256703 Using Path Analysis To Test A Hypothesis On The Theory Of Change In Hemoglobin A1C (HbA1C) Among Clients In A Culturally Tailored Diabetes Intervention For African Americans And Latinos

Tuesday, October 30, 2012 : 12:50 PM - 1:10 PM

Brandy Sinco, MS , School of Social Work, University of Michigan, Ann Arbor, MI
Michael Spencer, PhD , School of Social Work, University of Michigan, Ann Arbor, MI
Nicolaus Espitia, MSW , School of Social Work, University of Michigan, Ann Arbor, MI
Edith Kieffer, PhD , School of Social Work, University of Michigan, Ann Arbor, MI
Gloria Palmisano, BS, MA , REACH Detroit Partnership, Detroit, MI
Melissa A. Valerio, PhD, MPH , Hbhe, University of Michigan, Ann Arbor, MI
Phillip Chapman, PhD , Statistics, Colorado State University, Fort Collins, CO
Objective. To use path analysis to test a hypothesis about the theory of change in Hemoglobin A1c among clients in a culturally tailored diabetes intervention. The hypothesis was that changes in self-efficacy, diabetes-related distress, and knowledge would lead to change in self-management behavior, which would lead to reduction in HbA1c. Methods. Two cohorts were combined for a sample size of 326. Path analysis was chosen because it enables the analyst to test the order in which measurable variables affect each other. Mardia's multivariate kurtosis, along with univariate skewness and kurtosis, were used to check for multivariate normality. MAR (Missing at Random) was evaluated by testing if pre-intervention means differed significantly by whether post-intervention values were present. The structural equation model was estimated by FIML (Full Information Maximum Likelihood). Goodness of fit was evaluated with Joreskorg-Sorbom GFI for absolute fit, Bentler's CFI for comparative fit, RMSEA for parsimony, and SRMR for prediction. Results. All post-intervention measures were strongly correlated with pre-intervention values. Post-intervention HbA1c dropped by 0.5 per unit increase in self-management behavior. Program attendance in group, versus individual format, was associated with a 6 point drop in diabetes distress and significant increase in knowledge of diabetes management. Greater self-efficacy was associated with higher attendance. Based on GFI=.9928 and CFI=.935, the model explained 99.25% of the generalized covariance and was a 93.5% improvement over the null model. Conclusion. Path Analysis is an effective method to model the process of change in key outcome variables and to measure quantitative intervention effects.

Learning Areas:
Social and behavioral sciences
Systems thinking models (conceptual and theoretical models), applications related to public health

Learning Objectives:
1)_Explain how path analysis can be used to test a hypothesis on the process of change in key variables in a health intervention. 2)_Describe procedures for testing the assumptions of multivariate normality and MAR (Missing at Random) for a structural equation model. 3)_Define degree of freedom and power calculations for structural equation models. 4)_Discuss the goodness of fit indices and graphical techniques for evaluating the goodness of fit for a structural equation model.

Keywords: Statistics, Diabetes

Presenting author's disclosure statement:

Qualified on the content I am responsible for because: I have masters degrees in mathematics and statisics, and have been involved with analysis of the REACH project data for several years. I have been a co-author on many publications involving statistical analysis.
Any relevant financial relationships? No

I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.

Back to: 4248.0: Student Award Presentation