Case-mix adjustment for evaluation of mortality in cystic fibrosis

Comparison of patient mortality rates in cystic fibrosis (CF) obtained from different institutions requires the use of case-mix adjustment methods to account for baseline differences in patient and disease characteristics. There is no current professional consensus on the use of case-mix adjustment methods for use in comparing mortality rates in CF. Characteristics used for this case-mix adjustment should include those that are different across institutions and are associated with patient survival. They should not include characteristics of disease severity that may be a consequence of effectiveness of treatment. The goal of these analyses was to identify a set of these characteristics of patients or disease that would be useful for case-mix adjustment of CF mortality rates. Data from the Cystic Fibrosis Foundation Patient Registry and from the United States Census of the Population (1990) were used in these analyses. Kaplan-Meier techniques, the log-rank test, and Cox proportional hazards regression were used to estimate survivorship, calculate hazard ratios (HR), 95% confidence intervals (CI(95%)), and to conduct tests of statistical significance. The data set included all 30,469 CF patients seen at CF Care Centers from 1982-1998. There were 5,906 deaths during 508,721 person-years of follow-up. In multivariate analyses, female gender (HR 1.30, CI(95%) (1.16, 1,47), P < 0.001), nonwhite race (HR 1.48, CI(95%) (1.07, 2.04), P = 0.018), Hispanic ethnicity (HR 1.85, CI(95%) (1.42, 2.43), P < 0.001), and symptomatic presentation (respiratory, gastrointestinal, respiratory and gastrointestinal, meconium ileus, and other symptomatic presentations; HRs 1.38-1.83; P values, 0.028 to < 0.001) were associated with higher risk of death. The homozygous Delta F508 genotype (HR 1.36, CI(95%) (1.19, 1.55), P < 0.001) and neither mutation being Delta F508 (HR 1.40, CI(95%) (1.15, 1.71), P = 0.001) were also associated with higher risk of death. Patients diagnosed after 36 months of age had almost 50% reduction in risk of death compared to those diagnosed before 6 months of age (HR 0.52 CI(95%) (0.44, 0.61), P < 0.001). When patients living in zip codes with a median household income > $50,000/year (corrected for the 1999 consumer price index) were compared with those living in areas with a median household income < $20,000/year, it was apparent that those in the wealthier areas had a 40% reduced risk of death (HR 0.60, CI(95%) (0.44, 0.82), P = 0.001). All of these characteristics were independently significant predictors of death, and all of these characteristics differed significantly across the CF Care Centers. This case-mix adjustment model uses patient and disease characteristics available at the time of diagnosis of CF, and is not believed to be influenced by subsequent treatment to predict the risk of death. If these case-mix adjustment methods are adopted broadly, they will make it possible to study treatment effects and differences in mortality outcomes, while adjusting for baseline differences in patient and disease characteristics.