The study objective was to evaluate the capability and the consistency of the triage nurse to categorize correctly emergency patients and its impact on the waiting time for physician examination over a period of 3 years. The study was performed at the emergency department of the Barzilai Medical Center, Ashkelon, Israel. A retrospective review of the medical records was performed. All patients who were examined by a triage nurse during 2 randomly chosen consecutive weeks during the years 1995 and 1998 participated. All the medical records were reviewed by the authors and the following information was extracted from the medical records: nurse triage category, time of initial evaluation by a triage nurse, duration of employment of the nurse in the ED, and her experience as a triage nurse, time of initial examination by a physician, the total length of stay in the ED, the history taken by the triage nurse and the physician, and the physician's urgency category. Patient in urgency category 1 is a patient whose condition may deteriorate if not examined within 1 hour; patient in category 2 is a patient whose condition may deteriorate if not examined within 2 hours; category 3 is all the rest. Any deterioration and or delay of treatment of the patients were also recorded. Data concerning patients with an initial complaint of chest pain were extracted separately. The data were analyzed using the SPSS software and the results were tested by the student t test and chi square test. Interobserver agreement was measured using the kappa value. A total of 2,886 completely full medical records were reviewed by the authors: 1,310 records from period I (1995) and 1576 from period II (1998). Of the patients 92% and 88.2% were classified by the triage nurse as category 3 in periods I and II respectively, 7% and 9.8% as category 2, and 1% and 2% as category 1 respectively. Full agreement of triage category between nurse and physician was found in 90.5% of the cases in period I and 93% in period II (kappa = 0.90 and kappa = 0.93 respectively). In period I, 70% of the patients in category 1 were examined by a physician in 1 hour versus 100% in period II. Almost all the patients in category 2 were examined within 2 hours (98%, 97%), and 98% of those in category 3 were examined within 3 hours. The average waiting time for physician examination in category 1 patients dropped from 43.1 minutes in period I to 18.2 minutes in period II. The average waiting time for the triage nurse was 9 minutes in period I, and 7.42 minutes in period II. The average length of stay in the ED in period I was 1 hour and 24 minutes and 1 hour and 30 minutes in period II. Of the anamneses taken by the triage nurse 91.8% were fully identical with the physicians' anamneses, but in period II this percentage jumped to 98%. Patients with chest pain were categorized correctly by the triage nurse in 76.8% of the cases in period I and 72.4% in period II, with an overtriage of 18.6% and 20.7% respectively (kappa = 0.75, kappa = 0.70 respectively). In our study, nurse triage was safe and effective in classifying patients to urgency categories. The results are consistent and even improved over a 3-year period. The rates of incorrect classification, deterioration, and delay of treatment of patients because of incorrect triage are very low. Most of the patients were examined by the physician within the expected time. Triage nurse predicted correctly the urgency category of patients with chest in most of the cases and the rate of missing acute coronary events was very low.