Maximum Individualized Change Analysis

University Of South Florida, Tampa FL

Investigators

Abstract

DESCRIPTION (provided by applicant): Project Summary Problems frequently encountered social and behavioral science researchers concern heterogeneity of the: 1) individuals needing treatment, 2) interventions responding to individuals'specific needs, and 3) array of potential outcomes resulting from treatment (Goldenberg, 1978;Gordon, Powell, Rockwood, 1999;Rockwood, Joyce, &Stolee, 1997). A common response to these problems is to identify various outcome measures that encompass the variability in patients, treatments, and outcomes (Kazdin, 2005). In these studies, the patients treated are often assessed on all outcomes. Analyses often involve multivariate approaches or latent class models. A problem with this design and analysis strategy is that when patients are assessed on all potential measures, many are assessed on outcomes that their specific treatment plan was never designed to address. The project is designed to expand the research base on analytic methods to assess the statistical significance associated with interventions targeted at heterogeneous populations. Specifically, the project will continue the study of the maximum individualized change (MIC) procedure developed by Boothroyd, Banks, Evans, Greenbaum, and Brown (2004). The method was initially developed as an analytic alternative to the more traditional multivariate and latent model approaches often used in studies examining individually-tailored interventions. Our initial developmental work on this approach suggests that it offers a number of significant advantages over traditional statistical approaches in studies where a number of measures are used to assess potential treatment outcomes. These advantages include increased statistical power to detect smaller program effects and few problems associated with missing data. Despite the MIC's early promise, more basic work is necessary to determine whether this approach should become a standard analytic tool in future studies. The study will consist of two phases. First, an expanded simulation study of the MIC procedure will be conducted (Boothroyd et al. 2004) broadening the initial set of assumptions. Specifically, the impact of seven variables will be systematically examined. The variables include varying the: 1) level of correlation among outcome measures, 2) number of outcome measures/domains being assessed, 3) distribution of response effects, 4) number of outcome measures on which a person improves, 5) sample size, 6) weighting the outcome measures based on qualitative assessment regarding the likelihood of change, and 7) amount of missing data. In phase two, the MIC procedure will be used in the re-analysis of data collected as part of SAMHSA-funded multi-site national study examining the impact of managed care. In a recently published article summarizing the study findings, the authors concluded, "Managed care and fee-for-service Medicaid programs did not differ on most measures;however, a lack of sufficient power was evident for many measures" (Leff, et al., 2005).