Background: Although ICDs provide meaningful prolongation of life for patients in clinical trials at high risk for sudden cardiac death (SCD), data are lacking on the effectiveness of ICDs in patients who receive Medicare benefits.
Aim 1: To demonstrate differences in the incidence rate of death and cardiovascular hospitalizations between trial patients and Medicare patients who are undergoing ICD implantation.
Aim 2: To explore whether the incidence of death and cardiovascular hospitalizations after ICD implantation differ by gender, age category, and comorbidities.
Aim 3: To identify subgroups of patients undergoing ICD implantation for whom: 1) ICDs are not effective (expected survival < 18 months); and 2) ICDs are not cost-effective (expected survival < 5 years).
Aim 4: To estimate the incidence of various potential short and long-term adverse events following ICD implantation.
Aim 5: a) To directly compare the effectiveness of ICD versus medical management and b) to assess the impact of unmeasured confounding by indication or severity in a well-designed comparative effectiveness study.
Study design: Retrospective cohort study
Data Source: 1) the CMS ICD Registry, 2) the Heart failure/Myocardial infarction (HF/MI) registry and 3) Medicare eligibility, Part A, B and D data.
Participants, setting, and expected sample size: The study utilizes the following: a) >400,000 patients in the CMS ICD registry, or b) >50,000 similar patients in the Outcome HF/MI registry and c) patients from the entire Medicare data with heart failure or myocardial infarction who meet a claims-based definition for low-ejection fraction and other National Coverage Decision (NCD) criteria for ICDs but did not receive ICDs. We will further require that these patients: 1) be ≥65 years at the time of receiving ICD or a matched random date; and 2) are eligible for Medicare ≥12 months prior to the date of cohort entry.
Outcomes: The following are study outcomes: all-cause death, hospitalization for heart failure and other cardiovascular events (i.e., myocardial infarction; stroke). To define these events, we will use previously validated definitions with very high specificity.
Statistical methods: Survival analyses and logistic regression will be used to describe and predict outcomes. We will use Cox regression to compare ICD and non-ICD after adjusting for confounders using multivariate adjustment or high-dimensional propensity score (hd-PS).
Comparator groups: No ICDs with medical treatment
Expected Outputs: Scientific reports
Expected Date of Project Completion: 9/30/2011