 |
The 130th Annual Meeting of APHA |
|
The 130th Annual Meeting of APHA |
Helena Temkin-Greener, PhD1, Derick R. Peterson, PhD2, Alina Bajorska3, Stephen Kunitz, MD, PhD1, Diane Gross, PhD1, and Dana B. Mukamel, PhD1. (1) Department of Community and Preventive Medicine, University of Rochester Medical Center, 601 Elmwood Avenue, Box 644, Rochester, NY 14642, 585 275 7813, helena_greener@urmc.rochester.edu, (2) Department of Biostatistics, University of Rochester, 601 Elmwood Avenue, Box 630, Rochester, NY 14642, (3) Department of Community and Preventive Medicine, University of Rochester School of Medicine and Dentistry, 601 Elmwood Avenue, Box 644, Rochester, NY 14642
In this study, we examine risk-adjusted mortality across 28 PACE sites to understand the importance of person- and site-specific characteristics as determinants of mortality. PACE provides comprehensive medical, long-term care, and support services to older, frail Medicare beneficiaries. Participant level data, for 3308 individuals enrolled after January 1998 and before December 1999 were used to examine the relationship between mortality and the predictors. Cox proportional hazards models were used to estimate conditional hazard ratios (HR), the relative risk of mortality for each variable, holding the others fixed. We found statistically significant variation in risk-adjusted mortality for a number of variables, including: site of enrollment (HR ranging from 0.14 to 2.23, compared to the average); age (HR 1.65-3.49, compared to those <65); race (HR 0.64-1.36, compared to whites); ambulation (HR 1.60); severe cognitive impairment (HR 1.64, MSQ>8 compared to MSQ<3); diagnosis of cancer (HR 1.86 without and 15.11 with chemotherapy), and self-assessed health (HR 0.63-0.84, compared to non-reports). A likelihood-based pseudo-R2 goodness-of-fit statistic was used to compare the importance of participant versus PACE site characteristics in explaining variation in mortality. Of the entropy explained by the full model, 16% may be attributed to the variation across PACE sites, controlling for patient characteristics, while 71% can be attributed to patient characteristics, controlling for site. This study shows that while much of the variation in mortality remains unexplained even by the "full" model, differences between the sites reinforce the old adage that "location is destiny".
Learning Objectives:
Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.