Prolonged intensive care unit stay: predictors and impact on resource utilization

Objective

To evaluate the predictors of prolonged ICU stay its impact on resource utilization.

Methods

We collected the following data prospectively on all ICU admissions between 2/1999 and 9/2000: demographics, ICU stay, APACHE II and SAPS II scores, the main reason for ICU admission, re-admissions and ICU outcome. We documented the presence of coma, oliguria, coagulopathy and infection as well as the need for mechanical ventilation or vasopressors in the first 24 hours. ICU stay was considered prolonged if it exceeded 14 days. We calculated the utilization of ICU days and ventilator days. We identified predictors of prolonged stay using univariate analysis.

Results

Eleven percent of patients (104/947) stayed in ICU > 14 days. This group of patients used 45% (2880/6392) of ICU days and 56% (2556/4604) of ventilation days. Prolonged stay patients had higher APACHE II (21 ± 8 vs 19 ± 9, P = 0.016) and SAPS II scores (43 ± 16 vs 37 ± 20, P = 0.003), although ICU mortality was not different (19% vs 21%). Patients were more likely to have prolonged stay if the main reason for admission was respiratory (OR 2.2, CI 1.4–3.6) or trauma (OR 2.1, CI 1.4–3.4) and less likely if it was non-trauma surgical (OR 0.27, CI 0.13–0.54). Prolonged stay occurred more likely with re-admissions (OR 2.1, CI 1.1–3.8) and in patients with oliguria (OR 1.8, CI 1.1–3.1), coagulopathy (OR 1.5, CI 1.01–2.3), infection (OR 2.3, CI 1.5–3.5), mechanical ventilation (OR 1.9, CI 1.3–2.9) and vasopressor therapy (OR 1.8, CI 1.2–2.7).

Conclusions

Although patients with prolonged stay constitute a small fraction of ICU patients, they consume a significant proportion of ICU resources. Patients admitted for respiratory or trauma indications are more likely to have prolonged stay. Attempts to shorten ICU stay, such as by development of clinical pathways, should especially target these patients. Caring after some of these patients in a step-down unit may have a great impact on resource utilization.