Health Services Research最新文献

{"title":"Where Can Artificial Intelligence Assist Cancer Care?: Examining Patient-Centered Communication Dimension Effects.","authors":"Qiwei Luna Wu, Yue Liao, Grace Ellen Brannon","doi":"10.1111/1475-6773.14653","DOIUrl":"10.1111/1475-6773.14653","url":null,"abstract":"<p><strong>Objective: </strong>To explore how aspects of patient-centered communication (PCC) may directly or indirectly predict patients' preferences for artificial intelligences (AIs) versus human medical professionals, based on the stimulus-organism-response model.</p><p><strong>Study setting and design: </strong>As AI gains popularity and researchers explore its application in the medical context, it is important to understand how current patient-provider dynamics involving high technology (e.g., telehealth communication) may shape patients' perceptions of future use of AI, especially in the context of cancer care where patient satisfaction and sense of care continuity are important. Participants were recruited from an online panel in China (June 2024). Structural equation modeling analyzed the relationships among variables, including six PCC dimensions (i.e., exchanging information, fostering healing relationships, making decisions, managing uncertainty, responding to emotions, and enabling patient self-management), communication outcomes (i.e., patient satisfaction, sense of care continuity), and patients' preference of AIs vs. human medical professionals.</p><p><strong>Data sources and analytic sample: </strong>Primary data were collected from an online panel of 495 Chinese cancer patients in China, representative of the gender and age distribution of the overall Chinese population due to quota sampling.</p><p><strong>Principal findings: </strong>Direct predictors of preference for replacing human medical professionals with AIs included lower patient satisfaction (β = -11, p < 0.05), lower ease of use (β = -0.1, p < 0.05), better care continuity (β = 0.15, p < 0.01), providers' attending to emotions (β = 0.17, p < 0.05), and less enablement in self-management (β = -0.17, p < 0.01). Patient satisfaction, ease of use, and care continuity mediated the relationships between different PCC dimensions and patients' preferences for AI use.</p><p><strong>Conclusions: </strong>PCC and communication outcomes are associated with cancer patients' preferences in future AI use. Our study sheds light on how clinicians may improve their communication to educate patients on navigating the cancer care continuum using AI technology.</p>","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":" ","pages":"e14653"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12967909/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144235980","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Closing the Gap: Optimizing Hospital Discharge Planning for Improved Outcomes.","authors":"Weihan Chu, Charles Liao, Arman Sharma, David Svec","doi":"10.1111/1475-6773.70101","DOIUrl":"10.1111/1475-6773.70101","url":null,"abstract":"","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":"61 2","pages":"e70101"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13077775/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147583197","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Determining the Survival Impact and Cost-Effectiveness of Multi-Gene Panel Sequencing in Metastatic Colorectal Cancer With Super Learning Approaches.","authors":"Emanuel Krebs, Deirdre Weymann, Howard J Lim, Stephen Yip, Dean A Regier","doi":"10.1111/1475-6773.70009","DOIUrl":"10.1111/1475-6773.70009","url":null,"abstract":"<p><strong>Objective: </strong>To determine the effectiveness and cost-effectiveness of multi-gene panel sequencing compared to single-gene KRAS testing for metastatic colorectal cancer (mCRC).</p><p><strong>Study setting and design: </strong>British Columbia, Canada (BC) is a provincial single-payer public healthcare system, and it was the first province to publicly reimburse multi-gene sequencing for mCRC. Panels expand treatment de-escalation by expanding RAS testing for more precise targeting of anti-EGFR therapies. Reimbursement of panels remains unequal across healthcare systems given uncertain clinical and economic impacts. Our quasi-experimental study design followed the target trial emulation approach, emulating random treatment assignment with two different methods to examine the sensitivity of estimates: inverse probability of treatment weighting estimated with super learning (SL-IPTW) and 1:1 genetic algorithm-based matching, a machine learning approach. We then estimated mean three-year survival time and costs (public healthcare payer perspective; 2021CAD) and calculated the incremental net monetary benefit (INMB) for life-years gained (LYG) at $50,000/LYG using weighted linear regression and nonparametric bootstrapping, also accounting for inverse probability of censoring weights. Our sensitivity analysis estimated LYG using targeted minimum-based loss estimation (TMLE), a doubly robust approach that also uses super learning.</p><p><strong>Data sources and analytical sample: </strong>Patient-level linked administrative health databases capturing cancer and non-cancer care for all BC adults with a metastatic colorectal cancer between 2016 and 2019.</p><p><strong>Principal findings: </strong>Our study included 892 patients (84.3%) receiving multi-gene panels and 166 (15.7%) receiving single-gene testing. INMB estimates were similar for SL-IPTW ($20,397; 95% CI: $9317, $34,862) and matching ($19,569; 95% CI: $8509, $31,447), with 99.3% and 98.8% probabilities, respectively, of panels being cost-effective. We found statistically significant survival benefits with LYG of 0.31 (SL-IPTW; 95% CI: 0.04, 0.54), 0.25 (matching; 95% CI: 0.03, 0.47) and 0.19 (TMLE; 95% CI: 0.02, 0.37).</p><p><strong>Conclusions: </strong>Survival impacts were robust to super learning approaches. Real-world evidence demonstrates that reimbursing multi-gene sequencing for more precise targeting of mCRC treatments provides value for healthcare systems and clinically important benefits to patients.</p>","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":" ","pages":"e70009"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12967917/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144849688","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adulthood Health Insurance Source for Previous Criminal Legal System Involved Pediatrics.","authors":"Ian A Silver, Samantha A Tosto, Jamie Newsome, Alicia D McKay","doi":"10.1111/1475-6773.70106","DOIUrl":"10.1111/1475-6773.70106","url":null,"abstract":"<p><strong>Objective: </strong>To test the effects of criminal legal system (CLS) involvement during adolescence on health insurance enrollment source during adulthood.</p><p><strong>Study setting and design: </strong>The National Longitudinal Survey of Youth-1997 (NLSY97) is a United States birth cohort study used to conduct our current study. Three discrete treatment groups were created: youth arrested before age 18, youth incarcerated in juvenile facilities before age 18, and youth incarcerated in adult facilities before age 18. The control group included youth with no CLS contact before age 18. Gradient boosted inverse probability weighting (IPW) was used to decrease the differences between the groups across potential confounders. The primary outcome for the current study was participants' health insurance source as an adult (2005 to 2021): no health insurance, Medicaid insurance, employer-provided insurance, family member provided insurance, self-purchased insurance, and other insurance policies.</p><p><strong>Data sources and analytical sample: </strong>The NLSY97 is a secondary data source collected from 1997 to 2021. Of the initial sample of 9907 eligible individuals, 8984 (90.7%) individuals participated in the NLSY97, and 7826 individuals were included in the current study.</p><p><strong>Principal findings: </strong>The analytical sample was 50.5% male, 57.7% white, 27.8% Black, and aged from 12 to 41 during the study period. Being arrested (b = 0.241, p < 0.001, IRR 95% CI = 1.249,1.295; b = 0.261, p < 0.001, IRR 95% CI = 1.264,1.332), incarcerated in a juvenile facility (b = 0.403, p < 0.001, IRR 95% CI = 1.466,1.528; b = 0.431, p < 0.001, IRR 95% CI = 1.489,1.590), or incarcerated in an adult facility (b = 0.616, p < 0.001, IRR 95% CI = 1.816,1.890; b = 0.337, p < 0.001, IRR 95% CI = 1.355,1.447) before age 18 was associated with an increased number of years without health insurance and Medicaid insurance (respectively).</p><p><strong>Conclusion: </strong>This study highlights that youth involved in the CLS before age 18 might rely on Medicaid insurance or possess no health insurance during adulthood, highlighting the importance of state-run medical insurance programs for this population.</p>","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":"61 2","pages":"e70106"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13052054/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147619265","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Correction to \"National Trends in Market Competition for Hospital-Based Pediatric Services: 2011-2018\".","authors":"","doi":"10.1111/1475-6773.70118","DOIUrl":"https://doi.org/10.1111/1475-6773.70118","url":null,"abstract":"","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":"61 2","pages":"e70118"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13100971/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147789764","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The Use of Glucagon-Like Peptide 1 Agonists Among Non-Diabetics: Evidence From Medicare Part D.","authors":"Minji Kim, Kieran Allsop, Joseph F Levy","doi":"10.1111/1475-6773.70098","DOIUrl":"10.1111/1475-6773.70098","url":null,"abstract":"<p><strong>Objective: </strong>To assess the extent of off-label glucagon-like peptide-1 receptor agonist (GLP-1) prescribing among individuals without diabetes in Medicare Part D.</p><p><strong>Study setting and design: </strong>This cross-sectional study included Medicare Part D beneficiaries who initiated GLP-1s. We used a difference-in-differences event-study design with insulin initiators as a control group to interpret time trends in prescribing without diabetes evidence.</p><p><strong>Data sources and analytic sample: </strong>We identified beneficiaries initiating semaglutide or tirzepatide between 2019 and 2023 using a 20% sample of Part D event data. Nondiabetic use was defined as the absence of a diabetes diagnosis or a prescription history of diabetes medications within 1 year before or after 6 months of the first GLP-1 fill.</p><p><strong>Principal findings: </strong>Among 126,263 new GLP-1 users, 9.0% were prescribed without diabetes, rising to 13.8% in 2023. This practice was more common among younger, female beneficiaries with obesity. From 2021 to 2023, the share of initiators without diabetes evidence increased by 13.3 percentage points (95% CI, 12.9-13.6), beyond changes observed among insulin initiators.</p><p><strong>Conclusions: </strong>This growth underscores the substantial budgetary implications for Medicare, particularly amid ongoing policy discussions around expanding coverage of GLP-1s for weight management.</p>","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":"61 2","pages":"e70098"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13077783/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147437890","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enumerating the Oncology Specialist Workforce in Medicaid: Applying a Triangulated Approach.","authors":"Anushree Vichare, Mandar Bodas, Clese Erikson, Pavani Chalasani, Qian Eric Luo","doi":"10.1111/1475-6773.70029","DOIUrl":"10.1111/1475-6773.70029","url":null,"abstract":"<p><strong>Objective: </strong>To develop a novel method for enumerating the oncology specialist workforce triangulating taxonomy codes, board certification data, and clinical diagnosis codes in Medicaid claims, and to describe oncology specialists' Medicaid participation, their patient panels, and ascertain the concentration of types of cancers they treated.</p><p><strong>Study setting and design: </strong>We identified oncology specialists using multiple data sources and conducted an exploratory analysis of their patient panels using multi-state Medicaid claims data. We used cluster analysis of diagnosis code patterns in claims to accurately determine the concentration of cancers by site in oncologists' panels.</p><p><strong>Data sources and analytic sample: </strong>We used data from 2016 to 2020 Transformed Medicaid Statistical Information System (T-MSIS) and physician certification data. We included board-certified oncology physicians specialized in medical and radiation oncology, hematology, hematology-oncology, gynecologic oncology, and pediatric hematology-oncology. To identify surgical oncologists, we combined board certification and Medicare Provider Enrollment, Chain, and Ownership System (PECOS) data. We identified Medicaid beneficiaries with malignant neoplasms by cancer site using ICD-10-CM codes.</p><p><strong>Principal findings: </strong>In 2016, about 89% of oncology specialists participated in Medicaid; this proportion decreased slightly to 86% in 2020. The trends in Medicaid participation and the mean number of beneficiaries differed by oncology specialty. Panels of pediatric hematologist-oncologists had a higher proportion of Hispanic Medicaid beneficiaries with cancer (26%) relative to other specialists. Cluster analysis identified 565 out of 5395 medical oncologists that had high concentration (at least 58%) of breast cancer patients in their panels. Among 6970 hematologist-oncologists, 269 had high concentrations in breast cancer (more than 60%), and 944 in hematological cancer (more than 59%).</p><p><strong>Conclusions: </strong>Our study offers a pragmatic approach to understand the oncology specialist workforce available to Medicaid beneficiaries. The findings provide baseline estimates to track this workforce and provide policymakers with an opportunity to develop targeted strategies to improve access to cancer care.</p>","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":" ","pages":"e70029"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12968063/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144857064","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Income and Geographic Disparities in in-Hospital Mortality Following Cardiovascular Procedures: Evidence From the National Inpatient Sample, 2016-2022.","authors":"Alon Bergman, Ashwin Nathan","doi":"10.1111/1475-6773.70117","DOIUrl":"10.1111/1475-6773.70117","url":null,"abstract":"<p><strong>Objective: </strong>To examine whether income- and geography-related disparities in in-hospital mortality after major cardiovascular procedures arise from differences in patient acuity, hospital characteristics, or inequities within hospitals.</p><p><strong>Study setting and design: </strong>This observational study analyzed national data on eight major cardiovascular procedures performed between 2016 and 2022. We used multivariable logistic regression with progressive adjustment for demographics, clinical severity (All Patient Refined Diagnosis Related Groups [APR-DRG] risk and severity scores), and hospital characteristics.</p><p><strong>Data sources and analytic sample: </strong>We analyzed secondary data from the National Inpatient Sample including 1,120,235 discharges (weighted N = 5,906,795) representing adults undergoing percutaneous coronary intervention, coronary artery bypass grafting, carotid endarterectomy/stenting, surgical valve replacement, transcatheter valve procedures, non-carotid endarterectomy, aneurysm repair, or peripheral bypass. Patient income was proxied using ZIP code-level median household income quartiles. Geographic location was classified as large metropolitan (≥ 1 million population), smaller metropolitan (50,000-999,999), or non-metropolitan.</p><p><strong>Principal findings: </strong>Lowest-income patients presented with mean APR-DRG risk scores 0.15-0.25 points higher than highest-income patients. After full adjustment with hospital fixed effects, in-hospital mortality was 0.67% points higher (95% CI: 0.08-1.26) among lowest-income patients. Geographic patterns were complex: after adjusting for hospital characteristics, non-metropolitan location was associated with 0.48% points higher mortality, though this was not statistically significant (95% CI: -0.01 to 0.97), and smaller metropolitan areas with 1.03% points higher mortality (95% CI: 0.30-1.76). Between-hospital differences explained 11.6% of mortality variance.</p><p><strong>Conclusions: </strong>Socioeconomic and geographic disparities in mortality following major cardiovascular procedures persist after adjustment for clinical and hospital factors. These disparities remain, with slightly larger point estimates, in within-hospital analyses, suggesting that hospital-level differences alone do not account for observed inequities. Interventions should address both social determinants and intra-hospital inequities. Multilevel interventions targeting both social determinants and within-hospital processes may be needed.</p>","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":"61 2","pages":"e70117"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13086525/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147700708","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Post-Hospital Access to Preferred and High-Quality Skilled Nursing Facilities for Patients With Opioid Use Disorder.","authors":"Talia S Benheim, Momotazur Rahman, Andrew R Zullo, Ashley Z Ritter, Simeon D Kimmel, Patience M Dow","doi":"10.1111/1475-6773.70110","DOIUrl":"10.1111/1475-6773.70110","url":null,"abstract":"<p><strong>Objective: </strong>To examine whether Medicare beneficiaries with opioid use disorder (OUD) encounter limited access to hospitals' highest-volume (i.e., \"preferred\") or high-quality skilled nursing facilities (SNFs) compared to beneficiaries without OUD.</p><p><strong>Study setting and design: </strong>We estimated within-hospital disparities in access to preferred and high-quality SNFs by OUD status using linear probability models and discrete choice models (McFadden-style conditional logistic regression). We defined preferred status using shared hospital-SNF discharge volume and quality using CMS star ratings. In choice models, we matched patients with and without OUD 1:1 on discharging hospital and date, and applied inverse probability weighting and propensity score subclassification to address confounding.</p><p><strong>Data sources and analytic sample: </strong>We used 2017-2021 Medicare inpatient claims to identify Medicare beneficiaries ages 18+ discharged to a SNF following hospitalization.</p><p><strong>Principal findings: </strong>In the full sample (N = 6,490,593), patients with OUD were 2.5 and 3.6 percentage points (pp) less likely to enter preferred and high-quality SNFs, respectively. Among those discharged to preferred SNFs, patients with OUD were 2.0 pp less likely to enter high-quality preferred SNFs. In the matched subsample (n = 156,610), the marginal effect of preferred status on a person being discharged to their closest SNF was 1.1 pp lower for patients with OUD than those without OUD (p < 0.05), but with no significant disparity after inverse probability weighting. When the closest SNF's quality rating increased by 1 star, the probability of entry increased by 0.7 pp for people without OUD but decreased by 0.2 pp for people with OUD (difference = 0.9 pp, p < 0.001), a difference that persisted after weighting.</p><p><strong>Conclusions and relevance: </strong>Publicly-reported star ratings had weaker associations with the SNF placements of Medicare beneficiaries with OUD compared to those without OUD, and preferred referral networks alone did not eliminate these gaps. Regulatory and reimbursement reforms that support SNFs in developing OUD-related care capacity and that promote equitable admissions deserve attention.</p>","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":"61 2","pages":"e70110"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13077784/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147583176","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine Learning Risk Stratification for Older Breast Cancer Survivors: Clinical Care Implications.","authors":"Stephanie B Wheeler, Jason Rotter, Lisa P Spees, Caitlin B Biddell, Justin G Trogdon, Catherine M Alfano, Deborah K Mayer, Michaela A Dinan, Larissa Nekhlyudov, Sarah A Birken","doi":"10.1111/1475-6773.70005","DOIUrl":"10.1111/1475-6773.70005","url":null,"abstract":"<p><strong>Objective: </strong>To develop and validate a clinical risk prediction algorithm to identify breast cancer survivors at high risk for adverse outcomes.</p><p><strong>Study setting and design: </strong>Our national retrospective analysis used cross-validated random forest machine learning models to separately predict the risk of all-cause death, cancer-specific death, claims-derived risk of recurrence, and other adverse health outcomes within 3 and 5 years following treatment completion.</p><p><strong>Data sources and analytic sample: </strong>Our study used the Surveillance and Epidemiology End Results (SEER) registry-Consumer Assessment of Healthcare Providers and Systems (CAHPS) survey (SEER-CAHPS) linked data for survivors diagnosed between 2003 and 2011, with follow-up claims data to 2017.</p><p><strong>Principal findings: </strong>Within the 3-year follow-up period, 372/4516 survivors (mean age 75.1; 81.7% white) in the primary cohort (8.2%) died, 111 from cancer (2.5%), 665 (14.7%) experienced cancer recurrence, and 488 (10.8%) were hospitalized for adverse health outcomes. The algorithm's prediction resulted in 91.9% out-of-sample accuracy (the percent of observations classified correctly) and a 37.6% Cohen's Kappa (i.e., improvement over an uninformed model). Out-of-sample accuracy was 97.5% (44% improvement) for predicting cancer-specific death, 85% (26% improvement) for recurrence, and 89% (28% improvement) for other adverse health outcomes. Important predictors across outcomes included geographic region, age, frailty, comorbidity, time since diagnosis, and out-of-pocket cost responsibility.</p><p><strong>Conclusions: </strong>Machine learning models accurately predicted relevant adverse survivorship outcomes, driven primarily by non-cancer specific factors. Breast cancer survivors at high risk for adverse outcomes may benefit from more intensive care, whereas those at low risk may be more appropriately managed by primary care.</p>","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":" ","pages":"e70005"},"PeriodicalIF":3.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12967901/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144651272","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}