The literature argues that regions with greater availability of health care services have high‐intensity patterns of care that raise costs but do not improve outcomes. It has been proposed that limiting the supply of medical services may improve efficiency by reducing unnecessary and discretionary care. However, other research suggests that specialized services improve health. Meanwhile, physician supply is being actively increased. This analysis selected patients newly diagnosed with coronary artery disease (CAD) to assess the impact of seeing a cardiologist on testing patterns and risk of myocardial infarction (MI) using both standard risk‐adjustment and instrumental variable (IV) methods.
Patients continuously enrolled in Market‐Scan claims data from 2005 to 2007 residing in 1 of 15 study metropolitan statistical areas (MSAs) were included. American Medical Association Master File physician supply data and Census Bureau data were used to determine the cardiologist density in these MSAs (range, 5‐21 per 100,000 population). Patients newly diagnosed with CAD by a general internist or family physician between January 1, 2006, and June 30, 2006, were categorized based on whether they saw a cardiologist (C+) within 6 months of diagnosis. Claims 1 year prior to the CAD diagnosis were used to determine comorbidities, which were mapped into binary indicator variables using Elixhauser Comorbidity Software. Outpatient and inpatient claims were analyzed in the year subsequent to diagnosis. The number of each test [electrocardiogram (EKG), stress test (ST), echocardiogram (EC)] was the dependent variable in separate linear regressions. Cardiac catheterization (CC) and MI were dichotomized (≥1/0) and analyzed using logistic regression. The cardiologist density in each patient's MSA was used as an IV, which predicted the chance of seeing a cardiologist (F = 25, P < 0.0001).
Of 46,605 patients newly diagnosed with CAD, 57% saw a cardiologist (mean age, 55.4 vs. 53.2 years) within 6 months. Comorbidities were more common among C+ patients, a higher proportion of whom had ≥2 EKGs (43% vs. 28%), ≥2 STs (20% vs. 14%), ≥3 ECs (32% vs. 19%), and ≥1 CC (18% vs. 9%). Regression, adjusting for age, sex, and comorbidities, showed that C+ was associated with having more EKGs (P < 0.001), STs (P < 0.001), ECs (P < 0.001), CC (OR, 2.20; CI, 2.16‐2.24), and MI (OR, 1.67; CI, 1.53‐1.81). Using the predicted probability of C+, IV regression results were consistent with those of the standard risk‐adjustment analysis, except that a higher cardiologist density was associated with fewer EKGs (P < 0.01) and reduced chance of MI (OR, 0.83; CI, 0.70‐0.96).
Standard risk‐adjustment methods revealed a higher rate of all tests and MI among CAD patients seen by a cardiologist, suggesting a higher‐intensity practice style. IV analysis suggests that these findings are accurate, except that the higher rate of EKGs and MI may be a result of unmeasured factors. Specialty care is higher intensity but may improve outcomes.
G. Ruhnke ‐ none