Administrative databases, case-mix adjustments and hospital resource use: the appropriateness of controlling patient characteristics.

Hospital administrative databases are used for studying resource use and medical outcomes. The ability to use administrative data to make comparisons among providers requires accurate adjustment of rates based on case mix. Adjustment for case mix often incorporates pre-existing patient conditions such as comorbidities. We tested the hypothesis that some circulatory comorbidities can appear positively or negatively associated with percutaneous transluminal coronary angioplasty (PTCA), not for clinical reasons, but because of the population used for modeling. When statistical models included all discharges with a principal or primary diagnosis of coronary atherosclerosis, or angina with coronary atherosclerosis, multivariate analysis revealed that discharges with dysrhythmias and the more severely ill were less likely to receive PTCA. However, when analysis excluded discharges treated with options (e.g. bypass) reserved for patients with more severe conditions, the presence of dysrhythmias and more severe illness increased the odds of receiving PTCA. Variability involving the direction of association between patient characteristics and a specific intervention illustrates that rates adjusted for patient characteristics cannot be properly interpreted without a clear understanding of the rationale underlying strategies for case-mix adjustment.