End-of-life spending analysis of randomized trial of machine learning nudges to prompt serious illness communication among patients with cancer

6515 Background: Early serious illness conversations (SICs) between oncology clinicians and patients in the outpatient setting may improve mood and quality-of-life among patients with cancer. However, the impact of early SIC “nudges” on end-of-life spending is unknown. Methods: This was a secondary analysis of a stepped-wedge randomized trial (NCT03984773) that randomized 9 medical oncology practices and their high-risk patients at a large academic institution to a behavioral intervention to increase SICs (performance reports and peer comparisons; precommitments for high-risk patients; weekly opt-out text prompts before high-risk encounters) vs. standard of care, between June 2019 and April 2020. We identified high-risk patients using a machine learning (ML) algorithm predicting 180-day mortality. This secondary analysis included 1187 (957 intervention, 230 control) patients with complete data who died by December 2020. We abstracted spending (defined as inflation-adjusted reimbursements for acute care [inpatient + ED], office/outpatient care, intravenous chemotherapy, other therapy [e.g. radiation], long-term care, and hospice) from the institution’s accounting system; we captured spending at University of Pennsylvania inpatient, outpatient, and pharmacy settings. To evaluate intervention impacts on spending, we used a two-part model: first, logistic regression to model zero versus nonzero spending, and second, generalized linear models with gamma distribution and log-link function to model daily mean spending in the last 180 days of life. Models were adjusted for clinic and wedge fixed effects and clustered at the oncologist level. Results: Median age at death was 68 years (IQR 15.5), 317 (27%) patients were non-White, and 448 (38%) patients had a SIC prior to death. The intervention was associated with lower mean spending in the last 180 days of life (mean daily spending $377.96 [intervention] vs. $449.92 [control]; adjusted mean difference -$75.33, 95% CI -$136.42, -$14.23 , p = 0.016) (Table), translating to $13,559 total adjusted savings. Intervention patients incurred lower mean daily spending for chemotherapy (adjusted difference -$51.35, p < 0.001), office/outpatient care (-$14.59, p < 0.001), and other therapy (-$10.35, p = 0.043). The intervention was not associated with differences in end-of-life spending for acute care utilization, long-term care, and hospice. Results were consistent for spending in the last 30 and 90 days of life and after adjusting for age, race, and ethnicity. Conclusions: A ML intervention to prompt SICs led to end-of-life savings, driven by decreased chemotherapy and outpatient spending. Clinical trial information: NCT03984773 . [Table: see text]