Effects of ICU Capacity Strain on Patient Outcomes and ICU Discharge Timing
Abstract & Commentary
By David J. Pierson, MD, Editor
SYNOPSIS: Data from 155 ICUs over an 8-year period showed that patients admitted on days with ICU capacity strain (high census, new admissions, and/or higher patient acuity) experienced higher in-hospital mortality. However, unit strain on the day of ICU discharge was associated with fewer hours in the ICU and a slightly increased likelihood of ICU readmission, but there was no adverse effect on any patient outcome examined.
SOURCES: Gabler NB, et al. Mortality among patients admitted to strained intensive care units. Am J Respir Crit Care Med 2013;188:800-806. Wagner J, et al. Outcomes among patients discharged from busy intensive care units. Ann Intern Med 2013;159:447-455.
Two studies by the same group of investigators, published simultaneously in different journals, used a large nationwide database (the Cerner Corporation's Project IMPACT) to examine the influence of ICU capacity strain (that is, ICU census, new admissions, and average patient acuity) on patient outcomes. Project IMPACT includes data from 155 ICUs in the United States, and is considered to be representative of critical care in this country: 73% of its units are in community hospitals and 58% are in urban centers. The data used for these retrospective cohort studies came from patients cared for from 2001-2008, a period when a closed model for daytime intensivist staffing was in effect in 7% of the participating ICUs.
Gabler and colleagues sought to determine whether transient increases in ICU strain, present on the day a patient is admitted to the unit, influence in-hospital mortality. Their study population consisted of 264,401 patients, of whom 36,465 (14%) died in the hospital. By several measures of how busy the unit was at the time, being admitted at a time of increased ICU capacity strain was associated with an increased likelihood of dying in the hospital. This association was greater in ICUs with a closed physician staffing model: odds ratio (OR) 1.07; 95% confidence interval, 1.01-1.11 for closed units, vs OR 1.01 (95% CI, 0.99-1.03) for open units (P = 0.02). These findings persisted despite several statistical procedures for reducing the effects of confounding variables.
To test the hypothesis that ICU capacity strain may result in patients being moved out of the ICU too soon, with adverse consequences, Wagner et al examined the same three aspects of ICU strain (ICU census, new admissions, and average acuity), assessed on the day of ICU discharge, in relation to ICU length of stay and post-ICU discharge outcomes. Their study population, from the same 2001-2008 Project IMPACT database used by Gabler et al, consisted of 200,730 patients discharged from the 155 ICUs to hospital floors. In these patients, they sought to determine associations between ICU capacity strain metrics and ICU length of stay, 72-hour ICU readmissions, subsequent in-hospital death, post-ICU discharge length of stay, and hospital discharge destination. Examination of the data included multiple adjustments for patient characteristics and severity of illness, as well as for whether they had received mechanical ventilation or vasoactive infusions during their ICU stay.
Of the 200,730 patients who survived their ICU stay and were discharged to a hospital floor or step-down unit, 3% were readmitted to the ICU within 72 hours, 4% died in the hospital, and 63% were discharged from the hospital to their homes. Increases in the three measures of ICU strain were associated with shorter preceding ICU length of stay (P < 0.001 for each), and increased likelihood of ICU readmission (all, P < 0.05). The magnitude of the former effect was numerically small but statistically significant: increases in all strain variables from the 5th to the 95th percentiles of the data distributions resulted in a mean decrease in expected ICU length of stay of 6.3 hours (CI, 5.3-7.3 hours). However, no strain variable was associated with increased odds of subsequent death, decreased likelihood of being discharged home, or longer total hospital length of stay.
The authors interpret their findings as indicating that although patients tend to be moved out of the ICU more quickly when the unit is strained by admissions, census, and/or acuity, and this slightly increases the likelihood of ICU readmission, there are no adverse effects on patient outcomes. One implication of this is that shortening ICU stays in the manner observed in this study — although by little in terms of hours per patient — could have an important effect nationwide in increasing critical care capacity as demands for this resource continue to increase.
The associations between ICU strain and patient outcomes in these studies are statistically strong but small in magnitude. However, as pointed out by Gabler et al in their discussion, "Small changes in mortality risk, when estimated precisely (i.e., with narrow confidence intervals) as in this study, may have a large cumulative impact and may be important to ICU patients and providers."
The fact that the increase in mortality observed during times of ICU capacity strain was greater for units with a "closed" model for physician staffing than for "open" units deserves comment. Granted, only 7% of the ICUs in the database were closed during the period covered by the study. However, the sample included more than 19,000 patients admitted to closed units. One might think that such units would be more likely to be situated in academic hospitals that manage sicker patients and have a higher proportion of emergently admitted patients, exacerbating the unit strain. However, there is no indication that this was the case, and Gabler et al performed multiple procedures to eliminate such confounders. The authors' explanation, which seems plausible to me, is that closed units may be less flexible than open units in terms of physician staffing at times of short-term surges in census, new admissions, and acuity. Closed units have been shown to produce better patient outcomes under static conditions, but it may be that it is easier to accommodate increased demands for physicians when things get busy in an open-unit environment. This finding deserves further examination.