Excess length of stay and mortality due to Clostridium difficile infection: a multi-state modelling approach.

BACKGROUND The burden of healthcare-associated infections, such as healthcare-acquired Clostridium difficile (HA-CDI), can be expressed in terms of additional length of stay (LOS) and mortality. However, previous estimates have varied widely. Although some have considered time of infection onset (time-dependent bias), none considered the impact of severity of HA-CDI; this was the primary aim of this study. METHODS The daily risk of in-hospital death or discharge was modelled using a Cox proportional hazards model, fitted to data on patients discharged in 2012 from a large English teaching hospital. We treated HA-CDI status as a time-dependent variable and adjusted for confounders. In addition, a multi-state model was developed to provide a clinically intuitive metric of delayed discharge associated with non-severe and severe HA-CDI respectively. FINDINGS Data comprised 157 (including 48 severe) HA-CDI cases among 42,618 patients. HA-CDI reduced the daily discharge rate by nearly one-quarter [hazard ratio (HR): 0.72; 95% confidence interval (CI): 0.61-0.84] and increased the in-hospital death rate by 75% compared with non-HA-CDI patients (HR: 1.75; 95% CI: 1.16-2.62). Whereas overall HA-CDI resulted in a mean excess LOS of about seven days (95% CI: 3.5-10.9), severe cases had an average excess LOS which was twice (∼11.6 days; 95% CI: 3.6-19.6) that of the non-severe cases (about five days; 95% CI: 1.1-9.5). CONCLUSION HA-CDI contributes to patients' expected LOS and risk of mortality. However, when quantifying the health and economic burden of hospital-onset of HA-CDI, the heterogeneity in the impact of HA-CDI should be accounted for.