Impacts of Methicillin-resistant Staphylococcus aureus on Length-of-stay among Texas hospital inpatients Using ICD10CM Codes




O'Neill, Liam Ph.D.
Park, Saehwan


Journal Title

Journal ISSN

Volume Title



Objective: Patients who contract Methicillin-resistant Staphylococcus aureus (MRSA) during their hospital stay will have a significantly longer length of stay, resulting in increased costs and worse outcomes. In practice, however, it may be difficult to estimate the additional health care costs that can be attributed to a MRSA infection. Moreover, it is ethically impossible to conduct a randomized controlled trials (RCT) with a sufficiently large sample. The endogenous nature of hospital acquired infections (HAI) and limitations of ICD-9-CM administrative data have also posed challenges to researchers. The purpose of this study is to examine the effects of different categories of MRSA infections on length of stay among hospitalized patients in Texas, using ICD10CM data which have recently become available. Design: We used Texas Health Care Information Collection (THCIC) inpatient database for the fourth quarter in 2015. We only included hospitalized patients whose length of stay exceeded one day. The final sample included 654,074 discharges. The dependent variable was each patient’s length of stay (LOS). Explanatory variables included five different MRSA types (MRSA sepsis, MRSA pneumonia, MRSA unspecified sites, other MRSA infections, and MRSA colonization), recommended by ICD-10-CM guidelines. In order to properly assess MRSA effects, we controlled patients’ medical status using major diagnosis categories and Charlson comorbidity index. Other patient-level confounders were adjusted as well, including age, gender, race and ethnicity, admission sources, and admission types. Negative binomial models were used for our analysis. In additions, we used propensity-score matching (PSM) to reduce the bias due to confounders and to reasonably infer causality. Findings: The mean LOS among MRSA patients varied across different types of MRSA, ranging from 8.9 days (MRSA colonization) to 20.6 days (MRSA pneumonia), while average LOS among non-MRSA patients was 6.0 days. Our multivariate model indicated that MRSA infections significantly increased the length of stay and the effects by 52% up to 131% depending on MRSA types. Matched comparison revealed endogeneity, showing that the effect of MRSA infections reduced to 28%-77%, but were still significant, except MRSA colonization (p=0.119). The differences in days of LOS were 8.6 days for MRSA pneumonia (IRR=1.768; p Conclusions: While most types of MRSA increased LOS significantly, our study also confirmed endogenous relationship between length of stay and MRSA infection, potentially because developed MRSA may prolong hospitalization, which may itself increase the risk of exposure to other pathogens. This bi-directional relationship is likely to result in over-estimations, causing exaggerated benefit predictions. Nevertheless, our results verified that MRSA accounted for 4-8 days (32-77%) of unnecessary hospitalization. The use of ICD10CM data overcomes some of the limitations of previous studies, i.e., those based on ICD9CM, as it includes five sub-categories of MRSA infections. Implications: Economic analyses for interventions, programs, and structural investments which consider prevention of HAIs can use our results to estimate expected benefits of reduced MRSAs. Hospital managers and health care professionals can better manage various MRSAs by understanding different risks and impacts. A perennial difficulty in studying MRSA infections is the lack of accurate and reliable data. The use of ICD10CM combined with public-reporting of MRSA infections shows great promise toward improving patient safety.