A Mathematical Model to Prioritize Healthcare Facilities for High Prevention Impact on Healthcare-Associated Infections

BACKGROUND: Targeting prevention efforts at healthcare facilities with a high preventable burden of healthcare-associated infections (HAI) should be a cost effective way to reduce morbidity, mortality, and avoidable healthcare costs. However, it is unclear how to best identify outlier facilities to maximize impact of prevention dollars. The standardized infection ratio (SIR) may not necessarily indicate a large number of HAIs, especially in facilities with low exposure (device days). We examined several measures that may help prioritize facilities to maximize public health impact of limited prevention resources.

METHODS: Central line-associated bloodstream infection (CLABSI) data reported to NHSN in 2011 were used to calculate facility-specific SIRs; 388 facilities were included for which the SIR>1 (regardless of significance testing). Facility-specific measures included: cumulative attributable difference (CAD) (excess number of HAI, [observed – expected]), etiologic fraction (EF) [CAD/observed], SIR, and device days. Six mathematical models were used to identify outlier facilities using a random sample of 100 facilities to simulate a typical “state” experience. Models include single or multiple measures, each with an arbitrary cut-point to define outlier facilities. The outlier facilities identified in each model were characterized by their median CAD, EF, SIR, device days and observed HAIs. Bootstrap simulation (1000 runs per model) was performed and the median measures were compared between models by Wilcoxon-Mann-Whitney test.

RESULTS: Relying only on significant test of SIR>1 (model-1), facilities identified as outliers had a median of 7.2 excess HAIs (CAD) and 14.5 observed HAIs; Model-2 containing only CAD measure resulted in identifying facilities with higher median CAD (10.8 excess HAIs), device days (9218) and observed HAIs (31) than other models (p<0.05). Facilities selected by the models with multiple measures (model 4-6) had either similar or lower median CAD than that obtained by model-1. When ranking facilities by the CAD only model, the excess number of HAIs (275.8) among the 10 highest ranked facilities accounted for 23% of the total excess number (1,211) of HAIs in these 388 facilities, and therefore, targeting prevention efforts at these 10 facilities could reduce excess HAIs by 23%.

CONCLUSIONS: The CAD measure may help prioritize prevention activities to yield the highest impact on overall HAI experience for a state or group. It also can serve as an indicator of public health impact in terms of excess HAI than expected at national, state and facility level. The same principles could be extended to other HAIs, such as catheter-associated urinary tract infections and surgical site infections.