Quantifying Additional Healthcare Resource Consumption Associated with MRSA Infection

Abstract

 This study tried to quantify the additional healthcare resources (indicated by length of hospital stay and healthcare expenditure) consumed by MRSA infections. The study included patients who had been discharged from our hospital (a 380–bed tertiary hospital) between December 2012 and December 2014. The database for analysis involved the combination of 2 administrative datasets: JANIS infection surveillance data for all admissions, and data from a government survey on the post-implementation effects of the diagnosis procedure combination system. The data were analyzed through propensity score matching. Propensity scores were estimated using a logistic regression model in which the independent variable was a dichotomous MRSA infection variable (1: infected; 0: uninfected). Using the propensity score, 1:1 matching was performed between cases (MRSA infection) and controls (no MRSA infection). The data was also analyzed using another matching method that addressed time-dependent bias. The additional healthcare resources associated with MRSA infections were calculated from the differences in the mean quantities of resources consumed between cases and controls. The total of 24,538 patients in the study included 47 identified as MRSA-infected patients. Including time-dependent bias, infected patients were associated with an additional length of stay of 13.1 days (95% confidence intervals [CI] 3.7–22.4, p=0.008) and an additional incremental healthcare cost of 1.07 million yen (95% CI 0.317–1.822, p=0.007). Excluding time-dependent bias, the additional length of stay was 21.2 days (95% CI 11.7–30.8, p<0.001) and the additional healthcare cost was 1.61 million yen (95% CI 0.643–2.570, p=0.001). The additional healthcare resource consumption associated with MRSA infections was estimated using propensity score matching using 2 matching methods that differed according to whether or not time-dependent bias was included. The 2 methods produced different estimates, indicating that failure to address time-dependent bias may lead to overestimates of the additional healthcare resources consumed. These estimates have possible applications in evaluating the cost-effectiveness of infection control and prevention measures.