Journal Article

Four New Patient-Reported Outcome Measures Examining Health-Seeking Behavior in Persons With Type 2 Diabetes Mellitus (REDD-CAT): Instrument Development Study

Moldovan, Ioana, Carlozzi, Noelle E, Kallen, Michael A, Bragg, Alexa, De La Cruz, Barbara A, Martin-Howard, Jessica, Miner, Jennifer A, Mitchell, Suzanne E, Troost, Jonathan P, Jack, Brian W

2024

Overview

The management of type 2 diabetes mellitus (T2DM) includes mastery of complex care activities, self-management skills, and routine health care encounters to optimize glucose control and achieve good health. Given the lifelong course of T2DM, patients are faced with navigating complex medical and disease-specific information. This health-seeking behavior is a driver of health disparities and is associated with hospitalization and readmission. Given that health-seeking behavior is a potentially intervenable social determinant of health, a better understanding of how people navigate these complex systems is warranted. To address this need, we aimed to develop new patient-reported outcome (PRO) measures that evaluate health-seeking behavior in persons with T2DM. These new PROs were designed to be included in the Re-Engineered Discharge for Diabetes-Computer Adaptive Test (REDD-CAT) measurement system, which includes several other PROs that capture the importance of social determinants of health. Overall, 225 participants with T2DM completed 56 self-report items that examined health-seeking behaviors. Classical Test Theory and Item Response Theory were used for measurement development. Exploratory factor analysis (EFA; criterion ratio of eigenvalue 1 to eigenvalue 2 being >4; variance for eigenvalue 1 ≥40%) and confirmatory factor analysis (CFA; criterion 1-factor CFA loading <.50; 1-factor CFA residual correlation >.20; comparative fit index ≥0.90; Tucker-Lewis index ≥0.90; root mean square error of approximation <0.15) were used to determine unidimensional sets of items. Items with sparse responses, low-adjusted total score correlations, nonmonotonicity, low factor loading, and high residual correlations of high error modification indices were candidates for exclusion. A constrained graded response model was used to examine item misfit, and differential item functioning was examined to identify item bias. Cronbach α was used to examine internal consistency reliability for the new PROs (criterion ≥0.70), and floor and ceiling effects were examined (criterion ≤20%). Four unidimensional sets of items were supported by EFA (all EFA eigenvalue ratios >4; variance for eigenvalue 1=41.4%-67.3%) and CFA (fit statistics all exceeded criterion values). This included (1) \"Health-Seeking Behavior: PCP-Specific\" (6 items); (2) \"Health-Seeking Behavior: General Beliefs\" (13 items); (3) \"Health-Seeking Behavior: Family or Friends-Specific\" (5 items); and (4) \"Health-Seeking Behavior: Internet-Specific\" (4 items). All items were devoid of differential item functioning for age, sex, education, or socioeconomic status factors. \"Health-Seeking Behavior: General Beliefs\" was developed to include both a computer adaptive test and a 6-item short form version; all other PROs were developed as static short forms. The psychometric reliability of these new PROs was supported; internal consistency ranged from acceptable to excellent (Cronbach α=.78-.91), and measures were free of significant floor or ceiling effects (floor effects range: 0%-8.9%; ceiling effects range: 0%-8.4%). The new REDD-CAT Health-Seeking Behavior PROs provide reliable assessments of health-seeking behaviors among those with T2DM.

• Publisher: JMIR Publications
• Subject: Achievement tests / Behavior / Data warehouses / Decision making / Diabetes / Health care access / Health care policy / Health education / Health Insurance Portability & Accountability Act 1996-US / Health literacy / Illnesses / Informed consent / Original Paper / Patients / Professional ethics / Reading