By Eric Walter, MD, MSc

Pulmonary and Critical Care Medicine, Northwest Permanente and Kaiser Sunnyside Medical Center, Portland, OR

Dr. Walter reports no financial relationships relevant to this field of study.

SYNOPSIS: There is widespread variability between hospitals in rates of ICU admission. High ICU utilization hospitals were more likely to use invasive procedures and incurred higher costs than low ICU utilization hospitals with no difference in mortality.

SOURCE: Chang DW, Shapiro MF. Association between intensive care unit utilization during hospitalization and costs, use of invasive procedures, and mortality. JAMA Intern Med 2016;176:1492-1499.

ICU services comprise 13.4% of total hospital costs and more than 4% of national health expenditures. Yet, the decision as to which patients should be cared for in the ICU largely is subjective. Chang and Shapiro used administrative data to compare ICU use across 94 hospitals in Washington state and Maryland for diabetic ketoacidosis (DKA), pulmonary embolism (PE), upper gastrointestinal bleed (UGIB), and congestive heart failure (CHF). The primary outcomes were risk-adjusted mortality, use of invasive procedures, and hospital costs. Invasive procedures were defined as use of central venous catheters for any of the diagnoses, mechanical ventilation in DKA, thrombolytics in PE, and esophagogastroduodenoscopy in UGIB. Analyses were adjusted for both patient level and hospital level factors. Logistic regression models were used to predict ICU admission rates for each hospital during hospitalizations for each diagnosis. Hospitals also were dichotomized into higher (> 50th percentile for predicted ICU utilization rate) and lower ICU utilization (50th percentile and below) groups.

There was wide variability in rates of ICU admission for each diagnosis (16.3%-81.2% for DKA, 5.0%-42% for PE, 11.5%-51.2% for UGIB, and 3.9%48.8% for CHF). High ICU utilization was associated with increased use of invasive procedures in all four conditions. Increased ICU utilization was associated with higher hospital costs despite comparable lengths of stay. Severity of illness was lower among patients in high ICU utilization hospitals. Hospital mortality did not differ between hospitals with high and low ICU utilization. Correlations between ICU utilization rates for all four conditions were high.


The ICU is inherently a heterogeneous unit. In the same ICU, we may admit a patient with severe hypotension next to a patient with malignant hypertension. The unit admits a patient presenting with severe bleeding, followed by a patient suffering from portal venous thrombosis. This heterogeneity may explain why so many ICU trials have produced negative results. The variability in admission rates described by Chang and Shapiro highlight another layer of ICU heterogeneity.

In many ways, these results should not be surprising. The decision to admit someone to the ICU is subjective. It will be influenced not only by the patient’s severity of illness, but also hospital size, number of ICU beds, bed availability, nurse ratios, physician comfort, reimbursement, and more. Chang and Shapiro found that institutional factors appeared to influence the decision to admit to the ICU more so than patient level factors. High ICU utilization hospitals admitted patients with all four studied conditions frequently to the ICU despite a lower severity of illness. The influence of these institutional factors may be understandable in some circumstances. In this study, smaller hospitals were higher ICU utilizers. It may make sense to admit a patient with DKA to the ICU in a smaller hospital in which floor nurses may not have the time or expertise to manage an insulin drip. However, these results also suggest that many patients may not need ICU care, as there was no difference in mortality between hospitals with high and low utilization. The decision to admit patients to the ICU may expose patients to potential harms, since invasive procedures such as central lines, thrombolytics, intubation, and EGD were used more often by high ICU utilizer hospitals.

These results present significant implications for future studies evaluating ICU outcomes and costs. There may be too much variability from hospital to hospital to compare ICU costs or outcomes directly across hospitals and regions without accounting for both institutional and patient level factors. We simply cannot compare sepsis-related organ failure assessment scores and Charlson comorbidity indices. Future studies must try to better understand the institutional factors that affect the decision to admit a patient to the ICU. At the individual level, ICU physicians and directors must critically consider the reasons why or why not staff chooses ICU level of care for a patient. With a better understanding of the factors that affect the decision to admit to the ICU, clinicians can better determine when ICU level care truly is needed.