ICU Patients With Low Severity of Illness: How Many?

Abstract & Commentary

Synopsis: In this regional sample, a large proportion of the patients admitted to ICUs had a low (< 0.3%) estimated probability of death and did not receive ICU-specific interventions.

Source: Rosenthal GE, et al. Arch Intern Med 1998; 158:1144-1151.

To examine variations in severity of illness among patients admitted to ICUs, Rosenthal and colleagues collected data from 38 ICUs in 28 hospitals in one metropolitan region for four years (1991-1995). Patients with a predicted risk of death of 1% or less were classified as having low severity of illness. Predicted risk of death was determined using a statistical model that included APACHE III Acute Physiology Score, ICU admission source, and ICU admission diagnosis. Data were abstracted from ICU medical records on standardized forms. Patients were excluded if they were in the ICU for less than four hours following surgery, had burn injuries (only 1 hospital provided this care), were admitted only for hemodialysis, or died within the first hour after admission. Patients with diagnoses managed in coronary care or cardiovascular ICUs (acute myocardial infarction, unstable angina, cardiac arrhythmias, open heart surgery) were also omitted since they would require ICU care independent of acute physiologic abnormalities.

The total sample of 104,487 patients had a mean age (± SD) of 61.9 ± 18.1 years, 52% were male, and overall mortality was 11.8%. Low-severity patients accounted for 20,451 (19.6%) of all admissions, including 23.6% of postoperative and 16.9% of nonoperative admissions. Alcohol and other drug overdoses accounted for 40.2% of nonoperative low-severity admissions, while laminectomy and carotid endarterectomy accounted for 52.3% of postoperative low-severity admissions. Overall mortality among low-severity patients was 0.3%. The most commonly used ICU-specific intervention in postoperative and nonoperative patients was intraarterial monitoring. Excluding intraarterial pressure monitoring, only 12.1% of all postoperative and 11.4% of all nonoperative lowseverity admissions received an ICUspecific intervention. Nonetheless, lowseverity admissions accounted for 11.1% of total ICU bed days. Substantial variability was observed across hospitals. For nonoperative admissions, rates of lowseverity illness ranged from 5.227.5% of all ICU admissions; for postoperative admissions, rates ranged from 9.468%. The mean rate of lowseverity admissions was lower in the five major teaching hospitals than in the 23 minor teaching and nonteaching hospitals in the study sample.


Online availability of medical data offers new possibilities for assessing how health care, including care in the ICU, is delivered. While it is well known that ICU care is costly, few studies have examined how ICU beds are used or whether practice varies among hospitals. Important findings in the present study are that ICU beds were often used for patients in whom the risk of death was low and for patients who did not receive any ICU specific intervention.

More than 62% of the patients with lowseverity illness who received ICUspecific interventions received only intraarterial monitoring—an intervention unlikely to provide important assessment data. No information was provided regarding staffing patterns in the study ICUs, so it was not possible to determine whether patients were admitted to ICUs in order to permit more vigilant observation by nursing personnel or to compensate for the lack of housestaff. Rosenthal et al noted that there were large variations in numbers of lowseverity patients admitted to teaching and nonteaching hospitals, but they did not note if there were variations among nonteaching facilities. This information would be helpful in eliciting factors that may have led to the observed differences.

As Rosenthal et al note, there are several methodological limitations in this study. Findings were based on medical record data rather than direct observation; low severity was defined using historical outcomes; and no information was recorded about organizational resources, such as the ability to provide intensive monitoring in nonICU settings. Nevertheless, findings of this study suggest the need to carefully assess ICU admission criteria in order to elicit whether tradition, rather than need, determines who gets admitted. Given the high costs of ICU care, these findings provide strong support for ongoing analysis of ICU admission policies with the goal of determining how best to use these expensive resources.