An Improved Model for Evaluating Change in Randomized Pretest, Posttest, Follow-Up Designs

Abstract

Randomized pretest, posttest, follow-up (RPPF) designs are often used for evaluating the effectiveness of an intervention. These designs typically address two primary research questions: (1) Do the treatment and control groups differ in the amount of change from pretest to posttest? and (2) Do the treatment and control groups differ in the amount of change from posttest to follow-up? This study presents a model for answering these questions and compares it to recently proposed models for analyzing RPPF designs due to Mun, von Eye, and White (2009) using Monte Carlo simulation. The proposed model provides increased power over previous models for evaluating group differences in RPPF designs.

Randomized clinical trial evaluating the efficacy of a brief intervention targeting anxiety sensitivity cognitive concerns.

“…A latent difference score approach, using growth curve modeling (Mara et al, 2012; Mun, von Eye, & White, 2009) was used to model the effects of the intervention on changes in ASI-3 levels from preintervention to postintervention as well as on changes from postintervention to Month 1 follow-up. This approach was used over more traditional methods such as analysis of variance for several reasons.…”

“…Models were identified by fixing the residual variances in change from pre- to postintervention (Change 1) and from postintervention to Month 1 follow-up (Change 2). Mara et al (2012) demonstrated improved modeling power when baseline performance was covaried out. Therefore, baseline ASI-3 levels were also included as a covariate.…”

Analyzing data from a pretest-posttest control group design

“…One limitation of our review is that we did not cover the Johnson-Neyman analysis, which is an alternative to ANCOVA when homogeneity of regression slopes fails (D'Alonzo, 2004). Another limitation is that we did not provide extensions to include follow-up designs (Mara et al, 2012;McArdle, 2009;Mun et al, 2009;Willoughby et al, 2007). In addition, we did not address multivariate analyses, structural equation modeling or hierarchical linear modeling.…”

Study protocol of a multi-center RCT testing a social-cognitive intervention to promote volunteering in older adults against an active control

“…For more refined analyses, latent change scores, modeling change in volunteering minutes from baseline to six weeks, three months, and six months is performed in Mplus 8 [69]. This latent change approach has several advantages over traditional variance analyses: It allows testing differences from baseline to all post-tests in a single model instead of multiple tests, has more power to detect treatment effects, is robust to non-normality, and provides information on individual variability and fit statistics [70, 71]. All models are statistically controlled for age, sex, marital status, education, subjective health, and number of chronic conditions.…”