A prospective comparative trial was conducted to determine the effect
of a physician's visual assessment of emergency patients on triage cat
egorization and ability at triage to predict admission. The setting wa
s a university, county, referral center and residency training site. P
articipants were a consecutive sample of emergency department patients
presenting between the times of 0700 and 2300 hours for 5 weeks. All
patients were assigned a triage category by an emergency nurse (RN) wh
o saw the patient and by an emergency physician (EP) who had the optio
n of performing a visual assessment. Triage categorization was compare
d for interobserver agreement (Kappa [kappa] statistic) and by ability
to predict admission (MacNemar's test). A total of 3,949 patients was
entered. The patients that physicians visually assessed were triaged
by nurses as more ill (P < .001) For triage categories visualized by t
he EP compared with RN categorization, interobserver agreement was 59.
8%, kappa = .21. For triage categories not visualized by EP compared w
ith RN categorization, interobserver agreement was 67.9%, kappa = .45
(P < .001). Sensitivity of EPs to predict admission is as follows: all
RN triage, 41.3; not seen by EP, 54.9; seen by EP, 69.3. Specificity
is as follows: all RN triage, 93.7; not seen by EP, 88.5, seen by EP,
83.9. When physician visual assessment was done, agreement between phy
sicians and nurses decreased by more than half. Physicians who include
d visual assessment in patient triage were less likely to agree with R
N categorization. A visual assessment by the physician improved the se
nsitivity for predicting admission with an only small cost in specific
ity. Copyright (C) 1997 by W.B. Saunders Company