Random forest modeling using socioeconomic distress predicts hernia repair approach

Abstract

Background: Surgical techniques for abdominal wall hernia repair have advanced, yet it is unclear if all patient populations experience these innovations equally. We hypothesized that in patients undergoing abdominal wall herniorrhaphy, there would be socioeconomic variation between robotic, laparoscopic, and open approaches.

Methods: We performed a retrospective review of patients undergoing abdominal wall herniorrhaphy at a tertiary care center from 2013 through 2019. Patients were stratified by approach: laparoscopic (LH), open (OH), or robotic (RH). Insurance type was categorized as private, Medicare, or Medicaid/uninsured. Using zip code data, we obtained a Distressed Communities Index (DCI), which is comprised of 7 unique socioeconomic variables. We employed random forest (RF) modeling to predict surgical approach and determined each factor's variable importance (VI) for our model.

Results: There were 559 patients; 39.7% (n = 222) LH, 33.3% (n = 186) OH, and 27% (n = 151) RH. The DCI (p < 0.01) and rates of poverty (p = 0.01), adults without diplomas (p < 0.01), and unemployment (p < 0.01) were highest in the OH group while job growth (p = 0.02) and median income ratio (p < .01) were highest in the RH group. The LH group had a greater proportion of privately insured patients than Medicaid/ uninsured patients (43.4% vs 15.9%, p < 0.01). The most important variables identified by our RF model were job growth (for RH), insurance type (for LH), and no high school diploma (for OH).

Conclusion: Insurance type, job growth, and educational attainment may influence operative approach and can contribute to the existing disparities in hernia surgery. Surgeons should address these inequalities and commit to parity in the delivery of surgical care.

Keywords: Distressed communities index; Hernia repair; Insurance; Random forest; Socioeconomic disparity.