Medical Intensive Care Unit Admitting Patterns in a National Cohort

Meeting: Hospital Medicine 2012, April 1-4, San Diego, Calif.

Abstract number: 97651

Background:

Critical care makes up nearly 1% of the US gross domestic product, but there is wide hospital variation in critical care resource use. It is possible that some hospitals have higher ICU admission rates because of patient factors beyond their control such as admitting diagnosis. Therefore, we sought to describe the association between hospital ICU admitting patterns and admitting diagnosis.

Methods:

We created a retrospective cohort of the first non–surgical admission of patients admitted from the Emergency Department or Outpatient Clinic to 120 Veterans Affairs (VA) acute care hospitals from July 2009 to June 2010. For each admission, we defined severity of illness as 30–day predicted mortality on admission (estimated using the validated VA–ICU severity score). To assess the diagnosis–specific association between ICU admission rate and severity across hospitals, we constructed separate multilevel models (random intercept and slope) for each of the nine most common admitting diagnoses (Table). We did not adjust for any covariates beyond severity and admitting diagnosis.

Results:

The 278,335 patients in our cohort had a median 30–day predicted mortality of 1.5%. The ICU admission rate for a median severity patient varied from 3.8% for pneumonia to 46.5% for acute myocardial infarction (Table). The odds ratio for the change in ICU admission rate for a one standard deviation change in severity of illness varied from 1.08 for chest pain to 2.39 for gastrointestinal hemorrhage. Compared to non–cardiac diagnoses (i.e., sepsis, chronic obstructive pulmonary disease, gastrointestinal hemorrhage, and pneumonia), ICU admission rates for cardiac diagnoses were not as strongly associated with changes in predicted mortality (Figure).

Conclusions:

Our results quantify the disparate way in which the ICU is used across diagnoses, a finding that will not surprise clinicians working in the hospital. However, it emphasizes that given current practice, measures of ICU utilization that aggregate across diagnoses will not work well as global measures of efficiency, even as it raises the larger question of whether these prevalent patterns represent the best use of an expensive resource.

Table 1Medical ICU Admitting Patterns by Diagnosis

  ICU admission rate for a median severity patient Odds ratio for change in ICU admission rate with a one standard deviation change in severity
Diagnosis Estimate (95% CI) Estimate (95% CI)
AMI 46.5% (40.5–52.6) 1.13 (1.02–1.25)
Dysrhythmia 21.6% (18.0–25.7) 1.32 (1.22–1.43)
GI Bleed 19.1% (16.4–22.0) 2.39 (2.12–2.70)
CAD 13.6% (10.9–16.7) 1.28 (1.12–1.47)
Sepsis 10.0% (8.8–11.2) 1.93 (1.85–2.01)
CHF 6.8% (5.5–8.4) 1.34 (1.21–1.49)
COPD 5.7% (4.7–6.8) 2.01 (1.76–2.30)
Chest pain 4.5% (3.2–6.4) 1.08 (0.95–1.24)
Pneumonia 3.8% (3.2–4.6) 1.93 (1.73–2.16)
Abbreviations: ICU is intensive care unit; CI is confidence interval; AMI is acute myocardial infarction; GI is gastrointestinal; CAD is coronary artery disease; CHF is congestive heart failure; COPD is chronic obstructive pulmonary disease; and PNA is pneumonia. Note: The median severity corresponds to a 30–day predicted mortality of 1.5% (calculated on admission).

Figure 1ICU Admitting Patterns across Severity of Illness: Cardiac (top panel) vs. Non–Cardiac (bottom panel) Diagnoses.

To cite this abstract:

Sales A, Kennedy E, Chen L, Hofer T. Medical Intensive Care Unit Admitting Patterns in a National Cohort. Abstract published at Hospital Medicine 2012, April 1-4, San Diego, Calif. Abstract 97651. Journal of Hospital Medicine. 2012; 7 (suppl 2). https://www.shmabstracts.com/abstract/medical-intensive-care-unit-admitting-patterns-in-a-national-cohort/. Accessed November 18, 2019.

« Back to Hospital Medicine 2012, April 1-4, San Diego, Calif.