Measurement properties and comparative performance of health literacy screening questions in a predominantly low income African American population with diabetes

Objective: To examine the measurement properties of the 16 screening questions (16-SQ) of inadequate health literacy (HL) and their briefer version (3-SQ), and identify the best screen for inadequate HL in non-white populations. Methods: Sample included 378 individuals with type-2 diabetes. We computed sensitivity, specificity, positive and negative likelihood ratios, and C-indices, using the s-TOFHLA as a reference measure. We also conducted exploratory factor analysis, and used structural equation modeling (SEM) for confirmatory purposes. Results: Mean age was 56.1 years, 69% were female, and 83% were African-American. 10% had limited HL (s-TOHFLA scores <23). Six questions (6-SQ) were identified and included in the final item-reduced factor analysis, which showed good fit in confirmatory SEM (chi-square. =. 9.5; P=. 0.305; RMSEA. =. 0.023). Weighted summative score of the 6-SQ and the item "difficulty understanding written information" performed better than the 3-SQ in identifying patients with inadequate HL (C-indices 0.67 versus 0.75). Conclusion: The weighted summative score of the 6-SQ and the item "difficulty understanding written information" performed better than the other items or combinations of these items in identifying individuals with inadequate HL. Practice implications: The proposed weighting of scores could be applied in studies using these screening questions for better classification of inadequate HL.