Abstract PR01: Implementation of a real-time AI-guided navigation service for pancreas cancer
Kristen M. John,Joseph Tenner,Yuddy Franco,Rolando Croocks,Melissa Perez,Tara McEvoy,Tiffany Zavadsky,Kristen Beyer,Rita Mercieca,Sandeep Nadella,Anthony Carvino,Matthew Barish,Daniel A. King
DOI: https://doi.org/10.1158/1538-7445.panca2023-pr01
IF: 11.2
2024-01-18
Cancer Research
Abstract:Patient navigation aims to overcome access barriers and optimize care delivery, but navigators cannot identify pre-diagnosed patients using traditional methods, hindering navigation efficacy. We observed care gaps among patients identified at our institution with suspected pancreatic cancer during January 2023: among 67 patients with new pancreas cancer, only 36% underwent biopsy with 22 days between radiology and biopsy; and 30% were seen by an outpatient oncologist with 32 days between radiology and visit. We sought to improve patient care outcomes through daily prospective AI-guided identification and navigation of patients with radiographic findings suspicious for pancreas cancer. In June 2023, we implemented an AI-guided daily workflow using the following steps: 1) at 24-hour intervals, an NLP model reviewed abdominal imaging reports and flagged reports containing language suspicious for pancreatic cancer; 2) a trained coordinator and GI oncologist validated reports with new masses and sent such reports to a GI navigator, and 3) the navigator facilitated appointments for proper follow-up, including GI services, the multi-disciplinary tumor board, and clinical trial coordinator. During June 2023, the model identified 69 patients with suspicion for new pancreas cancer. 19% were immediately eligible for navigation and scheduled for follow-up within our institution with an average of 8 days between report and navigation - 77% saw an outpatient oncologist with an average of 17.5 days between report and visit, 69% underwent biopsy with an average of 4.5 days between report and biopsy, and 54% were seen at multidisciplinary pancreatic cancer tumor board (pcTB) with an average of 16 days between report and pcTB. Furthermore, of these, we were able to enroll 4 patients to clinical trials and 5 accruals to biospecimen studies during this one-month period. 2 of these accruals were to the P-1000 study, which has been open at our institution for the last 2 years with a 0.5 per month accrual rate. Of the 81% who were not immediately eligible for navigation, 23% were already established with an oncologist in our system, 23% sought care elsewhere, 23% were not eligible for or did not desire further workup, and 30% were still undergoing in-patient workup. We show that an AI-guided workflow can transform referral patterns of pre-diagnosed pancreatic cancer patients, creating a new access stream for navigators to intervene much earlier in a patient's continuum of cancer care with resultant improvement in healthcare delivery, evidenced by an average of 4.5 days between imaging and biopsy and 17.5 days between imaging and outpatient oncology visit, compared to our traditional cancer care approach with averages of 22 and 32 days, respectively. Furthermore, we have rapidly accelerated pancreas cancer clinical trial enrollment and biospecimen accrual at our institution since implementation of this workflow with 4 patients enrolled and 5 biospecimen accruals in a month alone. We are now working to expand this innovative framework across additional cancer types. Citation Format: Kristen M. John, Joseph Tenner, Yuddy Franco, Rolando Croocks, Melissa Perez, Tara McEvoy, Tiffany Zavadsky, Kristen Beyer, Rita Mercieca, Sandeep Nadella, Anthony Carvino, Matthew Barish, Daniel A. King. Implementation of a real-time AI-guided navigation service for pancreas cancer [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Pancreatic Cancer; 2023 Sep 27-30; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(2 Suppl) nr PR01.
oncology