Lessons learned from a method for pairing student questionnaires from successive applications using name codes.
Int Conf AIDS 2002 Jul 7-12; 14:(abstract no. C10699)
Leenen I, Givaudan M IMIFAP, Mexico City, Mexico
ISSUES. When questionnaires are used to evaluate HIV/AIDS education programs, guaranteeing the confidentiality of the respondent's answers is a prerequisite to obtaining valid results. For longitudinal studies, this poses the problem of how to correctly identify subjects through subsequent measures, without affecting confidentiality. One solution is to ask subjects to convert their name into a code by an identical procedure in each application. However, even a simple converting procedure may induce errors and jeopardize the linking process. DESCRIPTION. We present a method to link questionnaires filled out by the same set of students in different applications. It takes into account a subject's name code as well as additional variables that may help in establishing a correspondence between questionnaires. Further, the method allows the user to adjust the probability of different types of errors in the name code and for assigning a different weight to the respective additional variables. Finally, the method includes a parameter of how stringent the procedure is in establishing a link. The method is illustrated using the data available from an HIV/AIDS prevention project in Toluca, Mexico. In this project, self-applied questionnaires were administered to high school students at 4 regular intervals over a 15-month period. Matching questionnaires based on exact correspondence of the name code resulted in 30% matches. With the proposed method, 56% were reliably matched (with a high stringency).
ISSUES: Successful matching of confidential questionnaires requires use of multiple variables that can account for several types of error.
RECOMMENDATIONS: Longitudinal studies that depend on linking individual responses from one round to the other must follow a clear procedure to match questionnaires in order to adequately evaluate the intervention. To this end, additional variables and possible errors should be considered and managed appropriately
