Members of the Gateway team in collaboration with the Harmonized Cognitive Assessment Protocol (HCAP) International Network recently published an article in the journal Alzheimer’s & Dementia that discusses theoretical and methodological considerations for researchers who wish to use the HCAP data for cross-national comparisons and offers a set of recommended best practices for these comparisons. In this blog post, we discuss the motivation behind this article and what we hope that HCAP data users will get out of it.

What is the HCAP?

The HCAP is designed to be a flexible, yet comparable instrument with which to measure cognitive function and status in diverse older populations around the world. It has been embedded in sub-studies of the US Health and Retirement Study and its International Partner Studies in 18 countries around the world, with future HCAP administrations planned for at least six additional countries. These existing and planned HCAP studies represent approximately 75% of the global population aged ≥65 years. The HCAP is thus poised to be a key data source for global dementia research in the years to come.

Why did we write the “Best Practices” article?

We wrote this article because we wish researchers to conduct thoughtful and high-quality cross-national comparisons using the HCAP data. The HCAP data are a publicly available resource and are becoming increasingly popular within the research community. The cross-nationally harmonized data on cognitive function provided by the HCAP are an incredible opportunity to gain new insights in dementia etiology around the world. However, cognitive function is an outcome that is highly sensitive to linguistic, cultural, social, and educational factors in its measurement. Cross-national diversity in such factors has the potential to introduce bias into the measurement and interpretation of HCAP data. To reduce such bias, the HCAP studies have conducted extensive adaptation and pre-testing work to ensure that their HCAP batteries are appropriate for their study populations. In addition, the Gateway team and the HCAP International Network have collaborated to conduct what is referred to as “pre-statistical” and “statistical” harmonization of some of the HCAP data that are currently publicly available to account for potential differences in the meaning and difficulty of HCAP cognitive test items that have been adapted across different countries. The processes of “pre-statistical” and “statistical” harmonization are described in more detail in the “Best Practices” article, as well as in other sources. An important note to make at this point is that the HCAP data released by individual studies are not the statistically harmonized scores, which, at the time of writing of this blog post, are currently available as user-contributed data for the first waves of each of the HRS-HCAP, ELSA-HCAP, LASI-DAD, CHARLS-HCAP, Mex-Cog, and HAALSI-HCAP studies.

Figure 1 from article: Map of countries with Harmonized Cognitive Assessment Protocol (HCAP) studies, according to stage of progress as of July 2023: funding applied for, funding received, data collection begun, data collection complete, and data collection complete and publicly released.

Despite adaptations of the HCAP battery during data collection, and post-hoc statistical harmonization procedures that have been completed for some HCAP studies, there may remain residual bias due to incommensurate measurement of cognitive function across studies. There are country-level differences observed in the means and distributions of the statistically harmonized HCAP data, which are visualized in the “Best Practices” articles. These differences may be due to a mixture of residual bias in measurement, in addition to true population differences. In the “Best Practices” article, we discuss how observed population differences in later-life cognitive function may be influenced by cross-population differences in exposures to risk and protective factors that people experience across their lifetimes. For example, a recent cross-national comparison of HCAP data conducted by our team found that population differences in educational attainment explain about 50-90% of observed population differences in later-life cognitive function across the US, England, China, Mexico, and India. We hope that HCAP data users will be thoughtful about considering the historical and contextual reasons why such differential population distributions in later-life cognitive function may arise and explore these reasons in their analyses, where possible.

Final thoughts

We hope that HCAP data users will read the “Best Practices” article before embarking on cross-national comparisons of risk factor associations with later-life cognitive function. In addition to an in-depth theoretical and methodological discussion, the article provides a straightforward list of recommendations, posed as questions that researchers may ask themselves as they plan and conduct their analyses. We hope that this article will promote high-quality analyses and fair use of the HCAP data around the world.