Oncoshare-Lung: Novel three-way linkage of neighboring academic and community medical centers to state cancer registry for lung cancer.

Abstract

e23292 Background: Clinical research conducted at academic medical centers is often limited by fragmented patient history, which can be enhanced by integrating data from neighboring community centers to obtain a more consistent, continuous record of care. Such electronic health records (EHR), if further linked to registry data, provide both patient-level clinical data and validated long-term outcomes. We demonstrate the replicable process of a novel EHR integration for lung cancer (LC) between Stanford Health Care (SHC), an academic medical center, and Sutter Health, a nearby multi-site community practice, linked together to the California Cancer Registry (CCR). Methods: Patients with LC treated at SHC and Sutter Health were identified using the Stanford Cancer Registry (SCR) and the California Cancer Registry (CCR), respectively. Overlapping patients were identified using a hash code generated with patient information and verified manually. Using the combined Oncoshare-Lung data, descriptive analyses were performed for patient and cancer characteristics, smoking history, and presence of second malignancies, and compared to nation-level cohort from Surveillance, Epidemiology, and End Results (SEER). Overall survival (OS) was examined between the two cohorts using Kaplan-Meier analysis. Presence of recurrence was manually identified for a subset of 500 patients at SHC using radiology and clinical reports. Results: A total of 44,317 unique patients were identified, with 6,318 SHC-only, 37,257 Sutter-only, and 742 overlapping patients whose initial primary LC (IPLC) were diagnosed between 2000-2021, including 13.2% Asian, substantially higher than SEER (4.1%), 7.3% Hispanic, 6.6% non-Hispanic Black, and 68.7% non-Hispanic White. LCs were 25.2% Stage 1, 6.9% Stage 2, 17.6% Stage 3, 42.0% Stage 4, and 50.9% adenocarcinoma, 16.6% squamous cell carcinoma, 7.0% non-small cell carcinoma, NOS, 8.3% small cell carcinoma, 9.0% carcinoid, 7.0% other. Of 61.0% patients with known smoking history, 23.2% were never-smokers vs. 76.8% ever-smokers. Manual review of 500 SHC patients identified 25.2% with LC recurrence, serving as ground truth for building algorithms to automate the process. Further, 2.9% developed SPLC. The median OS from IPLC of patients in Oncoshare-Lung was 1.6 years [IQR: 0.3-2.9] vs. 0.8 years [IQR: 0.2–2.7] in SEER (P < 0.0001). Conclusions: Oncoshare-Lung provides a unique opportunity to address critical, unanswered clinical questions by complementing long-term outcomes in registry data (e.g., OS and second malignancies) with longitudinal clinical data in EHR (e.g., smoking history and recurrence). It is well situated to investigate many pressing issues including understanding high risk of LC among never-smoking Asian patients. Key future work include curating detailed treatment history and automating identification of recurrence in EHR.