The calculation of health literacy levels in self-report surveys: a new method using rasch analysis

Abstract Background The classification of adults’ health literacy levels in self-report surveys are diverse, but follow the same logic. Usually, 3 or 4 levels will be computed (e.g., inadequate, problematic and adequate health literacy (HL)) and the cut-off values are based on a reached value from the sum or mean score, e.g., 50% (cut-off between inadequate and problematic HL) und 66% (cut-off between problematic and adequate HL). Using normative cut-off values for HL level computation is flawed, because test difficulty and item difficulty are neglected. Rasch-Thurstonian thresholds will be presented as a new method to compute HL levels, in which test and item difficulty are considered for the classification of peoples’ competence. Methods To demonstrate the new method, data from the German 2021 COVID-HL school principal study (N = 2186) will be shown. The HLS-COVID-Q22 was used to assess COVID-19 HL. Participants rated on a 4-point likert-type scale the difficulty of 22 items (very easy, easy, difficult, very difficult). The classic method was computed (3 levels with the cut-offs at 50% and 66% of the sum score). Rasch-Thurstonian thresholds were computed. Here, the classification of 4 COVID-19 HL levels are based on peoples’ cumulative probabilities of choosing one of the four response categories for the different items. Results According to the classic method, 7.3% of participants had inadequate self-reported COVID-19 HL, 34.3% problematic and 58.4% adequate. Using Rasch-Thurstonian thresholds, 0% were classified into Level 1, 3.4% into Level 2, 68.6% into Level 3 and 28% into Level 4 of self-reported COVID-19 HL. Conclusions The new proposed method to compute HL levels, Rasch-Thurstonian thresholds, has the advantage of taking item and test difficulty into account in comparison to commonly used methods. The present study is a contribution to health literacy research by discussing the advancement of currently applied methods. Key messages • Using the same cut-off values for level computation in different questionnnaires is flawed, because test difficulty and item difficulty are neglected. • Rasch-Thurstonian thresholds are an appropriate method to compute health literacy levels for self-report questionnaires.