Statistical Approaches for the Analysis of Combined Health-Related Factors in Association with Adult Cognitive Outcomes: A Scoping Review

Abstract

Background: Dementia prevention is a global health priority, and there is emerging evidence to support associations between individual modifiable health behaviors and cognitive function and dementia risk. However, a key property of these behaviors is they often co-occur or cluster, highlighting the importance of examining them in combination. Objective: To identify and characterize the statistical approaches used to aggregate multiple health-related behaviors/modifiable risk factors and assess associations with cognitive outcomes in adults. Methods: Eight electronic databases were searched to identify observational studies exploring the association between two or more aggregated health-related behaviors and cognitive outcomes in adults. Results: Sixty-two articles were included in this review. Fifty articles employed co-occurrence approaches alone to aggregate health behaviors/other modifiable risk factors, eight studies used solely clustering-based approaches, and four studies used a combination of both. Co-occurrence methods include additive index-based approaches and presenting specific health combinations, and whilst simple to construct and interpret, do not consider the underlying associations between co-occurring behaviors/risk factors. Clustering-based approaches do focus on underlying associations, and further work in this area may aid in identifying at-risk subgroups and understanding specific combinations of health-related behaviors/risk factors of particular importance in the scope of cognitive function and neurocognitive decline. Conclusion: A co-occurrence approach to aggregating health-related behaviors/risk factors and exploring associations with adult cognitive outcomes has been the predominant statistic approach used to date, with a lack of research employing more advanced statistical methods to explore clustering-based approaches.