Using quantitative and qualitative methods to evaluate survey item quality: A demonstration of practice leading to item clarity.

摘要

The purpose of this study was to propose and evaluate a procedure for revising an existing self-administered survey that is in need of item revision and/or scale reduction while maximizing validity and reliability. The procedure was demonstrated using the Client Evaluation of Self and Treatment (CEST; Joe, Broome, Rowan-Szal, & Simpson, 2002), a self-administered survey used in drug and alcohol treatment agencies. The procedure included confirmatory and exploratory factor analyses of a large dataset of completed CEST surveys, a readability analysis, and cognitive interviewing of two different groups of respondents to determine what problems they might have with CEST items. The cognitive interviewing revealed a number of issues that led to confusion among respondents, including items with two distinct concepts embedded, items containing absolutes and vague qualifiers, misinterpreted items, and terms and phrases respondents had difficulty understanding. The CEST was also judged to be long and potentially burdensome to respondents. Based on the results of this evaluation, a new survey---the Brief Assessment of Self in Context (BASIC)---also intended for use by substance abuse treatment providers, was constructed. First, factor analyses of the CEST and advice from an expert panel were used to determine which scales to retain. Next, quantitative analyses and cognitive interviewing helped determine which CEST items to retain and which to revise. Readability, sound item writing principles, and response format and scale requirements were also used to determine which items to include in the initial draft of the BASIC and guided item construction when needed. After the panel of experts provided feedback on the first revision, a final draft was prepared. Another round of cognitive interviewing was followed by administration of the final draft of the survey to a representative sample. The results indicated that the BASIC's items are clear, unambiguous, and easy to interact with and understand, and that the instrument is an improvement over the CEST. In brief, the procedure demonstrated in this study produced a psychometrically sound instrument composed of items that are easy for respondents to access.