Evaluating the Conceptual Structure of the Barriers to Preexposure Prophylaxis (PrEP) Access Survey: A Confirmatory Item Factor Analysis
Author(s)
Patrick Sullivan, DVM, PhD1, Jeno Millechek, MA2, Woodie Zachry, PhD3, SAEID SHAHRAZ, MD3, Alice Hsiao, PharmD3, JeanPierre Coaquira Castro, MPH3.
1Emory University, Rollins School of Public Health, Atlanta, GA, USA, 2University of North Carolina at Chapel Hill, Chapel Hill, NC, USA, 3Gilead Sciences, Inc., Foster City, CA, USA.
1Emory University, Rollins School of Public Health, Atlanta, GA, USA, 2University of North Carolina at Chapel Hill, Chapel Hill, NC, USA, 3Gilead Sciences, Inc., Foster City, CA, USA.
OBJECTIVES: To evaluate the conceptual structure of the Barriers to Pre-Exposure Prophylaxis (PrEP) Access Survey using pilot results. The survey comprises 35 questions assessing psychosocial and structural barriers individuals may experience when receiving their initial PrEP prescription. The hypothesized conceptual structure of the model is designed to allow for interpretation of responses at the domain level and item level.
METHODS: Fit of pilot study survey results (N=235) to 3- and 4-domain models of conceptual structure was assessed using confirmatory item factor analysis (CIFA). Excluding questions with binary responses or ‘skip’ options, items were sorted into concepts according to the Information-Motivation-Behavioral Skills Model, supplemented with concepts from validated instruments aligned with survey objectives. CIFA evaluated 21 items grouped into 3 domains as attitudes (n=6), financial burden (n=5), and self-efficacy (n=10); in the 4-domain model, two self-efficacy items were moved to a physician communication domain. Model parameters were estimated with robust maximum likelihood estimation (MLR). Goodness-of-fit was assessed with the comparative fit index (CFI), Tucker-Lewis index (TLI), and root mean square error of approximation (RMSEA); values ≥0.90 (CFI and TLI) and ≤0.08 (RMSEA) indicated an acceptable fit. Model fit was compared using likelihood ratio tests (LRT).
RESULTS: The 4-domain model fit pilot survey data significantly better than the 3-domain model (LRT=27.19, p<0.001). This fit was assessed as acceptable by RMSEA (0.058) and less than acceptable by CFI (0.852) and TLI (0.830). Standardized factor loadings revealed three items with a weak or non-significant relationship to their domain. Fit of the 4-domain model improved upon post-hoc removal of two of these items (CFI=0.890; TLI=0.871; RMSEA=0.055), with similar results when all three were removed.
CONCLUSIONS: CIFA results support using a 4-domain model to evaluate responses to a subset of items in the Barriers to PrEP Access Survey, allowing enhanced result interpretation to better inform clinical and patient decision-making.
METHODS: Fit of pilot study survey results (N=235) to 3- and 4-domain models of conceptual structure was assessed using confirmatory item factor analysis (CIFA). Excluding questions with binary responses or ‘skip’ options, items were sorted into concepts according to the Information-Motivation-Behavioral Skills Model, supplemented with concepts from validated instruments aligned with survey objectives. CIFA evaluated 21 items grouped into 3 domains as attitudes (n=6), financial burden (n=5), and self-efficacy (n=10); in the 4-domain model, two self-efficacy items were moved to a physician communication domain. Model parameters were estimated with robust maximum likelihood estimation (MLR). Goodness-of-fit was assessed with the comparative fit index (CFI), Tucker-Lewis index (TLI), and root mean square error of approximation (RMSEA); values ≥0.90 (CFI and TLI) and ≤0.08 (RMSEA) indicated an acceptable fit. Model fit was compared using likelihood ratio tests (LRT).
RESULTS: The 4-domain model fit pilot survey data significantly better than the 3-domain model (LRT=27.19, p<0.001). This fit was assessed as acceptable by RMSEA (0.058) and less than acceptable by CFI (0.852) and TLI (0.830). Standardized factor loadings revealed three items with a weak or non-significant relationship to their domain. Fit of the 4-domain model improved upon post-hoc removal of two of these items (CFI=0.890; TLI=0.871; RMSEA=0.055), with similar results when all three were removed.
CONCLUSIONS: CIFA results support using a 4-domain model to evaluate responses to a subset of items in the Barriers to PrEP Access Survey, allowing enhanced result interpretation to better inform clinical and patient decision-making.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
MSR92
Topic
Methodological & Statistical Research
Topic Subcategory
Survey Methods
Disease
Infectious Disease (non-vaccine)