3032 — Using Latent Class Analysis to Further Understand the Impact of Multiple Risk Factors in Primary Care

Funderburk JS (Center for Integrated Healthcare/Syracuse VAMC), Maisto SA (Center for Integrated Healthcare/Syracuse VAMC), Sugarman DE (Center for Integrated Healthcare/Syracuse VAMC), Krenek M (Center for Integrated Healthcare/Syracuse VAMC), Labbe AK (Center for Integrated Healthcare/Syracuse VAMC)

Objectives:
Significant advances have been made regarding interventions designed to address single health risk factors (e.g., smoking) for chronic disease, but there is considerably less knowledge about multiple risk factors. A greater understanding for the co-variation, prevalence, and impact of multiple risk factors could improve patient healthcare by guiding the design of integrated interventions. Specifically, the identification of specific clusters of risk factors that integrated interventions can target due to their significant effect on health and healthcare costs. In an effort to better understand the presentation of multiple risks in primary care, this study used latent class analysis (LCA) to examine the impact of the covariation of multiple risk factors in individuals presenting to primary care.

Methods:
Using VA's medical database, patients with a primary care encounter in VISN 2 from January 1 to June 30, 2005 (N = 10,043) were identified and information was collected regarding multiple risk factors commonly screened for in primary care (e.g., body mass index, blood pressure, smoking, alcohol use).

Results:
The results revealed three latent classes that were cross validated on an independent confirmatory sample. Then, clinical outcome data (e.g., number of primary care encounters) were obtained from each patient’s electronic medical record from July 1, 2005 through March 1, 2007. Regression analyses revealed that those individuals assigned to the latent class with the highest risk for health problems based on the number of positive risk factors had a higher number of primary care encounters compared to individuals assigned to the healthiest latent class. Further analyses are being conducted examining the relationship between specific latent classes and the number of behavioral health and emergency room encounters.

Implications:
This study exemplifies how LCA can be used to identify prevalent risk factor clusters that have significant effects on health outcomes, so integrated interventions can target these behaviors.

Impacts:
In addition, the data indicate the potential impact of efforts directed toward the development of empirically validated interventions targeting multiple risk factors in primary care.