OBJECTIVES:

The objective of this study was to identify which social determinants of health (SDoHs) are more impactful to the health outcomes of participants with obesity versus participants without obesity.

METHODS:

In order to detect differences in how each SDoH impacts health outcomes, our team compared results of statistical inference between disease and non-disease cohorts. The variables used for this process were comprised of demographic, socioeconomic, survey, and electronic health record data supplied by the All of Us (AoU) database. The AoU database, consisting of this data plus whole genome and Fitbit data, is collected from participants across the United States with a focus on diversity and inclusion. Forward stepwise linear regression and principal component analysis was used to examine the importance of SDoH categories on health outcomes, while lasso regression was used to identify the individual survey questions that are important to these outcomes.

RESULTS:

A review of the forward stepwise regression results for our obesity models yielded notable differences in SDoH importance from the non-obesity models. This method indicated that social and community context, such as discrimination or supportive relationships, had a greater impact on general health and quality of life for obese participants than non-obese participants. In contrast, features related to neighborhoods and built environment played a greater role in determining these outcomes for non-obese participants.

Results from the lasso regression models found survey questions relating to supportive relationships (i.e. “How often do you feel isolated from others?”) to be important to an obese participant’s general health, whereas education and neighborhood conditions were also important for non-obese participants.

CONCLUSIONS:

This study was able to highlight key factors in determining participant health outcomes, which can be taken for further, more focused study with a policy perspective in mind.