Cross-national statistical harmonization of the Center for Epidemiologic Studies Depression (CES-D) scale among older adults in China, England, India, Mexico, South Africa, and the United States

Objectives: We examined differential item functioning (DIF) of the Center for Epidemiologic Studies Depression Scale (CES-D) items by country and statistically harmonized common cross-national factor scores for the CES-D to aid further cross-national research. Study Design and Setting: Data were from Harmonized Cognitive Assessment Protocol (HCAP) studies in China (N = 9639), England (N = 1262), India (N = 4048), Mexico (N = 1918), South Africa (N = 631), and the United States (N = 3321). Multiple indicators, multiple causes models were estimated to test DIF in the CES-D items by country. DIF items were defined as having an odds ratio (OR) outside the range of 0.75–1.25 in multiple indicators, multiple causes models. We evaluated DIF impact and identified salient DIF by examining whether the difference between DIF-adjusted factor scores and non-DIF–adjusted factor scores exceeded a threshold of 0.30 standard deviation (SD) units. Confirmatory factor analysis was used to create DIF-adjusted, cross-nationally harmonized CES-D factor scores. Results: Controlling for underlying depressive symptoms, HCAP participants in India had higher odds of reporting being not hopeful about future (OR = 1.38, 95% confidence interval [CI]: 1.34–1.42), not enjoying life (OR = 1.43, 95% CI: 1.38–1.48), and being unhappy (OR = 1.29, 95% CI: 1.25–1.34), compared to HCAP participants in the United States. These identified DIF items artificially increased mean harmonized CES-D factor scores by 0.48 SD units in the India HCAP, with over 50% of the factor scores increased by over 0.30 SD units, indicating salient DIF in the India HCAP. Conclusion: Our findings demonstrate cross-national heterogeneity in the expression of depressive symptoms. We provide DIF-adjusted CES-D factor scores to improve the quality of cross-national comparisons in aging research.