Measurement invariance of the Hopkins Symptoms Checklist: a novel multigroup alignment analytic approach to a large epidemiological sample across eight conflict-affected districts from a nation-wide survey in Sri Lanka

摘要

The alignment method, a novel psychometric approach, represents a more flexible procedure for establishing measurement invariance in geographically, ethnically, or linguistically diverse samples, especially in large epidemiological surveys. Although the Hopkins Symptoms Checklist (HSCL-25) has been used extensively in the field to assess anxiety and depressive symptoms, questions remain about the comparability of findings when the instrument is applied across regions in large-scale national surveys. The present study is the first in the field to apply the alignment method to test the structure and measurement invariance of the anxiety and depression dimensions of the HSCL-25 amongst Sri Lankan subpopulations (n?=?8456) stratified by geographical regions, levels of past exposure to conflict, and ethnic composition. Multigroup CFA analysis yielded non-converging models requiring substantial modifications to the models. As a result, multigroup alignment analysis was applied and the results supported the bifactorial structure and measurement invariance of the HSCL-25 across eight (severe and moderate) conflict-affected districts. The alignment analysis based on a good-fitting configural model yielded a metric non-invariance of 22.22% and scalar non-invariance of 5.88% (both under the established 25% threshold). The bifactorial model outperformed the tripartite and other models. In comparison to the anxiety items, the depressive items showed higher levels of metric non-invariance across districts. Our findings demonstrate the methodological feasibility of applying the alignment method to test the structure and invariance of the HSCL across ethnically diverse populations living in conflict-affected districts in Sri Lanka. Further studies are needed to examine ethnicity and language factors more critically.