Consequences of discharges from intensive care at night

Lancet. 2000 Apr 1;355(9210):1138-42. doi: 10.1016/S0140-6736(00)02062-6.


Background: It is generally believed that pressure for beds on intensive-care units (ICUs) has increased in the UK. This study used discharge at night as a proxy measure to investigate pressure.

Methods: Night was defined in two ways: "out of office hours' from 2200 to 0659 h and "the early hours of the morning" from 0000 to 0459 h. The rate of discharge at night was compared for 21 295 adult admissions to 62 ICUs covering the period 1995-98 with 10806 admissions to 26 ICUs covering the period 1988-90. With data solely from 1995-98, the consequences of discharge at night and premature discharge were investigated.

Findings: Overall, 2269 (21.0%) admissions did not survive the ICU in 1988-90 compared with 4487 (21.1%) in 1995-98. Of ICU survivors, 2.7% were discharged at night (2200-0659 h) in 1988-90 compared with 6.0% in 1995-98. In 1995-98, night discharges (2200-0659 h) had a higher crude (odds ratio 1.46, 95% CI 1.18-1.80) and case-mix adjusted (1.33, 1.06-1.65) ultimate hospital mortality. Higher odds ratios were observed when the definition of night was 0000-0459 h. Premature discharge was commoner at night, 42.6% vs 5.0% and its importance was apparent when incorporated into the logistic-regression model (premature discharge 1.35, 1.10-1.65; night discharge 1.17, 0.92-1.49).

Interpretation: Night discharges from ICU are increasing in the UK. This practice is of concern because patients discharged at night fare significantly worse than those discharged during the day. Night discharges are more likely to be "premature" in the view of the clinicians involved. The implication of these results is that many hospitals have insufficient intensive-care beds. In deciding whether or not to invest more resources in intensive care we must, however, consider the cost-utility of this particular service compared with other ways that additional resources could be used.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Databases, Factual
  • Diagnosis-Related Groups
  • Hospital Mortality
  • Humans
  • Intensive Care Units / statistics & numerical data*
  • Intensive Care Units / supply & distribution
  • Length of Stay / statistics & numerical data
  • Logistic Models
  • Outcome Assessment, Health Care
  • Patient Discharge / statistics & numerical data*
  • Time Factors
  • United Kingdom / epidemiology