PNS207 MORTALITY PREDICTION IN MEDICARE BENEFICIARIES: THE UTILITY OF HEALTH RELATED QUALITY OF LIFE MEASURES

Abstract

This study was designed to assess the strength, and two-year mortality estimation probabilities of health related quality of life (HRQOL) measures in a recent cohort of Medicare beneficiaries. The Medicare Health Outcomes Survey (HOS) cohort 15 (2012-2014) data was queried for all complete cases for age, gender, race, education, BMI category, marital status and patient reported measures of HRQOL including measures of physical functioning and mental health (primarily depression). The strength of association between HRQOL measures were assessed in multivariate models and a 2-year mortality prediction model was built using logistic regression. For this study, 296,320 patients were identified, and 221,492 complete cases were included for analysis. The respondents were predominantly 65 years or older (87%) and the group’s gender composition was fairly balanced between male and female (50.9% and 49.1%). Three mental health variables were moderately positively correlated to physical functioning (0.3) while one was negatively correlated (-0.06). Association of predicted probabilities and observed responses showed a c-statistic of 72%. The odds ratio of mortality between the levels of physical functioning were (3.8; 1 vs 2, 9.2; 1vs 3 and 2.45; 2 vs 3). The likelihood of mortality increases significantly with deteriorating physical functioning. In the absence of other comorbidities while adjusting for some demographic covariates, mortality increases as physical functioning and mental health declines in Medicare managed care population. Future studies should consider comorbidities that may influence the sensitivity and specificity of the model.