Benchmarking Antibiotic Use in Finnish Acute Care Hospitals by Patient Casemix-adjustment

 Background: Only a few reports on risk-adjusted models of variability of antibiotic use have been presented and most of them used hospital characteristics or case mix index (CMI), an economic parameter calculated using diagnosis-related groups (DRGs) for benchmarking.

Objective: We studied how the ranking order of antibiotic use in different Finnish acute care hospitals changed compared to the observed rankings when the use data was adjusted according to various patient case-mix variables.

Methods: Data on antibiotic use was collected during the national healthcare-associated infection (HAI) prevalence survey in 2005 in Finland in all five tertiary care, all 15 secondary care and 10 (25% of 40) other acute care hospitals. Use of antibiotics in Anatomical Therapeutic Chemical Codes J01 and P01 was measured by use-days/100 patient-days during a 7-day window period and the prevalence of >2 antimicrobials during the study day. Case-mix-adjusted antibiotic use was calculated by using a multivariate logistic regression model and an indirect standardisation method. Parameters in the model included age, sex, severity of underlying disease as measured by McCabe classification and Charlson index, intensive care, hematology, preceding surgery, respirator, central venous and urinary catheters, community-associated infection, pneumonia and bacteraemia as a HAI, other types of HAI and contact isolation due to MRSA.

Results: Of the parameters included in the model, all but preceding surgery, intensive care, respirator and Charlson index remained significant predictors of antibiotic use. The mean observed antibiotic density was 62 use-days/100 patient-days (range by hospitals, 46-99) and the case mix-adjusted use density ranged from 49 to 114. The ranking order changed one position in 12 (40%) hospitals and >2 positions in 15 (50%) hospitals when the case-mix adjusted figures were compared to those observed. In 24 hospitals (80%), the observed antibiotic use density was lower and in 15 (50%) the use density ranking position was lower than expected by the case-mix adjusted use density or ranking order, respectively. The mean of 15% of patients in all hospitals (range by hospitals, 7-29%) received > 2 antimicrobials during the study day.

Conclusions: Patient casemix-adjustment of antibiotic use ranked the hospitals differently from the ranking by observed use. It might be a useful tool for benchmarking hospital antibiotic use. To explore differences in use or possibilities for improvement, adjustment is essential. The variables we used in the model were all related to patient characteristics. However, the best set of easy and widely available parameters that would describe both patient material and hospital activities (e.g. organization type, use of step down units) in terms of antibiotic use is yet to be determined.