Background: Possible effect modifiers are often considered as confounders when applying pre-defined risk-adjustment models. The aim was to provide evidence of effect modification by gender in comparative evaluations of hospitals on 30-day in-hospital mortality after acute myocardial infarction (AMI).
Methods: Ninety-two Italian hospitals discharging more than 300 patients with a diagnosis of AMI during 2004 were considered. Patients discharged or transferred within 48 h of hospital admission were excluded. Comorbidities recorded in previous and current admissions were used to define patients' health status and to build the adjustment model, in which an interaction term (gender by hospital) was introduced to test the presence of effect modification. The end point was the 30-day in-hospital mortality after AMI.
Results: The study population consists of 38,544 incident events of AMI from 92 Italian hospitals. Eleven hospitals showed a significant effect modification by gender. In one of them, the overall mortality rate was comparable with that of the reference category, but a significant excess risk for women was found [odds ratios (ORs) = 2.3; P < 0.01]. In 10 hospitals, the overall adjusted ORs presented a significant excess mortality compared with the benchmark: three had a significant excess mortality only among females (ranging from 230 to 370%), four only among males (ranging from 110 to 200%), and three among both genders.
Conclusions: An effect modification by gender was found. The results suggest that in comparative hospital performances evaluation, stratification by gender is desirable to investigate possible differences in attitudes and practices of health services in the treatment of men and women.