Association between strained capacity and mortality among patients admitted to intensive care: A path-analysis modeling strategy

Purpose: To evaluate the associations between strained ICU capacity and patient outcomes. Methods: Multi-center population-based cohort study of nine integrated ICUs in Alberta, Canada. Path-analysis modeling was adopted to investigate direct and indirect associations between strain (available beds <= 1; occupancy >= 95%) and outcomes. Mixed-effects multivariate regression was used to measure the association between strain and acuity (APACHE II score), and both acuity and strain measures on ICU mortality and length of stay. Results: 12,265 admissions comprise the study cohort. Available beds <= 1 and occupancy >= 95% occurred for 22.3% and 17.0% of admissions. Lower bed availability was associated with higher APACHE II score (p < 0.0001). The direct effect of <= 1 available beds at ICU admission on ICU mortality was 11.6% (OR 1.116; 95% CI, 0.995-1.252). Integrating direct and indirect effects resulted in a 16.5% increased risk of ICU mortality (OR 1.165; 95% CI, 1.036-1.310), which exceeded the direct effect by 4.9%. Findings were similar with strain defined as occupancy >= 95%. Strain was associated with shorter ICU stay, primarily mediated by greater acuity. Conclusions: Strained capacity was associated with increased ICU mortality, partly mediated through greater illness acuity. Future work should consider both the direct and indirect relationships of strain on outcomes. (C) 2017 Elsevier Inc. All rights reserved.