Workgroup Selection
Workgroups are an integral part of this conference. Members of workgroups design and implement analyses during the conference to test scientific hypotheses with a goal of developing and submitting manuscripts. The conference provides an intensive start-up to this process, but it is expected that the workgroups will continue to interact after the meeting to complete and refine analyses and develop manuscripts. Our experience is that organizing workgroups in advance facilitates their coordination and productivity during and after the conference.
The 2024 conference will feature data from the Harmonized Cognitive Assessment Protocols (HCAP) and Alzheimer’s Disease Sequencing Project (ADSP). HCAP data will include cognitive data from the United States, Mexico, England, South Africa, China, Chile, and India, as well as background demographics and harmonizable risk factors from parent HRS studies in these countries. ADSP data will include cognitive and imaging/biomarker data from ADNI in the United States and KBASE in South Korea.
Potential Workgroups
- Analysis of variation in associations with cognitive outcomes attributable to intersectional and national identity using a MAIHDA framework. Popularly used for the study of socioeconomic disparities in health, multilevel analysis of individual heterogeneity and discriminatory accuracy (MAIHDA) models introduce an opportunity to apply multilevel analysis to explore cognitive health differences within and between strata defined by factors such as countries, states/provinces within countries, and sociodemographic dimensions. By applying the MAIDHA analytic framework within a cross-national framework, this workgroup will provide quantitative and theoretical means of documenting intersectional cross-national disparities in cognitive functioning. Data on risk factors are available from International Partner Study core interviews and HCAP interviews.
- Importance: The 2020 Lancet Commission’s report on dementia identified 12 potentially modifiable risk factors for dementia. There has been a boon of subsequent studies reporting population attributable risks for dementia due to these and other risk factors. Understanding cross-national differences in variation of modifiable risk factors for cognitive impairment and dementia from an intersectional perspective could inform the tailoring of public health messaging and intervention efforts.
- Possible Datasets: HCAP and ADSP
- Intra-national and cross-national differential item functioning in cognitive testing. Do measurement differences in cognitive abilities obscure our ability to detect cross-national associations? Previous published studies have applied MIMIC modeling to investigate subnational and cross-national measurement differences in cognitive testing in the HCAPs. Although most studies find some measurement differences that do not affect summary factors too much, there is likely residual cross-country differential item functioning due to lingering effects of illiteracy, innumeracy, and cultural differences in how cognition is measured based on cognitive tests derived from Western/HIC settings. Research on the detection of measurement differences could readily be extended to subsets of populations (e.g., literate groups from LMICs compared to HICs), or to within-country comparisons (e.g., languages, sensory function, literacy, or other potential sources of measurement differences within a country). In addition to substantive questions, this work group may explore and compare results from novel approaches to detection of differential item functioning in addition to MIMIC models, such as penalized SEM, alignment analysis, and MAIDHA.
- Importance: Although best practices for cross-national research advise that countries should be used as interaction variables with exposures to investigate cross-national differences in cognitive aging, there are considerable cross-national differences in levels of cognitive performance that imply residual differences in testing not detected by existing tests of differential item functioning.
- Possible Datasets: HCAP and ADSP
- Triangulating cross-national comparisons using cognitive tests with those using imaging and biomarker data. There is a plethora of demographic, socioeconomic, cardiovascular, behavioral, and other exposure variables that may have differing relationships cross-nationally with cognitive functioning in older adults. There are many scientific motivations for these types of comparisons, such as triangulating estimates across countries thought to have different confounding structures, or building evidence for the presence of effect modification across countries with different cultures and contexts. However, in cross-national research, there is variation in cognitive testing due to factors unrelated to cognitive functioning, such as differences in language, literacy, numeracy, and other sources. This complicates studies of cross-national differences in associations. One approach is to test for measurement differences in testing using available data on cognitive tests (e.g., work group 2). Another approach is to replicate associations based on cognitive tests with imaging and biomarker-based data. For example, are potential cross-national differences in the association of sex or education with cognitive performance also present when using imaging data? Deviations could indicate test bias, or cognitive reserve. This work group will leverage data from the ADSP’s ADNI and KBASE harmonized data.
- Importance: Cross-national comparative analyses of risk factor associations provides an opportunity to identify drivers of both individual variation and population average differences. Identifying drivers of country-level differences in cognitive function could be key to identifying successful population interventions, as such causes may be masked on the individual level if they are relatively homogeneously distributed within a country.
- Datasets: ADSP
- Simulation study to examine effects of differential precision of cognitive outcomes on risk factor associations. Measurement error can cause problems in a wide variety of ways, such as bias, imprecision (e.g., larger standard errors around effect estimates), and random error. Bias often arises from measurement error in exposures – either differential or nondifferential by levels of the outcome – rather than outcomes. While measurement error in outcome variables typically leads to imprecision in estimates, but not bias, what happens if measurement precision of a cognitive outcome varies longitudinally, ranging from extremes of r=0.6 (e.g., a core wave interview) to r=0.9 (e.g., an HCAP interview)? This simulation group will leverage an existing platform to answer this and related questions.
- Importance: The Health and Retirement Study in the US and its international partner studies in Mexico, England, SHARE countries, China, and India include a brief cognitive assessment consisting of a recall on a single-trial 10-word list, questions about orientation to time/place, and tests of attention/calculation. Each of these studies have more detailed HCAP exams. Many researchers may be encouraged to link cognition from Core surveys to cognition from HCAPs, thus timely guidance is needed as to the soundness of this approach.
- Datasets: Simulation data, with inputs from HCAPs
- Language. The role of language in cognitive aging is multifaceted and likely varies across contexts. For instance, everyday aspects of language such as bilingualism and literacy are thought to be positively associated with cognition and with life course socioeconomic factors, yet it is not clear whether these associations are similar across international contexts. Similarly, language as cognition could likely vary substantially across contexts. Most cognitive batteries evaluate diverse aspects of language abilities (i.e., semantic vs non-semantic) and we construct language measures based on this broader set. However, it is unclear if the different language measures load the same way onto the latent construct of language, or how language relates to outcomes across culturally and linguistically diverse populations. This workgroup will evaluate different aspects of language as they relate to cognitive outcomes cross-nationally.
- Datasets: HCAP and ADSP