Journal of Clinical Epidemiology 56 (2003) 238–247 0895-4356/03/$ – see front matter © 2003 Elsevier Science Inc. All rights reserved. doi:10.1016/S0895-4356(02)00613-3 Asthma-like symptoms assessment through ECRHS screening questionnaire scoring Mario Grassi*, Cristiana Rezzani, Ginevra Biino, Alessandra Marinoni Università Deglo Studi Di Pavia, Dipartimento di Scienze Sanitaire, Applicate E Piscocomportamentali, via Bassi, 21, 27100 Pavia, Italy Received 20 June 2001; received in revised form 15 November 2002; accepted 20 November 2002 Abstract The definition of asthma has always been a matter of discussion. The European Community Respiratory Health Survey (ECRHS) is a multina- tional survey designed to compare the prevalence of asthma in subjects aged 20 to 44 years throughout European countries. In each center a repre- sentative sample completed a self-administered screening questionnaire composed of nine dichotomous items. We propose a method for construct- ing a score for the ECRHS screening questionnaire and we validate the method with reference to the clinical diagnosis available for the Italian centers. Clinical diagnosis was made by a group of medical experts after examining the responses to a standardized clinical interview, respiratory function tests, and allergy tests. Before constructing a summary score, the number of latent factors/dimensions explaining correlations among the ob- served items was recognized. We identified only one factor/dimension underlying the screening questionnaire, so a summary score was determined by the Homogeneity Analysis by Alternating Least Square (HOMALS). Using best Receiver Operating Characteristic (ROC) threshold the diagnos- tic test of the score had sensitivity and specificity of 75.1 and 80.1%, respectively. The method of scoring is easily reproducible, and has the advan- tage of optimizing information recoded by the questionnaire. © 2003 Elsevier Science Inc. All rights reserved. Keywords: Asthma; Screening questionnaire; Unidimensionality; Score; ECRHS 1. Introduction The main reason for conducting standardized interna- tional prevalence studies of asthma is the current concern that the prevalence of the disease may be increasing [1–7]. Although methodologic differences in these studies make it difficult to compare the magnitude of discrepancies in asthma prevalence between countries, the trend of increas- ing prevalence in countries of widely differing lifestyles and ethnic groups is generally consistent. The principal problem in assessing asthma prevalence is that the definition and classification of asthma remains con- troversial. In fact, the diagnosis of asthma involves an overall assessment of the patient’s medical history, physical exami- nation, and laboratory test results, and there are no standard- ized methods for combining the information from these vari- ous sources. Usually, two methods have been used in the past to measure asthma prevalence: questionnaires and physio- logic measurements. Standard written questionnaires have been the principal instrument for measuring asthma symptom prevalence in community surveys, and in homogeneous pop- ulations these have been standardized, validated, and shown to be reproducible [8–10]. As an example, in the European Community Respiratory Health Survey (ECRHS) stage I a mailed questionnaire was administered. The ECRHS is a multinational survey designed to assess and compare the prevalence of asthma in 20- to 44-year-olds, in several Euro- pean areas [11]. The design of ECRHS involved two stages. In the first stage a short questionnaire of nine items, called the ECRHS screening questionnaire, was mailed to all the study participants. In the second stage, a 20% random sample of re- sponders to the mailed questionnaire and an additional sam- ple of symptomatic individuals were invited to the local chest clinic to undergo a standardized clinical interview, called the long questionnaire, allergy tests, and bronchial challenge test with methacholine. Many studies have been conducted on the ECRHS screening questionnaire [8,12,13], but almost all of them consider the prevalence or sensitivity and specificity of symptoms corresponding to individual items without con- sidering the questionnaire as a whole, and thus without computing a score. The construction of a total score allows combining information contained in the questionnaire. Nev- ertheless, a total score is possible only if the questionnaire is unidimensional, that is, if the associations between the ob- served variables are due to the presence of an unobserved * Corresponding author. Tel.: +39-382-507537; fax: +39-382-507570. E-mail address: mariog@unipv.it (M. Grassi).