Measuring religiosity in longitudinal secondary datasets: Comparison of single-item measures with validated scales in a UK cohort study (ALSPAC).

Background: Many studies use single-item variables to measure religiosity, such as religious belief, identity or service attendance. However, there are many different hypothesised dimensions of religiosity and it is often unclear how these single-item measures may map onto these theorised constructs. ALSPAC (Avon Longitudinal Study of Parents and Children) primarily relied on single-items to measure religiosity, but more recently has included validated questionnaires (DUREL [Duke University Religion Index] and I/EMSS [Intrinsic/Extrinsic Measurement: I/E-Revised and Single-Item Scales]). This paper aims to: i) assess whether the validated measures load together as expected in ALSPAC; and ii) understand which religiosity dimensions the single-item measures relate to. Methods: Twenty religiosity questions were asked to ALSPAC offspring and parents approximately 28 years after the offspring's birth. We used three exploratory factor analyses to assess how the different items related to one another on: i) the pre-validated DUREL and I/EMSS measures to examine whether they represent distinct dimensions, as intended ; ii) all 20 religiosity measures; and iii) the pre-validated measures and the single-item measures also used at previous ALSPAC timepoints (13 measures). Results: The first factor analysis showed that, beyond a single religiosity factor, these pre-validated items did not always work as intended. For instance, intrinsic religiosity items loaded together, but extrinsic religiosity items were often separated. The second and third analyses showed that single-item measures did not relate well to hypothesised dimensions of religion but did form two broad factors of belief-based and behaviour-based items. Results were broadly comparable across both ALSPAC generations. Conclusions: These results show that pre-validated measures of religiosity do not always behave as expected in ALSPAC, while the single-item measures do not easily map onto specific dimensions of religiosity. These results will help researchers better understand the ALSPAC religiosity data and inform analyses using these data.