Does a stepped approach using mixed-mode data collection reduce attrition problems in a longitudinal mental health study?

Abstract

According to epidemiological standards for longitudinal studies, any appropriate attempt should be made to reach a high response rate. This also concerns re-contacting respondents in follow-up waves. In the Netherlands Study of Depression and Anxiety (NESDA), respondents, who refused to visit the study site for a follow-up interview and would have been lost in a single step approach, were offered an interview at their home (in a second step) or an interview by telephone (in a third step). The stepped approach intended to increase response rates, but as a by-product, the switching of the interview response setting introduced effects due to the three different modes for obtaining data. This study aimed to evaluate the benefits and detriments of the introduction of a stepped approach using mixed-mode data collection. The attained attrition in the stepped approach was compared to the attrition rate that would have been achieved if the design had remained in a single mode. Logistic regression models were used to study if the attrition was related to patient characteristics, i.e. the attrition was selective. Propensity score matching was used to study if similar respondents reacted differently to different modes. The stepped approach using mixed-mode collection reduced the attrition from 22% to 13%, while the selectivity of the attrition was reduced on almost all socio-demographic variables, although the selectivity increased in variables that reflect the respondents’ mental health. Propensity score matching demonstrated no evidence of mode effects. Although the introduction of the stepped approach using mixed mode data collection did not completely solve the problem of attrition, we conclude that adding different modes was worthwhile.