1058

Abstract

Introduction: Because hospital mortality rates of sepsis are widely reported, individual ICUs are tasked to reduce the mortality rate of patients with sepsis. However, factors outside the control of intensivists may undermine efforts to achieve mortality reductions. Methods: This is a retrospective single-center observational study in an academic MICU from January 2010 to June 2012 during refinement of sepsis bundles. We examined outcomes of MICU patients with ICD-9 codes for severe sepsis/septic shock and ICD-9 codes for infection and acute organ failure. Results: Per 6-month block, over the 30 months of the study, the number of cases of sepsis increased by 35% (327 to 469) and cases of infection with organ failure increased 28% (390 to 498). Mean age, gender and the likelihood of undergoing mechanical ventilation for both groups was unchanged. For patients coded with sepsis, mean ICU and hospital LOS decreased from 6.1 to 4.6 days (p<0.01) and 13.2 to 11.1 days (p< 0.01) respectively. For patients coded for infection and organ failure, ICU and hospital LOS decreased from 4.5 to 4 days (p<0.01) and 13.5 to 11.8 days (p<0.01). Mortality rates for sepsis decreased non-significantly from 37% to 31.6% (p 0.1) and for sepsis and organ failure from 14.9% to 11.3% (p 0.6). The observed to expected mortality ratio (O/E) decreased for sepsis (1.58 to 1.29) and for infection with organ failure (1.44 to 0.85). Conclusions: Because of relatively small numbers of sepsis cases seen in a single ICU, coding variability and changes in severity of illness, unadjusted mortality is an inadequate sole measurement of the clinical improvement achieved from adherence to sepsis bundles. While crude mortality rates did not statistically change, we found significant decreases in ICU and hospital LOS and substantial decreases in the O/E mortality for cases of sepsis and infection with organ failure. We confirmed previous findings that reliance on ICD-9 codes for sepsis may underestimate the true incidence of sepsis in the ICU.