# Harmonization of Cognitive Data Across Cohort Studies

## TL;DR

This paper introduces a harmonized cognitive score to compare cognitive decline across two large US aging studies, despite differences in testing methods.

## Contribution

The novel contribution is a harmonized general cognitive factor score derived using item response theory and item banking across two studies.

## Key findings

- Harmonized scores showed lower values in dementia groups compared to cognitively normal participants in both HRS and NHATS.
- Cognitive decline was steeper for older individuals, those with less education, and those with health conditions like hypertension or diabetes.

## Abstract

Harmonizing longitudinal population-based studies affords the ability to compare cognitive decline even when protocols for measuring cognition differ between cohorts. We derived a longitudinally harmonized general cognitive factor score based on cognitive tests from the Health and Retirement Study (HRS) and the National Health and Aging Trends Study (NHATS), two of the largest population-based studies with longitudinal cognitive testing in the US. Using item response theory methods supplemented by an item banking approach, we leveraged common and unique tests administered in 2012-2018 Core interviews of the two studies to derive a harmonized general cognitive factor score. We evaluated concurrent and predictive criterion validity of harmonized scores by: 1) cross-sectionally comparing scores at baseline (2012) with cognitive impairment status and demographics; and 2) longitudinally assessing associations of demographics and risk factors. Cross-sectionally, the harmonized score was lowest among dementia/probable dementia groups in both HRS (z=-1.55 SD) and in NHATS (z=-2.21 SD) compared to cognitively normal participants in HRS (z = 0.27 SD) and NHATS (z=-0.05 SD). Longitudinally, in both studies, cognitive decline is steeper for participants at older ages within participants, with less education, and those reporting hypertension, stroke, or diabetes, relative to those without. This harmonized cognitive score can be a resource for aging researchers to examine and compare rare risk factors for cognitive performance and decline across large population-based studies. It returns patterns of change across demographic and health groups consistent with prior literature.