Cardiovascular Disease Risk and Noncardiovascular Chronic Disease Burden by Housing Status Authors: Haoyun Hong, Zihang Gao, Sadiya S. Khan, Yingying Sang, Haoyuan Wang, Chuan Hong, Michael J. Pencina, Jennifer L. Hall, Juan Zhao Published: 2026 Cardiovascular disease (CVD) remains the leading cause of death globally, accounting for more than 19.4 million global deaths in 2021, and 941,652 deaths in the United States. 1 Clinical models, such as the Predicting Risk of Cardiovascular Events (PREVENT) equations, are used to estimate 10‐year CVD risk and guide preventive efforts. 2 Social and environmental conditions, such as housing status were associated with health risk through economic, behavioral, and neighborhood pathways. 3 , 4 However, little is known about whether housing stability adds value beyond clinical predictors or extends to broader non‐CVD chronic disease burden. We examined whether housing status (owing or renting) aids in clinical CVD risk prediction and parallels disparities in non‐CVD health conditions. This study used deidentified electronic health record (EHR) data from the Truveta Inc. data platform, which compiles medical records from 30 US health care organizations. Individual socioeconomic data including housing status (ownership or renting) and income were linked from LexisNexis. Income was provided by LexisNexis as a continuous estimated value, although some records were reported as ranges. For those records, the midpoint of each range was used, and when multiple entries were available for the same individual, their values were averaged. This study performs analysis of deidentified EHR data accessed via Truveta Studio. Deidentification is attested to through expert determination in accordance with the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule. Because the study used only deidentified patient records, it did not require institutional review board approval. The data used in this study are available to all Truveta subscribers and may be accessed at studio.truveta.com. Eligible patients included adults (aged ≥18 years) with at least one outpatient visit between January 1, 2010, and December 31, 2016, without prior CVD and with complete PREVENT variables (systolic blood pressure, total cholesterol, high‐density lipoprotein cholesterol, body mass index, serum creatinine, diabetes, smoking status, use of antihypertensive medication or lipid‐lowering medication). Patients with missing housing status were excluded. Baseline was the first outpatient encounter with a serum creatinine measurement during the baseline period (2010–2016). The primary outcome was incident CVD, defined as a composite of myocardial infarction, stroke, or heart failure diagnosis after the baseline visit. The secondary outcome was the total count of incident major chronic non‐CVD conditions occurring after baseline visit, including chronic obstructive pulmonary disease, cancer, arthritis, mental health disorders, asthma, liver disease, gastrointestinal disorders, neurological disorders, osteoporosis, thyroid disorders, and HIV/AIDS. Diagnoses were identified through International Classification of Diseases, Ninth Revision (ICD‐9)/International Classification of Diseases, Tenth Revision (ICD‐10), codes. CVD risk at baseline was calculated using the base PREVENT model. The observed CVD/non‐CVD burden was predicted using the PREVENT risk and housing status. Associations between housing status and CVD outcome were examined using logistic regression. Non‐CVD chronic disease burden was assessed using a Poisson regression model. Models were adjusted for age, sex, PREVENT variables, income, and health care utilization (defined as total number of outpatient visits divided by follow‐up years). Continuous covariates were modeled flexibly using restricted cubic splines to reduce residual confounding. The cohort included 91 488 individuals with a mean±SD age of 54.5±11.9 years, with a median follow‐up of 8.7 years. Sex, race, and ethnicity were 54.6% female, 64.4% non‐Hispanic White, 8.0% non‐Hispanic Black, 7.5% Hispanic, and 20.2% other (Other refers to other race/ethnicity categories including Asian, Native Hawaiian or Other Pacific Islander, American Indian or Alaska Native, and individuals not mapped to standard race/ethnicity categories), respectively. A total of 17% of patients were renters and 83% were homeowners, with 66.8% of the cohort having commercial insurance, 6.5% government coverage, and 26.7% unknown status of insurance (ie, missing or unspecified insurance in EHR data). Figure 1A shows that under the same level of predicted CVD risks, CVD incidence was higher among renters than homeowners. Similarly, renters had a higher average number of non‐CVD conditions than homeowners at the same predicted risk level (Figure 1B). Adjusted analysis confirmed findings that renting was associated with higher odds of CVD outcomes compared with owning (odds ratio [OR], 1.368 [95% CI, 1.235–1.512]), after adjusting for PREVENT risk factors. This association remained significant after further adjustment of income and health care utilization (OR, 1.334 [95% CI, 1.201–1.489]). For non‐CVD outcomes, renting was also associated with an increase in non‐CVD events (incidence rate ratio [IRR], 1.05 [95% CI, 1.04–1.06]). The association remained significant after adjusting for income and healthcare utilization (IRR, 1.02 [95% CI, 1.001–1.03]). A, Average observed CVD conditions by predicted PREVENT risk, stratified by housing status. Error bars indicate 95% CIs. Statistical significance was assessed using Wilcoxon rank‐sum tests (*P<0.05, **P<0.01, ***P<0.001). At the highest risk level (≥25%), >20% of homeowners experienced myocardial infarction, stroke, or heart failure during follow‐up. B, Average observed non‐CVD conditions by predicted PREVENT risk, stratified by housing status. Error bars indicate 95% CIs. Statistical significance was assessed using Wilcoxon rank‐sum tests (*P<0.05, **P<0.01, ***P<0.001). CVD indicates cardiovascular disease; and PREVENT, Predicting Risk of Cardiovascular Events. Using longitudinal EHR data linked with individual‐level social determinants of health, we found that renters experienced higher cardiovascular and noncardiovascular burden compared with homeowners, even at similar predicted risk. This suggests that housing stability may reflect a level of structural vulnerability not captured by clinical models. Incorporating social determinants such as housing status into EHRs may improve risk stratification and reduce health disparities. Limitations of this study include the following: (1) housing status was binary, without data on homelessness or unstable housing; (2) complete case analysis may introduce selection bias toward individuals with greater access to health care; (3) residual confounding from unmeasured or imprecisely captured factors (eg, behavioral, environmental, or health care access variables) may still exist; and (4) we performed risk‐stratified analysis and modeled incident CVD and non‐CVD burden without using time‐to‐event methods, which may not fully account for differences in follow‐up duration or the timing of event occurrence. Therefore, observed associations between housing status and health outcomes should be interpreted as correlation rather than causation. Sources of Funding The study was funded by the Doris Duke Foundation. Disclosures None. Supporting information STROBE Checklist