Objectives: To determine the accuracy of predictions
of the need for hospital admission and, if
admitted, length of stay (LOS) made early in an
emergency attendance by emergency department
(ED) doctors, nurses, patients and relatives, and
the characteristics of ED presentations predictive
of admission and short stays (~3 days).
Methods: Prospective collection of predictions by
medical and nursing staff, patients and relatives of
ED departure status and LOS (1 day, 2-3 days, 4-
7 days or longer) of a convenience sample of
adults presenting with medical symptoms. Predictions
were made before full medical assessment
and matched against actual departure status and
LOS. Vital signs and demographics were
recorded.
Results: Seventy five percent (2159/2904; CI
73%?77%) of all admission predictions in 704
patients were correct with 85% (575/673; CI 81%-
88%) of doctors? predictions correct. Thirty-five
percent (361/1024) of all LOS predictions for 331
patients were correct with 46% (122/268; CI 40%-
52%) of doctors? predictions correct. Risk factors
for short-stay over longer admission included age
less than 65, normal oxygen saturations and selfreferral.
Conclusion: Emergency admissions can be predicted
with reasonable accuracy but LOS is difficult
to predict. Development of a prediction tool
may facilitate streaming and appropriate use of
short-stay units.