Clinical and Cost Consequences of Non-Adherence in the End-Stage Renal Disease Population.

Abstract

Results from this study indicate that non-adherence in hemodialysis (HD) sessions marginally affects health care costs and does not significantly increase the likelihood of kidney transplantation failure. This dissertation explores the impacts of non-adherence in HD sessions on health care costs and kidney transplantation failure, and uses a conventional non-adherence measure, a broader data set which includes HD patients nationwide, and rigorous statistical models to tackle these research questions. Informed policy recommendations are especially important because of the difficulty for dialysis patients to adhere to treatments. There are three principal findings from my first paper, which investigates the impact of non-adherence on separately billable (SB) Medicare Allowable Payment (MAP). (1) There is a negative association between non-adherence and SB MAP, consistent across three different regression estimations: OLS, 2SLS, and the log-linear regression. (2) The OLS estimation is more reliable and stable. (3) Results from three lagged effect models do not support the hypothesis that non-adherent dialysis patients would use more SB drug injectables in the following periods. My second paper explores the association between non-adherence and composite rate (CR) costs for dialysis facilities. There are four principle findings. (1) Since CR costs represent a skewed distribution, a log transformation of CR costs might be a better measure for the dependent variable. (2) The explanatory power increases significantly for log-linear models comparing to that for OLS models. (3) There is no association between non-adherence and CR costs except for the log-linear model without facility control variables. (4) Adding facility control variables significantly increases the explanatory power for log-linear models. My third paper explores whether non-adherence is a contributing factor to kidney transplantation failure. The results from the Cox proportional hazards models consistently show that non-adherence in hemodialysis sessions does not have a significant influence on kidney transplantation failure after controlling for none, some, or a full list of patient characteristics. The coefficient estimates from a binary non-adherence measure also confirm this finding. Findings from this dissertation may provide valuable information for dialysis patients, dialysis facilities, and policy makers when faced with concerns related to non-adherence.