Many clinical questions remain unanswered about comparative effectiveness of therapeutic options in actual clinical practice. Few randomized clinical trials (RCTs) compare effectiveness of new drugs to older alternative drugs. When the Medicare Modernization Act Prescription Drug Benefit (Part D) is implemented in 2006, there will be even greater need for data on comparative effectiveness. We need to apply newer methodologies that will improve our ability to reliably answer questions on comparative effectiveness usjng large observational datasets. Increasingly effective treatment of ischemic heart disease (IHD) has led to better survival of patients with IHD, but more patients living with heart failure (HF). RCTs have demonstrated that 3 beta-blockers carvedilol, metoprolol succinate, and bisoprolol fumarate improve survival in patients with HF and left ventricular systolic dysfunction. While older, generic beta-blockers (e.g., atenolol, propranolol, timolol) have been shown in RCTs to prolong survival in patients after myocardial infarction, there are no data on whether they improve survival in HF patients. A class of methods-inverse probability weighted (IPW) estimators-is useful in observational analyses with time dependent outcomes, confounding, and censored data. Through a collaboration with Medical Review of NC, a QIO, we propose to apply these estimators to address questions of comparative effectiveness in elderly HF patients who are dually eligible for both Medicare and Medicaid:
Aim 1 : Using IPW estimators, we will compare clinical outcomes (survival and rehospitalization for HF after discharge from an index HF hospitalization) in patients using carvedilol, metoprolol succinate, or bisoprolol fumarate (evidence-based beta-blockers) to outcomes in patients using all other beta-blockers;
Aim 2 : We will compare alternative methods appropriate for censored data including doubly robust estimators, Cox regression models, partitioned estimators, and hybrids of generalized estimating equations;
and Aim 3 : In the Duke Databank for Cardiovascular Disease, we will perform beta-testing of a generalized software program using IPW estimators and doubly robust methods to be developed by researchers at the University of North Carolina CERTs Center. ? ? ?
Showing the most recent 10 out of 14 publications
