EP763 - Transportation and Internet Access are Critical for Improving Quality of Care and Outcomes for Diverticulitis Patients

Purpose/Background: Optimal outcomes for patients with diverticulitis are dependent on reliable access to high quality care including use of mechanical bowel prep (MBP) with preoperative oral antibiotics (POA) for elective colectomies. Unwarranted variation in colorectal outcomes, however, exist by social, economic, structural, and environmental risk, all of which influence access to high quality health care. Specific mutable factors that can be targeted need to be identified. This study uses machine learning (ML) to evaluate a risk-adjusted outcomes registry linked to a validated socioeconomic (SES) predictive algorithm to elucidate mutable factors of social determinants of health.

Methods/Interventions: Addresses of elective colectomy patients diagnosed with diverticulitis treated at U.S. National Surgical Quality Improvement Program hospitals between January 1st, 2018 and June 30th, 2023 were geocoded to the block group level. Outcome-specific ML based logits were constructed and validated using 17 U.S. Census Bureau data components representing aspects of income, education, employment, housing, and household characteristics. Multivariable logistic regression was performed to evaluate the influence of ML logit values on receipt of MBP and POA, any major complication, and surgical site infection within 30-days of surgery. Mean Shapley values were calculated for each of the SES factors for each outcome to evaluate components’ relative contribution.

Results/Outcomes: Of the 29992 patients, 69% received MBP and 66% received POA. After adjusting for age, sex, race, body mass index, preoperative sepsis, comorbidities, American Society of Anesthesiologist class, and case complexity, higher SES was associated with use of MBP and POA. After adjusting for all covariates including MBP and POA, lower SES was associated with increased risk for any major complication and surgical site infection. Of the top contributing SES factors, potentially mutable factors included lack of access to a motor vehicle and to internet within a household.

Conclusion/Discussion: Addressing structural factors that impact patients from neighborhoods with lower SES should be considered as a key strategy for addressing colorectal disparities. Our results suggest that lack of access to transportation and information (i.e., internet), both of which can be targeted with interventions, influence access to best practices and optimal care for patients undergoing elective colectomies for diverticulitis. Knowing how each SES factor uniquely contributes to outcomes allows for more effective strategies to be deployed to improve quality of care.

Table. Validation results of the machine learning (ML) based SES predictive algorithm (model discrimination [c-statistic], model calibration [Hosmer-Lemeshow statistic], calibration P-values), and risk-adjusted odds ratios for the ML based continuous logit variables.
a Morbidity – a composite variable of complications not present at time of surgery including surgical site infection, wound disruption, pneumonia, unplanned intubation, postoperative ventilator dependence for > 48 hours, postoperative dialysis use, postoperative renal insufficiency, urinary tract infection, stroke, cardiac arrest, myocardial infarction, and sepsis or septic shock.
b ML logits were found, by examination of mean Shapley values, to be directly related to census components indicating lower socioeconomic status.
c ML logits were found, by examination of mean Shapley values, to be directly related to census components indicating higher socioeconomic status.

Figure. Heat map of mean Shapley values representing the relative contribution of each SES factor for each outcome. Top 7 contributing factors are bolded.
SES socioeconomic status; SSI surgical site infection; FPL Federal Poverty Line