Discipline: Biological Sciences
Subcategory: Physiology and Health
Session: 1
Room: Council
Nicholas Skoufis - Vanderbilt University
Bariatric surgery is an important intervention in the treatment of obesity and related illnesses. However, weight loss outcomes following surgery varies significantly and social determinants of health at the level of patients’ home communities may play a role. Our objective in this investigation is to identify demographic and social factors within patients’ home communities that contribute to reduced weight loss after bariatric surgery. We hypothesized that factors such as poverty, demographics, and education may be associated with outcomes of bariatric surgery. We match deidentified data from 3000 surgery patients from Vanderbilt University Medical Center with census data from their home communities at the census-tract level from the CDC. We determined patients’ percent change in body mass index three months after surgery (pcBMI) and identified patients in the upper and lower quartiles of pcBMI, coding them as 1 and 0 respectively. To identify association, we performed linear regressions between the pcBMI coding and each census variable. We found that the percentage of the community belonging to racial minority groups (P_MINORITY) and the percentage of the community without insurance (P_UNINSURED) had significant association with the pcBMI coding (p < 0.05). From there, we looked for intersections of medical comorbidities associated with pcBMI. To identify these comorbidities, we first took the subset of patients who had a given comorbidity. Then, we grouped patients by quartile of a census variable. We used a t-test (p < 0.05) to compare the pcBMI of the upper and lower quartile of the census variable. Applying this method for each comorbidity and the census variables P_MINORITY and P_UNINSURED revealed situations in which social determinants of health and medical comorbidities were associated with poor weight loss outcomes. For instance, among patients with type 1 diabetes, patients from communities with lower rates of insurance experienced less weight loss than patients from communities with higher rates of insurance. Similarly, among female patients, patients from communities with higher percentages of minorities experienced less weight loss. Several other comorbidities were identified with this method. These findings help elucidate the role of community factors in facilitating weight loss and may help guide healthcare providers to give special attention to patients from more vulnerable communities. Limitations of this investigation include the possibility of unaddressed confounding medical factors. Additionally, the patients were predominantly from middle Tennessee. We are undertaking further investigations into using causal inference and machine learning methods for counterfactuals to establish a causal relationship between demographic factors and weight loss outcomes.
Funder Acknowledgement(s): NSF REU Site Award #2050895
Faculty Advisor: You Chen, you.chen@vumc.org
Role: I developed the method for analyzing the data and wrote programs in R to perform the analysis