Assessments of intensive care unit (ICU) performance currently rely on risk‐adjusted rates of in‐hospital mortality. These rates do not account for deaths occurring shortly after hospital discharge that may be attributable to the quality of care at the discharging hospital. Factors predicting early postdischarge mortality among ICU patients are unknown.
Thirty‐five California hospitals retrospectively collected data from 9751 adult ICU survivors with an ICU stay ≥ 4 hours. We excluded: (1) readmissions, (2) burn, trauma, and CABG patients, (3) unavailable National Death Index data, and (4) in‐hospital deaths < 30 days from the index ICU admission. Early postdischarge mortality was defined as a death occurring after hospital discharge and less than 30 days from the index ICU admission. We used multivariable discrete survival models to identify factors associated with early postdischarge mortality. We explored patient, hospital, and utilization factors. Patient factors included race, sex, insurance status, and severity of illness (Acute Physiology and Chronic Health Evaluation IV predicted mortality). Hospital factors included teaching status, ownership type, and number of beds. Utilization factors included discharge location and ICU LOS divided into quartiles. In addition, to explore how severity of illness modifies the effect of ICU LOS in predicting early postdischarge mortality, we introduced an interaction term between these 2 variables.
Eight thousand nine hundred and seventeen patients met our inclusion and exclusion criteria. The early postdischarge mortality rate was 4.0% (n = 355). Multivariable analyses demonstrated that among patient factors, early postdischarge mortality increased with greater severity of illness (highest quintile OR = 14.61, 95% Cl 8.01–26.63) and decreased with full‐code status (OR = 0.31, 95% Cl 0.21–0.47). Among hospital factors, nonteaching hospitals demonstrated higher odds of early postdischarge mortality (OR = 1.34,95% Cl 1.00–1.80). Among utilization factors, early postdischarge mortality increased for patients transferred to subacute facilities (OR 2.59, 95% Cl 1.88–3.56) compared to discharges home. Additionally, shortening the adjusted LOS increased the odds of early postdischarge mortality (OR 1.91, 95% Cl 1.39–2.61) compared to patients with the longest adjusted LOS. Finally, there was a statistically significant (P < 0.01) interaction between severity of illness and quartile of length of stay. Therefore, shortening the ICU LOS was more predictive of early postdischarge mortality in the most severely ill patients.
Multiple patient, hospital, and utilization factors are associated with early postdischarge mortality. Some factors, including shortening the length of stay and discharges to subacute facilities, may signal deficiencies in the quality of care, as well as deficiencies in publicly reported ICU performance measures currently based on in‐hospital mortality.
E. E. Vasilevskis, none; M. W. Kuzniewicz, none; M. L. Dean, none; T. Clay, none; D, J. Rennie, none; R. A. Dudley, none.
To cite this abstract:Vasilevskis E, Kuzniewicz M, Dean M, Clay T, Rennie D, Dudley R. Predictors of Early Postdischarge Mortality in Critically III Patients: Lessons for Quality Performance and Quality Assessment. Abstract published at Hospital Medicine 2009, May 14-17, Chicago, Ill. Abstract 99. Journal of Hospital Medicine. 2009; 4 (suppl 1). https://www.shmabstracts.com/abstract/predictors-of-early-postdischarge-mortality-in-critically-iii-patients-lessons-for-quality-performance-and-quality-assessment/. Accessed September 16, 2019.