We developed a statistical model that would identify and quantify the relative contributions of different factors hypothesized to impact the frequency of emergency center (EC) patients who leave without being seen (LWBS). We performed an analysis of the daily counts of patients that registered in our EC during a 21-month period who then LWBS. Candidate predictor variables included the number of patients seen, and the number admitted to the hospital, for each area of our EC, as well as the hours of faculty double coverage, and the day of the week. Univariate analyses were performed using standard methods. Multivariate analysis was performed using the general linear model. A backward selection procedure was used to eliminate statistically insignificant variables until all remaining independent variables had P-values ≤ .05. External validation and analysis of the stability of the estimated regression coefficients of the model were evaluated using bootstrap methods. Two-tailed tests and a type I error of 0.05 were used. During the period studied, 133,666 patients were registered in the EC and 9,894 (7.4%) left. Multivariate analysis identified six variables that were significantly associated with LWBS. The fitted model containing all six variables explained 52.8% of the variability observed in LWBS frequency. The most powerful predictor of LWBS was total number of patients cared for in the main ED. This accounted for 46.4% of the observed variation in LWBS. The total number of trauma and resuscitation patients, and the total number of observation unit admissions to the hospital were also associated with increased LWBS. More pediatric cases seen in the main ED, weekends, and additional faculty coverage were associated with fewer patients leaving. Efforts to decrease the LWBS rate will be most successful if they address the issue of main ED volume. (Am J Emerg Med 2000;18:767-772. Copyright © 2000 by W.B. Saunders Company)