Identifying High-Priority Profiles for Edentulism and Other Health Outcomes Through Population-Based Clustering.

Introduction: The aim of this study was to create meaningful clusters of individuals with similar features at the start of the study and to link these clusters to the rate of complete edentulism. These clusters are also linked to death, diabetes, cardiovascular disease, and general distress connecting edentulism to overall health. Methods: Using the 2019-2022 longitudinal Medical Expenditure Panel Survey, the authors identified baseline risk profiles for adults (aged ≥18 years) through weighted k-modes clustering. To further examine the relationship among these groups, a second-stage hierarchical clustering method was applied to visualize cluster similarities. Input variables were smokers, self-reported health and mental health status, functional limitations, SES, gender, employment, and age. Input variables were measured close to the start of the study. Profiles of clusters were determined on the basis of the modal characteristics of each input variable. External variables (edentulism, death, cardiovascular disease, and diabetes) were assessed over the aggregate 4-year study interval, with Kessler-6 scores for general distress evaluated near to the end of the study. Results: The second-stage hierarchical analysis identified 3 overall blocks, described as high, medium, and low at-risk, comprising 17 clusters from the first stage. The highest risk block consisted of 5 clusters, all characterized by unfavorable self-reported general health and mental health statuses, yet only 1 cluster identified smoking as a modal characteristic. Within this block, 3 clusters most at-risk had the highest rates of complete edentulism, death, and diabetes, while being among the top 5 clusters with the highest rates of cardiovascular disease and Kessler-6 scores. Although these 3 clusters represented only 8.4% of the total cohort, they accounted for 25.3% of the total edentulous population, 50.0% of all deaths, 22% of all those with diabetes, and 16% of those with cardiovascular disease. The remaining 2 clusters in this block shared similar baseline characteristics but featured a younger age distribution and were more favorable but non-negligible for rates of diabetes or cardiovascular disease than the other 3 clusters. Conclusions: Clustering provides a pragmatic framework to identify shared characteristics within populations of existing or emerging at-risk groups. By identifying these high-priority groups, healthcare systems can better tailor population health programs to those most in need. Prioritizing initiatives that provide affordable dental care at all ages, alongside oral health literacy programs and investment in preventive services, has the potential to improve oral health and overall health outcomes.