Exposure and Outcome

Alcohol use and misuse was estimated using validated Alcohol Use Disorders Identification Test Consumption (AUDIT-C) questions. Participants were categorized into five strata, with AUDIT-C scores of 1 to 3 designated a priori as the referent group. Two additional indexes of alcohol consumption were used in sensitivity analyses: CAGE questionnaire, and self-report of drinking six or more drinks on an occasion (binge drinking). Patients with “yes” responses to two or more CAGE questions were considered to have screened positive for alcohol-use disor-ders. All three questionnaires are depicted in Figure 1.

We defined a COPD exacerbation after the index date as either a primary inpatient discharge diagnosis of COPD (codes 491.x, 492.x, 493.2, 496.x) or an outpatient diagnosis of COPD accompanied by prescriptions for either antibiotics or prednisone within 2 days of the visit. If participants had multiple COPD exacerbations, we used only the first event after the index date.

Other Variables

Smoking status, obtained by self-report, was categorized as never, past, or current smoking. Both smoking intensity and years since quitting were reported in categories and modeled as continuous variables. As a proxy of COPD severity, we assessed the number of canisters of ipratropium bromide and P-agonists prescribed during 12 months prior to the index date. We adjusted for overall comorbidity using the Seattle Index of Comorbidity (SIC), a weighted score derived from self-report of conditions that incorporates history of previous myocardial infarction, cancer, chronic lung disease, chronic heart failure, pneumonia, and cerebral vascular accidents.

Statistical Methods and Analysis

All statistical analyses were performed using statistical software (STATA 8.0; StataCorp; College Station, TX). All statistical tests were two tailed, and p values < 0.05 were considered statistically significant. Table 1 was directly adjusted for age in quintiles. Cox proportional hazard models were used to estimate the risk of COPD exacerbation, adjust for confounding, and assess effect modification. Potential confounding variables were entered individually and en bloc. Sensitivity analyses were planned a priori to assess for the effects of COPD definitions and age-stratified analyses.