Background The provision of intensive care is a perplexing issue for c
linicians and the public. Concerns about the apparent lack of beds and
the appropriateness of the patients admitted are tempered by the high
cost of providing this service. As part of a study commissioned by th
e UK Department of Health, we tested the hypothesis that there is exce
ss mortality among patients who are refused admission to intensive-car
e units. Methods All referrals to six intensive-care units with differ
ent numbers of beds were monitored during a 3-month period. Data on mo
rtality 90 days after first referral were obtained from family physici
ans for all patients known to be alive at hospital discharge. We adjus
ted, where possible, for confounding, including for age, sex, appropri
ateness of referral, disease severity, surgery and emergency categorie
s, and bed provision. We did multivariate analysis by multiple logisti
c regression to compare the adjusted 90-day mortality rates for patien
ts who were refused admission and for those admitted. Findings 480 pat
ients were admitted and 165 were refused admission. 90 days after refe
rral there had been 178 (37%) deaths among the admitted group and 75 (
46%) among the refused group. After multivariate adjustment, 113 patie
nts appropriately referred for intensive care but refused admission to
their first-choice intensive-care unit had a relative risk of death o
f 1.6 (95% CI 10-2.5), compared with the group of appropriately admitt
ed cases with medium APACHE II scores for disease severity. Age, the a
ssessed need for treatment or monitoring interventions, and emergency
status also contributed to differences in mortality among all referral
s. Bed provision did not contribute significantly to excess mortality.
Interpretation Although this study is observational and case-mix adju
stment is incomplete, we found a higher rate of attributable mortality
in patients who were refused intensive care, particularly for emergen
cy cases. We question whether the provision of more beds alone would b
e a solution and conclude that there is an urgent need for more approp
riate admission and discharge criteria.