Predicting Risk of Long-Term Institutionalization Among Community Dwelling Veterans Before the COVID-19 Pandemic.

IF 3.2 2区医学 Q2 HEALTH CARE SCIENCES & SERVICES

Health Services Research Pub Date : 2025-07-26 DOI:10.1111/1475-6773.70016

Bruce Kinosian, Susan Schmitt, Matthew Augustine, Scotte Hartronft, Rajesh Makineni, Kimberly Judon, Gregory Krautner, Cheryl Schmitz, Mary K Goldstein, Ciaran S Phibbs, Orna Intrator

{"title":"Predicting Risk of Long-Term Institutionalization Among Community Dwelling Veterans Before the COVID-19 Pandemic.","authors":"Bruce Kinosian, Susan Schmitt, Matthew Augustine, Scotte Hartronft, Rajesh Makineni, Kimberly Judon, Gregory Krautner, Cheryl Schmitz, Mary K Goldstein, Ciaran S Phibbs, Orna Intrator","doi":"10.1111/1475-6773.70016","DOIUrl":null,"url":null,"abstract":"Objective: To identify risk of long-term institutionalization (LTI) among Veterans receiving care in the Veterans Health Administration (VA).Study setting and design: We developed the \"Predicted Long-term Institutionalization\" (PLI) risk model for Veterans alive in the community at the end of fiscal-year (FY) 2017 followed for LTI in nursing home (cumulative NH days allowing any acute care and up to 7 days in community > 90 days) during FY2018-FY2019.Data sources and analytic sample: PLI used demographics, diagnoses, prior hospital and nursing home (NH) use, and risk indices for death and frailty from VA and Medicare claims and Minimum Data Set data. Development of PLI used multiple iterations to maximize sensitivity, constrained by achieving a number needed to screen (≤ 8), including age normalization to minimize algorithmic bias. We combined the elevated risk (ER) and common risk (CR) strata-specific predictions from the logistic regression models to identify three tiers of PLI: low risk, moderate risk, and high risk. We describe Veterans' outcomes in FY2018/2019 (LTI, death, hospitalization and VA cost) across the three PLI tiers.Principal findings: For identifying Veterans in LTI, compared to a baseline model that used only VA data as predictors (sensitivity 23%, specificity 98%), calibrating separate ER and CR strata increased sensitivity to 30%, the addition of Medicare data increased sensitivity to 33%, and age-normalization with differential risk strata thresholds increased sensitivity to 41% (specificity 96.6%). The final PLI model (c-statistic = 0.87) identified 3.5% of Veterans in PLI-high risk (13% LTI rate), who accounted for 41% of new LTI, 22% of decedents, 19% of VA cost, and 11% of hospitalizations in FY2018-2019.Conclusions: The PLI score identifies Veterans at high risk of LTI for further assessment and targeting of resources to support continued community residence.","PeriodicalId":55065,"journal":{"name":"Health Services Research","volume":" ","pages":"e70016"},"PeriodicalIF":3.2000,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Health Services Research","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1111/1475-6773.70016","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"HEALTH CARE SCIENCES & SERVICES","Score":null,"Total":0}

引用次数: 0

Abstract

Objective: To identify risk of long-term institutionalization (LTI) among Veterans receiving care in the Veterans Health Administration (VA).

Study setting and design: We developed the "Predicted Long-term Institutionalization" (PLI) risk model for Veterans alive in the community at the end of fiscal-year (FY) 2017 followed for LTI in nursing home (cumulative NH days allowing any acute care and up to 7 days in community > 90 days) during FY2018-FY2019.

Data sources and analytic sample: PLI used demographics, diagnoses, prior hospital and nursing home (NH) use, and risk indices for death and frailty from VA and Medicare claims and Minimum Data Set data. Development of PLI used multiple iterations to maximize sensitivity, constrained by achieving a number needed to screen (≤ 8), including age normalization to minimize algorithmic bias. We combined the elevated risk (ER) and common risk (CR) strata-specific predictions from the logistic regression models to identify three tiers of PLI: low risk, moderate risk, and high risk. We describe Veterans' outcomes in FY2018/2019 (LTI, death, hospitalization and VA cost) across the three PLI tiers.

Principal findings: For identifying Veterans in LTI, compared to a baseline model that used only VA data as predictors (sensitivity 23%, specificity 98%), calibrating separate ER and CR strata increased sensitivity to 30%, the addition of Medicare data increased sensitivity to 33%, and age-normalization with differential risk strata thresholds increased sensitivity to 41% (specificity 96.6%). The final PLI model (c-statistic = 0.87) identified 3.5% of Veterans in PLI-high risk (13% LTI rate), who accounted for 41% of new LTI, 22% of decedents, 19% of VA cost, and 11% of hospitalizations in FY2018-2019.

Conclusions: The PLI score identifies Veterans at high risk of LTI for further assessment and targeting of resources to support continued community residence.

查看原文本刊更多论文

在COVID-19大流行之前预测社区居住退伍军人长期机构化的风险

目的：了解在退伍军人健康管理局（VA）接受护理的退伍军人长期机构化（LTI）的风险。研究设置和设计：我们为2017财年（FY）末在社区生活的退伍军人开发了“预测长期机构化”（PLI）风险模型，随后在2018- 2019财年期间在养老院进行LTI（允许任何急性护理的累计NH天数和最多7天的社区bb0 - 90天）。数据来源和分析样本：PLI使用了人口统计学、诊断、以前的医院和疗养院（NH）使用情况，以及来自VA和Medicare索赔和最小数据集数据的死亡和虚弱风险指数。PLI的开发使用多次迭代来最大限度地提高灵敏度，受限于实现筛选所需的数量（≤8），包括年龄归一化以最大限度地减少算法偏差。我们结合了来自逻辑回归模型的高风险（ER）和普通风险（CR）的分层预测，确定了PLI的三个层次：低风险、中等风险和高风险。我们描述了退伍军人在2018/2019财年的结果（LTI、死亡、住院和VA费用）。主要发现：对于识别LTI退伍军人，与仅使用VA数据作为预测因子的基线模型相比（敏感性23%，特异性98%），校准单独的ER和CR层将敏感性提高到30%，增加医疗保险数据将敏感性提高到33%，年龄标准化与不同风险层阈值将敏感性提高到41%（特异性96.6%）。最终的PLI模型（c-statistic = 0.87）确定了3.5%的PLI高风险退伍军人（LTI率为13%），他们占2018-2019财年新LTI的41%，死者的22%，VA成本的19%和住院人数的11%。结论：PLI分数确定了LTI高风险的退伍军人，以进一步评估和确定资源目标，以支持继续社区居住。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Health Services Research 医学-卫生保健

CiteScore

4.80

自引率

5.90%

发文量

193

审稿时长

4-8 weeks

期刊介绍： Health Services Research (HSR) is a peer-reviewed scholarly journal that provides researchers and public and private policymakers with the latest research findings, methods, and concepts related to the financing, organization, delivery, evaluation, and outcomes of health services. Rated as one of the top journals in the fields of health policy and services and health care administration, HSR publishes outstanding articles reporting the findings of original investigations that expand knowledge and understanding of the wide-ranging field of health care and that will help to improve the health of individuals and communities.