Using ADI to Predict Emergency Department Overutilization in Post-Operative Epilepsy Patients

Introduction: Various studies have shown that racial and SDOH (Social Determinants of Health) disparities affect access to epilepsy physicians, medications, and surgical recommendations. Less is known about how these factors affect the post-operative course. The Area of Deprivation Index (ADI) is a numeric score that uses neighborhood-level data based on income, education, housing, and employment data. This score can be used to compare levels of disadvantage across the United States.

Methods: Patient data was obtained from a retrospective database consisting of epilepsy patients who have been treated at the Cleveland Clinic from 1997 to 2023. Emergency room (ED) use and readmission data were collected from EHR (Electronic Health Record) querying. ADI percentiles were calculated using GEOIDs closest to the date of surgery and mapped to the Wisconsin Neighborhood Atlas 2021 file. ADI was binned into quintiles as has been established in the literature. Comparison tests included ANOVA (Analysis of Variance), chi-square, Kruskal-Wallis, and Fisher’s exact.

Results: Our cohort of 3,663 was comprised of 1,896 (51.8%) males, with a mean age of 26.7 ± 17. A total of 3,402 (92.9%) were Caucasian, and the most common surgery was temporal lobectomy (45.9%). The median ADI was 65 (IQR = 38), and 331 patients did not have a present ADI value and were excluded from the analysis. Quintile 5 (most disadvantaged) had the highest proportion of Black/African Americans and the oldest population. 145 patients had at least 1 ED visit, and of these, 66 were readmitted from the ED within 24 hours. Logistic regression modeling identified the presence of medical comorbidities as predictive of ED use and did not find any factors predictive of ED readmission.

Conclusion : The results of this study currently show that patient disadvantage status is not associated with ED overutilization. Further studies will investigate how the relationships are affected when comorbidities are included in the regression modeling.