Objective: To determine the effect of treatment by a cardiologist on mortality of elderly patients with acute myocardial infarction (AMI, heart attack), accounting for both measured confounding using risk-adjustment techniques and residual unmeasured confounding with instrumental variables (IV) methods.
Data sources/study setting: Medical chart data and longitudinal administrative hospital records and death records were obtained for 161,558 patients aged > or =65 admitted to a nonfederal acute care hospital with AMI from April 1994 to July 1995. Our principal measure of significant cardiologist treatment was whether a patient was admitted by a cardiologist. We use supplemental data to explore whether our analysis would differ substantially using alternative definitions of significant cardiologist treatment.
Study design: This retrospective cohort study compared results using least squares (LS) multivariate regression with results from IV methods that accounted for additional unmeasured patient characteristics. Primary outcomes were 30-day and one-year mortality, and secondary outcomes included treatment with medications and revascularization procedures.
Data collection/extraction methods: Medical charts for the initial hospital stay of each AMI patient underwent a comprehensive abstraction, including dates of hospitalization, admitting physician, demographic characteristics, comorbid conditions, severity of clinical presentation, electrocardiographic and other diagnostic test results, contraindications to therapy, and treatments before and after AMI.
Principal findings: Patients admitted by cardiologists had fewer comorbid conditions and less severe AMIs. These patients had a 10 percent (95 percent CI: 9.5-10.8 percent) lower absolute mortality rate at one year. After multivariate adjustment with LS regression, the adjusted mortality difference was 2 percent (95 percent CI: 1.4-2.6 percent). Using IV methods to provide additional adjustment for unmeasured differences in risk, we found an even smaller, statistically insignificant association between physician specialty and one-year mortality, relative risk (RR) 0.96 (0.88-1.04). Patients admitted by a cardiologist were also significantly more likely to have a cardiologist consultation within the first day of admission and during the initial hospital stay, and also had a significantly larger share of their physician bills for inpatient treatment from cardiologists. IV analysis of treatments showed that patients treated by cardiologists were more likely to undergo revascularization procedures and to receive thrombolytic therapy, aspirin, and calcium channel-blockers, but less likely to receive beta-blockers.
Conclusions: In a large population of elderly patients with AMI, we found significant treatment differences but no significant incremental mortality benefit associated with treatment by cardiologists.