This was a longitudinal, retrospective cohort study using a 5% random sample of the US Medicare database, including patients aged ≥65 years from 01JAN2006-31DEC2013. Medicare provides health insurance coverage to ~ 42 million persons aged ≥65 years as well as nearly 9 million persons aged < 65 years with end-stage kidney disease or a disability. For each beneficiary, claims from all settings of care were linked to create a longitudinal record of their health encounters, diagnoses, and drug prescriptions. No patient identity or medical records were disclosed for the purposes of this study. Since the data used for this study were de-identified and only aggregate results were reported, the study was approved by the Institutional Review Board as exempt. Compliance with all applicable laws and the Health Insurance Portability and Accountability Act (HIPAA) regulations were maintained.
Patients were included in the study if they had ≥1 inpatient claim for HF (International classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM] code 428.xx) during the identification period (01JAN2007-31DEC2013); the discharge date of the first hospitalization was designated as the index date. Additionally, patients were required to have continuous enrollment in their Medicare health plan with medical and pharmacy benefits for ≥1-year pre-index date (baseline period: 1-year pre-index period including the index hospital admission). The follow-up period included the period after the index date until death, disenrollment from health plan, or end of the study period, whichever occurred first. Patients with evidence of HF during the 1-year period prior to the index hospital admission were excluded (Fig. 1).
HF patients with LVSD (ICD-9 CM codes 428.1, 428.20–428.23, 428.40–428.43) on the index date and without evidence of AF (ICD-9-CM code 427.31) in the baseline period (including the index hospital admission) were further stratified based on evidence of significant CAD in the baseline (including the index hospital admission) into CAD and non-CAD cohorts. We have included hospitalized HF patients without atrial fibrillation (AF) given the limited evidence on the burden of clinical outcomes as well as the conflicting evidence on the benefit of anticoagulation use in these patients [4]. Evidence of CAD was defined as having previously documented CAD (ICD-9-CM codes 410.x-414.x, 429.2, V45.81), history of prior coronary artery bypass graft (Current Procedural Terminology [CPT] codes 33,510–33,536 or ICD-9 CM procedure codes 36.10–36.17, 36.19), or history of percutaneous coronary intervention with or without stent (CPT codes 92,980–92,996 or ICD-9-CM codes 00.66, 36.01–36.09). Throughout the manuscript, included hospitalized HF patients with LVSD and without AF will be referred to as “HF with LVSD patients.” Additionally, newly-diagnosed HF patients identified with an inpatient or outpatient claim were included in a sensitivity analysis to see if there existed any differences in this population from the hospitalized HF population given the fact that the study included elderly patients aged ≥65 years.
Baseline measures
Patient demographics including age, sex, race, and US geographic region as of the index date were assessed. Clinical characteristics including Charlson comorbidity index (CCI) score, CHADS2 score (congestive HF, hypertension, age ≥ 75 years, diabetes mellitus, prior stroke, or transient ischemic attack [TIA]), comorbidities (hypertension, hyperlipidemia, arrhythmia, anemia, diabetes, trauma, chronic renal insufficiency, malignant neoplasm, pneumonia, peripheral artery disease, anasarca, chronic obstructive pulmonary disease, dementia, hepatic disease, rheumatoid arthritis, depression, coagulation defect, obesity, varicose veins, thrombophilia, inflammatory bowel disease, arterial embolic events, peptic ulcer, alcohol abuse, pulmonary edema, bleeding diathesis), and prior clinical events (IS, TIA, venous thromboembolism [VTE], major bleeding) during the baseline period were assessed.
Outcome measures
Main outcomes were occurrence of major adverse cardiovascular events including ACM, MI (ICD-9-CM codes: 410,412) and IS (ICD-9-CM codes: 433.01, 433.11, 433.21, 433.31, 433.81, 433.91, 434.01, 434.11, 434.91, 436). ACM rates, and cumulative incidence of MI and IS with death as a competing risk, were estimated by 60-day intervals during the first year of follow-up and by each year during the entire follow-up period (maximum 7 years) among newly-diagnosed “HF with LVSD patients” in the CAD versus non-CAD cohorts.
Statistical analysis
Descriptive statistics (means and standard deviations for continuous variables, numbers and percentages for dichotomous/polychotomous variables) were provided for all study variables, including baseline demographic and clinical characteristics in the CAD and non-CAD cohorts. Statistical tests of significance (chi-square for categorical variables and t-test for continuous variables) were conducted to assess differences between the cohorts. Propensity Score Matching (PSM) was used to achieve baseline balance for patient characteristics. The propensity score was calculated via a logistic regression model, and the covariates adjusted in the model included all demographics, CCI score, CHADS2 score, comorbidities, and prior baseline clinical events. Each CAD patient was matched to a non-CAD patient within 0.01 units of the propensity score. The adequacy of the matching procedure was assessed by standardized difference for each of the matching variables; a difference of < 10% is considered well balanced [11]. Between the CAD and non-CAD cohorts, Kaplan-Meier (KM) curves of ACM were compared using the log-rank test; cumulative incidence distribution for MI and IS was compared using Gray’s test [12, 13]. All analyses were conducted using SAS® statistical software (Version 9.3, SAS Institute, Cary, North Carolina, 2012).
