The 131st Annual Meeting (November 15-19, 2003) of APHA

Abstract #63076

Tanya Henneman, PhD¹, Tom Belin, PhD¹, Deborah Koniak-Griffin, EdD, RNC, FAAN², Evelyn Gonzalez-Figueroa, PhD, MPH², and William Cumberland, PhD¹. (1) Department of Biostatistics, University of California, Los Angeles, School of Public Health, Los Angeles, CA 90095-1772, (310) 825-4875, thenn@ucla.edu, (2) School of Nursing, University of California, Los Angeles, Box 956919, Los Angeles, CA 90095-6919

In health studies with missing data an available-case analysis can result in biased parameter estimates, the reduction of information or the elimination of potentially relevant variables from the analysis. Various imputation techniques have been developed in order to impute missing values and allow the researcher to perform statistical analyses appropriate for complete-data. With data collected in a couple-focused HIV/AIDS prevention program for young parents, we explore the use of a multiple imputation technique (Little and Rubin 1987) to assess the effect of an experimental intervention with outcome measurements missing for some individuals. Forty-nine couples were randomized according to site of recruitment into either a control or experimental group. Data were collected on subjects in a baseline questionnaire and at three different post-intervention time points. Outcome measures of interest are AIDS knowledge, behavioral intention to use condoms, the number of unprotected vaginal sex episodes in the past three months, and the number of sexual partners. We assume a multivariate normal model to impute values for missing outcomes (Schafer 1997). To determine if any significant differences in outcomes exist between the two treatment groups, we will apply a multiple imputation inference technique to combine parameter and standard error estimates from analyses made on imputed complete data sets.

Learning Objectives:

• 1. Develop a plan for handling missing data
• 2. Articulate and apply the multiple imputation procedure
• 3. Design and implement a statistical strategy for analyzing longitudinal data collected on couples

Keywords: HIV Interventions, Community-Based Partnership

Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.