669-P: Factor Analysis of the Diabetes-Specific Risk-Taking Inventory

Introduction: Adolescence is marked by increased risk-taking behaviors and challenges engaging in optimal type 1 diabetes (T1D) treatment. We developed the Diabetes-Specific Risk-Taking Inventory (DSRI) to assess risky adolescent T1D self-management behaviors. In the current study, we conducted confirmatory factor analyses (CFAs) for the DSRI and examined model fit.

Methods: We surveyed a national sample of 224 adolescents (age 15-18 years; M age 16.9 ± 1.1 years; 49% female, 11% identified as Black, Indigenous, or Persons of Color; 76.8% used insulin pumps; M HbA1c= 8.5% ± 1.3); from the T1D Exchange registry in a cross-sectional design. Participants completed the DSRI, and we collected the most recent HbA1c value from registry data. We performed CFAs in R using the Lavaan package to evaluate three models: 1) a one-factor structure, 2) a two-factor structure reflecting items higher and lower in risk (item level of risk assessed by surveying diabetes health care providers), and 3) another two-factor structure reflecting items that associate with higher HbA1c (>9%) and lower HbA1c (<9%).

Results: The DSRI measure contains 31 items; however, one item was removed from CFA analyses due to a nonsignificant factor loading value. The one-factor CFA model demonstrated poor fit χ²(399, n=224) =1714.94, p<0.001, RMSEA=0.12, CFI=0.52, TLI=0.49, as did the two-factor structure differentiating between higher and lower HbA1c χ²(399, n=224) =1121.02, p<0.001, RMSEA=0.090, CFI=0.72, TLI=0.69. The two-factor CFA reflecting higher and lower risk items model demonstrated improved fit χ²(399, n=224) =896.19, p<0.001, RMSEA=0.085, CFI=0.77, TLI=0.75.

Conclusions: The DSRI is a valuable self-report measure that providers may use to screen for high-risk behaviors not routinely assessed in T1D clinical care. The results suggest that the high-risk vs. low-risk two-factor model is slightly better fitting than the one-factor model; however, additional factor structures should be assessed to identify a more robust structural model.