Study of the psychometric properties of the HLS-EU-12 questionnaire in rural Bangladesh

by Fakir M. Amirul Islam

Poorer health literacy leads to limited knowledge of diseases and lower adherence to medication. Several tools are available for measuring health literacy, including the HLS-EU-Q12 questionnaire. This study aimed to examine the psychometric properties of the HLS-EU-Q12 questionnaire using Rasch measurement theory to assess general health literacy among adults in rural Bangladesh.

Baseline data was collected through a cluster randomized control trial (RCT) in the Narail district of Bangladesh from 1 December 2020 to 31 January 2021, involving 307 adults aged 30–75 with hypertension. The HLS-EU-Q12 questionnaire, which has 12 items, was validated using the Rasch measurement model and analyzed using RUMM2030. The analysis was focused on differential item functioning (DIF) across gender and age, targeting, multidimensionality, response dependency, and item categorization.

The sample consisted of almost equal proportions of females (50.2%) and males (49.8%), and control (50.8%) and intervention participants (49.2%). Initial examination indicated that the tool had a poor overall fit with the Rasch model, shown by a significant item-trait interaction (χ2 = 100.5 df = 48, p < 0.001). The reliability, measured by person separation index (PSI), was 0.746, which was considered satisfactory. The overall item fit residual (IFR) (M = 0.236, SD = 1.318) and the person fit residuals (PFR) (M = -0.186, SD = 1.03) were within the acceptable range of SD ± 1.4. All items were found to have ordered thresholds, suggesting that respondents had no difficulty differentiating between response options on the 4-point Likert-type scale. Only item 12 (“Join a sports club or exercise class if you want to”) had a fit residual value outside the acceptable range. Removing item 12 resulted in a good overall fit (χ2 = 60.35 df = 44, p = 0.05) and a slight improvement to the PSI (0.746 for 12 items vs.0.756 for 11 items). None of the items showed significant DIF for age and gender. Only two items showed residual correlation coefficient 0.20, indicating possible redundancy. The analysis demonstrated the scale’s unidimensionality, as shown by the lower bound of a binomial 95% confidence interval (CI) for the observed proportion (5.2%) being within the 95% confidence limit.

The study demonstrated the potential effectiveness of the HLS-EU-Q12 as a tool for assessing health literacy among adults in Bangladesh. However, further study is needed to evaluate the tool across different populations, including an in-depth investigation of item 12 to determine its inclusion or removal.

ClinicalTrials.gov NCT04505150. Registered on 7 August 2020.