HARMONIZATION OF COGNITIVE FUNCTION IN LONGITUDINAL INTEGRATIVE DATA ANALYSIS - Results From Two Chinese Datasets

Sep 29, 2026

Integrating multiple longitudinal studies is essential for understanding cognitive aging across population groups, yet conventional harmonization (e.g., z-scores or percentiles) may ignore differences in measurement properties and sample demographics, leading to bias. This study uses Item Response Theory (IRT) to harmonize general and domain-specific cognitive function across two major Chinese national aging datasets, aiming to improve comparability while maintaining reliability and validity in pooled analyses.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages2

Reading Time5 min

AccessFree to download

Objective

Method

Data sources

Analytic approach

Results

Discussion

Future work

672 Innovation in Aging, 2025, Vol. 9, No. S2 Abstract citation ID: igaf122.2294 HARMONIZATION OF COGNITIVE FUNCTION IN LONGITUDINAL INTEGRATIVE DATA ANALYSIS: RESULTS FROM TWO CHINESE DATASETS Chang Yu, University at Buffalo, SUNY, Getzville, New York, United States Objective: Integrating data from multiple longitudinal studies is crucial for understanding cognitive aging across different population groups. However, common harmonization methods, such as z-scores or percentiles, often fail to account for differences in measurement properties and sample demographics, potentially introducing bias. This study applies Item Response Theory (IRT) to harmonizing general and domain-specific cognitive function across two major Chinese national longitudinal aging studies. By developing a harmonized framework, this study aims to enhance comparability while ensuring reliability and validity in pooled analyses. Method: The sample consists of individuals who completed at least one wave of cognitive testing from two major datasets. The Chinese Longitudinal Healthy Longevity Survey Innovation in Aging, 2025, Vol. 9, No. S2 673 (CLHLS) (N = 38,760; 1998–2018; 8 waves) primarily focuses on older adults (65+), while the China Health and Retirement Longitudinal Study (CHARLS) (N = 19,150; 2011–2018; 4 waves) includes adults aged 45 and older. This study applies factor analysis to identify common latent cognitive constructsand employs IRT-based scale linking to align cognitive scores. Differential Item Functioning (DIF) analysis is conducted to examine variations in different datasets. Results: Preliminary findings suggest that factor score harmonization improves measurement alignment compared to traditional standardization techniques. However, differences in item properties and demographic distributions pose challenges, requiring further refinement of linking procedures. Discussion: This study advances longitudinal integrative data analysis in cognitive aging research, demonstrating the feasibility of factor score-based harmonization while highlighting potential limitations. Future work will refine linking approaches and assess how harmonization impacts cognitive trajectory analyses.

FAQs

Why can standard harmonization methods introduce bias when pooling longitudinal cognitive data?

Because z-scores or percentiles may not account for differences in measurement properties and sample demographics across studies, which can distort comparability.

Which datasets are used in the study to harmonize cognitive function?

The study uses the Chinese Longitudinal Healthy Longevity Survey (CLHLS) and the China Health and Retirement Longitudinal Study (CHARLS).

What is the main conclusion from the preliminary results?

Factor score harmonization improves measurement alignment compared with traditional standardization, but item property and demographic differences require further refinement of the linking procedure.

Sep 29, 2026

Upload by Aria Callaghan