in patients on antiretroviral therapy (ART), valued for its straightforward interpretability and flexibility in handling time-dependent covariates. Machine learning (ML) models have increasingly been adapted for handling temporal data, with added advantages of handling complex, nonlinear relationships and large datasets, and providing clear practical interpretations.<\/jats:p>\n <\/jats:sec>\n \n Objective<\/jats:title>\n This study aims to compare the predictive performance of the traditional CPH model and ML models in predicting treatment interruptions among patients on ART, while also providing both global and individual-level explanations to support personalized, data-driven interventions for improving treatment retention.<\/jats:p>\n <\/jats:sec>\n \n Methods<\/jats:title>\n Using data from 621,115 patients who started ART between 2017 and 2023, in Kenya, we compared the performance of the CPH with the following ML models\u2014gradient boosting machine, extreme gradient boosting, regularized generalized linear models (Ridge, Lasso, and Elastic-Net), and recursive partitioning\u2014in predicting first and multiple treatment interruptions. Explainable surrogate technique (model-agnostic) was applied to interpret the best performing model\u2019s predictions globally, using variable importance and partial dependence profiles, and at individual level, using breakdown additive, Shapley Additive Explanations, and ceteris paribus.<\/jats:p>\n <\/jats:sec>\n \n Results<\/jats:title>\n The recursive partitioning model achieved the best performance with a predictive concordance index score of 0.81 for first treatment interruptions and 0.89 for multiple interruptions, outperforming the CPH model, which scored 0.78 and 0.87 for the same scenarios, respectively. Recursive partitioning\u2019s performance can be attributed to its ability to model nonlinear relationships and automatically detect complex interactions. The global model-agnostic explanations aligned closely with the interpretations offered by hazard ratios in the CPH model, while offering additional insights into the impact of specific features on the model\u2019s predictions. The breakdown additive and Shapley Additive Explanations explainers demonstrated how different variables contribute to the predicted risk at the individual patient level. The ceteris paribus profiles further explored the time-varying model to illustrate how changes in a patient\u2019s covariates over time could impact their predicted risk of treatment interruption.<\/jats:p>\n <\/jats:sec>\n \n Conclusions<\/jats:title>\n Our results highlight the superior predictive performance of ML models and their ability to provide patient-specific risk predictions and insights that can support targeted interventions to reduce treatment interruptions in ART care.<\/jats:p>\n <\/jats:sec>","DOI":"10.2196\/78964","type":"journal-article","created":{"date-parts":[[2025,12,30]],"date-time":"2025-12-30T02:05:05Z","timestamp":1767060305000},"page":"e78964-e78964","source":"Crossref","is-referenced-by-count":3,"title":["Prediction of First and Multiple Antiretroviral Therapy Interruptions in People Living With HIV: Comparative Survival Analysis Using Cox and Explainable Machine Learning Models"],"prefix":"10.2196","volume":"14","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-4238-5880","authenticated-orcid":false,"given":"Donald","family":"Salami","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9481-9540","authenticated-orcid":false,"given":"Emily","family":"Koech","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7802-6206","authenticated-orcid":false,"given":"Janet M","family":"Turan","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5620-6436","authenticated-orcid":false,"given":"Kristen A","family":"Stafford","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3852-6695","authenticated-orcid":false,"given":"Lilly Muthoni","family":"Nyagah","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6487-0166","authenticated-orcid":false,"given":"Stephen","family":"Ohakanu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5415-8717","authenticated-orcid":false,"given":"Anthony K","family":"Ngugi","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1719-2796","authenticated-orcid":false,"given":"Manhattan","family":"Charurat","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1010","published-online":{"date-parts":[[2026,2,4]]},"reference":[{"issue":"2","key":"R1","doi-asserted-by":"publisher","first-page":"187","DOI":"10.1111\/j.2517-6161.1972.tb00899.x","article-title":"Regression models and life-tables","volume":"34","author":"Cox","journal-title":"J R Stat Soc Series B"},{"issue":"1","key":"R2","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1146\/annurev-statistics-033021-014043","article-title":"Fifty years of the Cox model","volume":"10","author":"Kalbfleisch","journal-title":"Annu Rev Stat Appl"},{"key":"R3","unstructured":"Cox DR Oakes D . Analysis of Survival Data. Chapman and Hall; 1987. URL: https:\/\/www.routledge.com\/Analysis-of-Survival-Data\/Cox-Oakes\/p\/book\/9780412244902?srsltid=AfmBOop9h4oTLfhsnPee_a42cacTol4DIAPZfcQtE7YWoTGaGLqw_ai1 [Accessed 11-02-2025] ISBN: 9780412244902"},{"issue":"1","key":"R4","doi-asserted-by":"publisher","first-page":"145","DOI":"10.1146\/annurev.publhealth.20.1.145","article-title":"Time-dependent covariates in the Cox proportional-hazards regression model","volume":"20","author":"Fisher","journal-title":"Annu Rev Public Health"},{"issue":"7","key":"R5","doi-asserted-by":"publisher","first-page":"121","DOI":"10.21037\/atm.2018.02.12","article-title":"Time-varying covariates and coefficients in Cox regression models","volume":"6","author":"Zhang","journal-title":"Ann Transl Med"},{"key":"R6","doi-asserted-by":"crossref","unstructured":"Salami D Koech E Turan JM et al. Using big data to evaluate the impact of differentiated care on first and recurrent HIV treatment interruptions in Kenya. SSRN. Preprint posted online on Dec 23, 2025. [doi: 10.2139\/ssrn.5954474]","DOI":"10.2139\/ssrn.5954474"},{"issue":"4","key":"R7","doi-asserted-by":"publisher","first-page":"1135","DOI":"10.1007\/s41745-021-00283-9","article-title":"Fifty years with the Cox proportional hazards regression model","volume":"102","author":"Andersen","journal-title":"J Indian Inst Sci"},{"issue":"3","key":"R8","doi-asserted-by":"publisher","first-page":"841","DOI":"10.1214\/08-AOAS169","article-title":"Random survival forests","volume":"2","author":"Ishwaran","journal-title":"Ann Appl Stat"},{"key":"R9","doi-asserted-by":"crossref","unstructured":"Bender A R\u00fcgamer D Scheipl F Bischl B . A general machine learning framework for survival analysis. Presented at: Joint European Conference on Machine Learning and Knowledge Discovery in Databases; Sep 14-18, 2020; Ghent, Belgium. [doi: 10.1007\/978-3-030-67664-3_10]","DOI":"10.1007\/978-3-030-67664-3_10"},{"key":"R10","unstructured":"Nagpal C Jeanselme V Dubrawski A . Deep parametric time-to-event regression with time-varying covariates. Presented at: Proceedings of AAAI Spring Symposium on Survival Prediction - Algorithms, Challenges, and Applications 2021; Mar 22-24, 2021; Palo Alto, CA, United States. URL: https:\/\/proceedings.mlr.press\/v146\/nagpal21a.html [Accessed 22-01-2026]"},{"issue":"1","key":"R11","doi-asserted-by":"publisher","first-page":"5","DOI":"10.1023\/A:1010933404324","article-title":"Random forests","volume":"45","author":"Breiman","journal-title":"Mach Learn"},{"issue":"5","key":"R12","doi-asserted-by":"publisher","first-page":"1189","DOI":"10.1214\/aos\/1013203451","article-title":"Greedy function approximation: a gradient boosting machine","volume":"29","author":"Friedman","journal-title":"Ann Statist"},{"key":"R13","unstructured":"Lundberg S Lee SI . A unified approach to interpreting model predictions. arXiv. Preprint posted online on May 22, 2017. [doi: 10.48550\/arXiv.1705.07874]"},{"key":"R14","unstructured":"Monitoring, evaluation, and reporting indicator reference guide MER 2.0. PEPFAR US President\u2019s Emergency Plan for AIDS Relief; 2024. URL: https:\/\/www.state.gov\/wp-content\/uploads\/2025\/01\/FY25-MER-v2.8-Indicator-Reference-Guide_508-Compliant.pdf [Accessed 19-12-2024]"},{"key":"R15","unstructured":"Maksymiuk S Gosiewska A Biecek P . Landscape of R packages for explainable artificial intelligence. arXiv. Preprint posted online on Sep 24, 2020. [doi: 10.48550\/arXiv.2009.13248]"},{"issue":"1","key":"R16","doi-asserted-by":"publisher","first-page":"1370","DOI":"10.1038\/s41598-023-28393-7","article-title":"Comparing machine learning approaches to incorporate time-varying covariates in predicting cancer survival time","volume":"13","author":"Cygu","journal-title":"Sci Rep"},{"issue":"1","key":"R17","doi-asserted-by":"publisher","first-page":"6968","DOI":"10.1038\/s41598-021-86327-7","article-title":"Explainable machine learning can outperform Cox regression predictions and provide insights in breast cancer survival","volume":"11","author":"Moncada-Torres","journal-title":"Sci Rep"},{"key":"R18","doi-asserted-by":"publisher","DOI":"10.3389\/fonc.2023.1049787","article-title":"A comparison of machine learning models and Cox proportional hazards models regarding their ability to predict the risk of gastrointestinal cancer based on metabolic syndrome and its components","volume":"13","author":"Tran","journal-title":"Front Oncol"},{"issue":"1","key":"R19","doi-asserted-by":"publisher","first-page":"129","DOI":"10.1186\/s12902-023-01368-5","article-title":"Comparison of the Cox regression to machine learning in predicting the survival of anaplastic thyroid carcinoma","volume":"23","author":"Xu","journal-title":"BMC Endocr Disord"},{"key":"R20","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pgph.0002105","article-title":"Historical visit attendance as predictor of treatment interruption in South African HIV patients: extension of a validated machine learning model","volume":"3","author":"Esra","journal-title":"PLOS Global Public Health"},{"issue":"1","key":"R21","doi-asserted-by":"publisher","DOI":"10.2196\/44432","article-title":"Predicting treatment interruption among people living with HIV in Nigeria: machine learning approach","volume":"2","author":"Ogbechie","journal-title":"JMIR AI"},{"issue":"1","key":"R22","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-022-16062-0","article-title":"Applying machine learning and predictive modeling to retention and viral suppression in South African HIV treatment cohorts","volume":"12","author":"Maskew","journal-title":"Sci Rep"},{"issue":"7","key":"R23","doi-asserted-by":"publisher","DOI":"10.1093\/ofid\/ofab130","article-title":"Development and validation of a multivariable prediction model for missed HIV health care provider visits in a large US clinical cohort","volume":"8","author":"Pettit","journal-title":"Open Forum Infect Dis"},{"issue":"1","key":"R24","doi-asserted-by":"publisher","first-page":"6421","DOI":"10.1038\/s41598-020-62729-x","article-title":"Predictive analytics for retention in care in an urban HIV clinic","volume":"10","author":"Ramachandran","journal-title":"Sci Rep"},{"issue":"1","key":"R25","doi-asserted-by":"publisher","DOI":"10.1186\/s12874-022-01679-6","article-title":"Survival prediction models: an introduction to discrete-time modeling","volume":"22","author":"Suresh","journal-title":"BMC Med Res Methodol"},{"issue":"2","key":"R26","doi-asserted-by":"publisher","first-page":"352","DOI":"10.1093\/biostatistics\/kxw047","article-title":"Survival trees for left-truncated and right-censored data, with application to time-varying covariate data","volume":"18","author":"Fu","journal-title":"Biostatistics"},{"key":"R27","unstructured":"Supporting re-engagement in HIV treatment services. World Health Organization. 2024. URL: https:\/\/www.who.int\/publications\/i\/item\/9789240097339 [Accessed 22-01-2026]"},{"key":"R28","unstructured":"Differentiation at re-engagement in HIV care: a multi-country workshop. International AIDS Society (IAS). 2024. URL: https:\/\/www.differentiatedservicedelivery.org\/resources\/differentiation-at-re-engagement-in-hiv-care-a-multi-country-workshop\/ [Accessed 17-11-2025]"}],"container-title":["JMIR Medical Informatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/medinform.jmir.org\/2026\/1\/e78964","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T18:11:47Z","timestamp":1782843107000},"score":1,"resource":{"primary":{"URL":"https:\/\/medinform.jmir.org\/2026\/1\/e78964"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,4]]},"references-count":28,"URL":"https:\/\/doi.org\/10.2196\/78964","relation":{},"ISSN":["2291-9694"],"issn-type":[{"value":"2291-9694","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,4]]},"article-number":"v14i2e78964"}}