Community Health and Prevention

Presentation Title (IN ALL CAPS)

USING ITEM RESPONSE THEORY (IRT) TO DESCRIBE AT-RISK PATIENTS PARTICIPATING IN A HEALTH COACHING PROGRAM

Departmental Affiliation and City, State, Zip for All Authors

Department of Biostatistics and Epidemiology, UNT Health Science Center, 3500 Camp Bowie Blvd., Fort Worth, Texas 76107; Department of Biostatistics and Epidemiology, UNT Health Science Center, 3500 Camp Bowie Blvd., Fort Worth, Texas 76107; Department of Behavioral and Community Health, UNT Health Science Center, 3500 Camp Bowie Blvd., Fort Worth, Texas 76107

Classification

SPH Student (For Competition)

Research Presentation Category

Community Health and Prevention

Brief Narrative or Summary

In this research project, we used Item Response Theory (IRT) models to evaluate the baseline characteristics of subjects involved in the m.chat program, a technology-assisted health coaching program designed to improve health and wellness indicators among people in permanent supportive housing.​ We report which variables had the greatest and least ability to differentiate subjects. Using these results, we can identify clients who are most at-risk and identify disparities and risk factors by such variables as race, gender etc.

Scientific Abstract

Item Response Theory (IRT) is a psychometric tool originally developed to assess ability, attitude and latent traits in education. In recent years, it has found wide-spread application in behavioral health, quality of life and clinical research. We used IRT models to evaluate the baseline characteristics of subjects involved in the m.chat program. M.chat is a technology-assisted health coaching program designed to improve health and wellness indicators among people in permanent supportive housing. The program was funded by a Medicaid 1115 Waiver to the State of Texas. Data was available on 115 baseline items for 174 subjects. For this analysis, we focused on four domains that are often associated with increased risk: (i) recent interactions with the criminal justice system, including arrests, being on probation or parole, or being incarcerated, (ii) satisfaction with current status including marital status, living status, or family and friend status, (iii) living with someone with a current alcohol or drug problem, and (iv) treatment and/or counseling for family, social or psychological problems. We used the Rasch model to analyze dichotomous outcomes and the graded response model for outcomes with more than two categories. Our preliminary findings indicate law enforcement-related items have the greatest ability to differentiate subjects where items pertaining to treatment and counseling for family, social and psychological problems have the least ability to differentiate subjects. These results might be used to identify clients who are most at-risk. Next, we plan to use this approach to identify disparities and risk factors by race, gender etc.

This document is currently not available here.

Share

COinS
 

USING ITEM RESPONSE THEORY (IRT) TO DESCRIBE AT-RISK PATIENTS PARTICIPATING IN A HEALTH COACHING PROGRAM

Item Response Theory (IRT) is a psychometric tool originally developed to assess ability, attitude and latent traits in education. In recent years, it has found wide-spread application in behavioral health, quality of life and clinical research. We used IRT models to evaluate the baseline characteristics of subjects involved in the m.chat program. M.chat is a technology-assisted health coaching program designed to improve health and wellness indicators among people in permanent supportive housing. The program was funded by a Medicaid 1115 Waiver to the State of Texas. Data was available on 115 baseline items for 174 subjects. For this analysis, we focused on four domains that are often associated with increased risk: (i) recent interactions with the criminal justice system, including arrests, being on probation or parole, or being incarcerated, (ii) satisfaction with current status including marital status, living status, or family and friend status, (iii) living with someone with a current alcohol or drug problem, and (iv) treatment and/or counseling for family, social or psychological problems. We used the Rasch model to analyze dichotomous outcomes and the graded response model for outcomes with more than two categories. Our preliminary findings indicate law enforcement-related items have the greatest ability to differentiate subjects where items pertaining to treatment and counseling for family, social and psychological problems have the least ability to differentiate subjects. These results might be used to identify clients who are most at-risk. Next, we plan to use this approach to identify disparities and risk factors by race, gender etc.