Objective: To develop an effective measure of the symptoms of eating disorders for use in community surveys. Method: A branched eating disorders instrument for administration by notebook computer was evaluated as a screen in 487 teenage schoolgiris. High scorers and a stratified sample of low scorers were evaluated in a second-stage interview with the Eating Disorders Examination (EDE). Results: The Branched Eating Disorders Test (BET) proved to have exceptional validity coefficients (sensitivity 1 OO%, specificity 99%, positive predictive value 70%). Discussion: This pilot validation suggests that the BET largely overcomes the problems of earlier instruments of limited screening utility when applied to community samples. 0 7 996 by John Wiley & Sons, Inc.
Over the last two decades the development of clinically validated instruments for the evaluation of symptoms of eating disorders has allowed an expansion of epidemiological research into community settings (Garner & Garfinkel, 1979; Garner, Olmstead, & Polivy, 1983; Whitaker et al., 1989). Although the application of clinically validated instruments in community studies of eating disorders has improved understanding of the disorders in the nonclinical setting, efficient case detection has continued to be a challenge to psychiatric epidemiologists. The relatively low prevalence of eating disorders in community samples substantially reduces the diagnostic and screening efficacy of instruments effective in clinical populations (Williams, Hand, & Tarnopolsky, 1982). Furthermore, the form of eating disorders in the community differs from that presenting in clinical settings where instruments are usually validated. Subclinical disorders are more common in the