Machine learning and national health data to improve evidence: finding segmentation in individuals without private insurance

Objective: Individuals without private health insurance have less access to healthcare, therefore are more prone to experience poor health when compared to those who have. Segmentation is an approach to find homogenous groups of people with the purpose of tailoring services and products. In public policy, segmentation might be used to identify characteristics and needs of specific groups and deliver targeted programs and spare costs. We aim to identify and describe segments within the uninsured population to aid targeted policy actions and improve health. Methods: We used secondary data collected from a representative, nationwide health survey (n=18,204). For the purpose of our analysis, we included data from individuals who answered “no” to the question: “Do you have private health insurance?” (n=12,134). Variables pertaining information on socio-demographic, health status, access and care were used. A multiple correspondence analysis was performed to find principal components followed by a hierarchical cluster. Results: We found three clusters. The first (54.12% of our sample) composed by a group of young, middle aged and professionally active individuals without health problems. The second (36.70%), a cluster of aging individuals composed especially by elderly women, either retired or fulfilling domestic tasks, with a long-term health problem. The last (9.17%) composed by elder people, with long-term health problem and scoring low in mental health related questions. Conclusion: Our study found three clusters (profiles of individuals) among the uninsured. Ultimately, our findings aim to support policy makers to deliver customized actions to improve health and provide cost-effective policies.