Bridging big data in the ENIGMA consortium to combine non-equivalent cognitive measures

Area of Science:

• Neuroscience
• Behavioral Science
• Data Science

Background:

• Replication and reliability issues in neuroscience research are often addressed by increasing sample sizes.
• Integrating multisite and multi-instrument data presents significant challenges.
• Auditory verbal learning tasks (AVLTs) are widely used but lack score standardization across different versions and study sites.

Purpose of the Study:

• To develop a method for linking and harmonizing scores from various auditory verbal learning tasks (AVLTs).
• To create a unified scale for latent verbal learning ability.
• To improve the reproducibility and comparability of memory test data across studies.

Main Methods:

• An international secondary analysis aggregated raw data from 53 studies (N=10,505) involving AVLTs.
• The ComBat-GAM algorithm was employed to remove site-specific effects while retaining instrument-specific variations.
• A continuous item response theory model estimated latent verbal learning ability, enabling score conversion across different AVLTs.

Main Results:

• Harmonization reduced total cross-site score variance by 37%, preserving meaningful memory effects.
• Age was the most significant factor influencing scores (-11.4%), while race/ethnicity was not significant.
• Validated conversion tools were developed, allowing researchers and clinicians to standardize AVLT scores.

Conclusions:

• Global harmonization initiatives can effectively address reproducibility challenges in the behavioral sciences.
• The developed methods and online tools facilitate score conversion across different AVLT instruments.
• This work establishes a foundation for more reliable and comparable memory research using Big Data.