A regression model was constructed to identify baseline variables independently and significantly associated with high cost. High cost was defined arbitrarily as a cost > $1,500; this cost equals approximately the 75% percentile. Variables eligible for inclusion in the model (p value < 0.25) were as follows: age, spirometry (no/yes), number of exacerbations the previous year, smoking (nonex-smoker vs active), chronic heart failure, ischemic heart disease, use of theophyllines, oral steroids, long-acting β2 agonists, ipratropium bromide, and inhaled steroids. Fifty-one patients were excluded from the model due to missing values, with a total of 1,459 patients being valid for analysis. The results of the regression model are presented in Table 7
. Chronic heart failure was strongly associated with higher cost (OR, 3.39; 95% CI, 2.33 to 4.95). Active smoking and frequent exacerbations in the past were also associated with higher cost (OR, 1.79; 95% CI, 1.34 to 2.39; and OR, 1.23; 95% CI, 1.15 to 1.31, respectively). Interestingly, having spirometry performed was inversely associated with cost (OR, 0.66; 95% CI, 0.51 to 0.86). The Hosmer-Lemeshow goodness-of-fit test indicated that the model was also well calibrated (p = 0.514), and the C statistics value was 0.76. In this test, a large p value indicates that the model is performing well, which means that there is no large discrepancy between observed and expected values.