Identifying Populations and Neighborhoods at High Risk for Hospital Admission-Driven Radiotherapy Interruption Using Geospatial Analytics

Abstract

Hospital admission driven radiotherapy interruption (HADRI) remains understudied despite its association with increased costs and inferior cancer outcomes. In this study we benchmark HADRI for patients receiving external beam radiotherapy (RT), identify predictors for HADRI, and geographically map HADRI at the neighborhood level across a large region served by a single academic referral center. Data was collected for all patients treated with RT at our center from January 1, 2015 to December 31, 2017. HADRI was defined as postponement of at least one RT treatment due to either hospital inpatient or emergency room admission. We examined differences in the proportion of patients with HADRI by cancer type, race, insurance, and patient predicted income (PPI) which was modeled using 2017 US Census data by patient residence census tract and categorized as low (<$34k), middle, or high (>$67k). Proportion of patients with HADRI were compared using Pearson’s Chi square testing and mapped by patient residence at the neighborhood (census tract) level. 82,773 total fractions (median 24, IQR 10-30) were delivered to 3729 patients with 7107 interruptions (8.6%) in 1928 patients (51.7%). 2022 (54.5 %) were Caucasian, 1577 (42.3%) were African American, and 120 (3.2%) were other races. Insurance status was defined as Commercial, Medicare, or Medicaid/Uninsured in 1794 (48.1%), 1503 (40.3%), and 432 (11.6%) patients, respectively. There were 727 interruptions (10.2%) due to hospital admission in 197 (5.3%) patients. Statistically significant elevations in HADRI were seen in patients treated for HN, lung, metastatic, and GYN cancers (19.9%, 18.5%, 10.3% and 8.9% of patients with each tumor type, respectively; p = <0.0001). Elevations in HADRI were seen between Medicaid/Uninsured patients v. those with Commercial or Medicare insurance (7.4% v. 3.5% p = <0.0001; OR 2.21 [95%CI 1.48-3.32]), African American v. Caucasian patients (5.1 % v 3.1% p = 0.002; OR 1.66 [95%CI 1.19-2.32]), and low PPI v. high PPI patients (5.5% v 2.4% p = 0.0002; OR 2.31 [95%CI 1.48-3.62]). The most common clinical causes for HADRI were: dehydration (22.3%); acute respiratory failure (13.6%); fever/sepsis (11.4%); renal disease (11.4%); radiation induced mucositis or dermatitis (9.1%); and PEG tube complication (6.8%). Spatial analysis revealed elevated rates of HADRI in low income, majority African American neighborhoods (p = 0.0002). Significant elevation in hospital admission driven RT interruption was seen in low income, racial minority, Medicaid/uninsured populations and patients with cancer sites associated with high morbidity. Geospatial analytics revealed neighborhood hotspots eligible for targeted intervention. Further investigation of hospital admission driven RT interruption could prevent treatment interruption and reduce costs in cancer care delivery in the era of value-based care.