The Impact of Federal Poverty Level on the Association between Insurance Status and Overall Survival among Adults with Cancer.

Medicaid insurance is associated with worse oncologic outcomes than private insurance and with similar outcomes as being uninsured, even after accounting for area-level socioeconomic status and stage at diagnosis. However, prior studies have not addressed potential confounding due to individual-level socioeconomic status, which determines Medicaid eligibility. Our objective was to quantify insurance disparities in overall survival after accounting for imputed individual-level federal poverty level. We collected data for patients 18-64 years old diagnosed with cancer in 2014-2016 from the Surveillance, Epidemiology, and End Results (SEER) Program. We analyzed these years as increased Medicaid eligibility income limits after 2014 increase comparability of socioeconomic status between the insurance groups. We imputed individual-level percent Federal Poverty Level (FPL) using a model trained on patients from the general U.S. population who participated in the 2016 Behavioral Risk Factors Surveillance Survey (BRFSS) with covariates also present in SEER: insurance status, age, race/ethnicity, sex, marital status, metropolitan residence, and state. Predictive accuracy was assessed with 5-fold cross-validation. The association of 1-year overall survival (OS) and insurance status was estimated using robust regression models accounting for imputation uncertainty, with and without adjustment for estimated individual-level FPL, cancer site, stage at diagnosis, county income and education, and the aforementioned covariates. A total of 308,557 BRFSS respondents were utilized to establish associations between covariates and FPL. The cross-validated accuracy of the model in predicting FPL <100%, 100-150%, 150-200%, and >250% was 84.9%, 79.6%, 61.6%, and 64.3%, respectively. A total of 416,784 cases in SEER were analyzed. The 1-yr OS for patients with private insurance, Medicaid insurance, and no insurance was 88.7%, 76.1%, and 73.7%, respectively. After adjusting for all covariates except individual-level FPL, there was worse 1-yr OS with Medicaid (-6.0%, 95% CI = -6.3 to -5.6, p<.001) and no insurance (-6.7%, 95% CI = -7.3 to -6.0, p<.001) compared to private insurance. However, after also adjusting for estimated individual-level FPL, there was no difference in 1-yr OS between Medicaid and private insurance (-0.4%, 95% CI = -1.9 to 1.1, p=.68), and uninsured individuals had significantly lower 1-yr OS than those with Medicaid (-2.1%, p<.001) or private insurance (-2.6%, p<.001). Subgroup analyses by cancer site showed similar patterns. These are the first data demonstrating similar cancer outcomes between patients with Medicaid and private insurance after accounting for estimated individual-level FPL and other covariates, more completely controlling for potential social determinants of health. Medicaid coverage may improve outcomes among low-income individuals without access to insurance. [ABSTRACT FROM AUTHOR]