Main Menu Browse by Day or Program Author Keywords

545 Multistate Modelling to Estimate the Excess Length of Stay Associated with MRSA Infection in Surgical Patients

Saturday, March 20, 2010
Grand Hall (Hyatt Regency Atlanta)
Giulia De Angelis, MD , Geneva University Hospitals, CH-1211 Geneva 14, Switzerland
Arthur Allignol, MSc , Freiburg centre for Data Analysis and Modelling, University of Freiburg, Freiburg, Germany
Anant Murthy, MSc , Johns Hopkins Bloomberg School of Public Health, Baltimore, MD
Martin Wolkewitz, PhD , Freiburg centre for Data Analysis and Modelling, University of Freiburg, Freiburg, Germany
Jan Beyersmann, PhD , Freiburg centre for Data Analysis and Modelling, University of Freiburg, Freiburg, Germany
Edith Safran, MD , Geneva University Hospitals, CH-1211 Geneva 14, Switzerland
Jacques Schrenzel, MD , Geneva University Hospitals, CH-1211 Geneva 14, Switzerland
Didier Pittet, MD, MS , Geneva University Hospitals, CH-1211 Geneva 14, Switzerland
Stephan Harbarth, MD, MS , Geneva University Hospitals, CH-1211 Geneva 14, Switzerland
Background: Current evidence on the excess length of stay (LoS) of nosocomial methicillin-resistant Staphylococcus aureus (MRSA) infections suffers from methodological limitations. Approaches which do not accurately take the timing of time-varying exposure, like nosocomial infection, into account, may overestimate the associated excess LoS. Multi-state model may help to avoid this time-dependent bias.

Objective: This study employed a multi-state model to estimate the excess LoS associated with nosocomial MRSA infection in a large cohort of surgical patients, using two different comparator groups (MRSA carriers and MRSA-free patients); and the excess LoS due to MRSA colonisation compared to MRSA-free patients.

Methods: We retrospectively analysed data from a prospective cohort study previously conducted at the University of Geneva Hospitals, Switzerland. 964 MRSA carriers, 167 (17%) of whom were infected, and 13’650 MRSA-negative control patients were included in two multistate models, considering either the detection of nosocomial MRSA infection or MRSA colonisation, as time-dependent exposures, while discharge or death was the study endpoint. The time point of any event referred to the time of admission to hospital, which was counted as day 1. Furthermore, the time point of MRSA colonisation after hospital admission was included as delayed study entry of MRSA carriers. The excess LoS was extracted computing the Aalen-Johansen estimator of the matrix of transition probabilities. 95% confidence intervals (95% CI) were derived by bootstrap re-sampling. Multivariate Cox regression analysis was used to assess the effect of MRSA infection on excess LoS.

Results: The median LoS of MRSA-infected, colonised and negative patients was 48, 13 and 6 days, respectively. After multistate modelling, the adjusted excess LoS of MRSA-infected compared to uninfected MRSA carriers was 5.9 (95% CI, 0.1-11.7) days, with deep surgical site infections causing the largest excess LoS (14 days, 95% CI, 3.2-24.8). Compared to MRSA-free patients, the excess LoS due to MRSA infection reached 16 days (95% CI, 9.4-22.7). The estimated effect of MRSA infection was most prominent on LoS in acute care wards, where MRSA-infected patients were estimated to stay 7.3 days (95% CI, 2.0-13.0) longer than patients only colonised by MRSA and 17 days (95% CI, 11.0-22.9) longer than MRSA-negative patients. Cox regression analysis confirmed that nosocomial MRSA infection decreased the discharge hazard of MRSA carriers (adjusted hazard ratio, 0.79; 95% CI, 0.65-0.95). Excess LoS due to MRSA colonisation was 3.8 (95% CI, 1.6-6.0) days.

Conclusions: MRSA represents an important health-economic burden, by significantly prolonging LoS, particularly in acute care wards. Multistate modelling is a promising statistical method to avoid time-dependent bias, providing accurate estimation of extra-LoS due to nosocomial infections.