Cancer ontology understanding among community oncologists: Insights from a survey linked to implementation of mCODE-informed electronic health record (EHR)

e18860 Background: Real-world data (RWD) will help recognize the gaps between evidence and practice, re-shaping clinical assessments. Data collected at the source in oncology-tailored EHRs are expected to facilitate generation of high-quality RWD. In the process of implementing a new EHR informed by mCODE standards in our network of oncology clinics, we defined mandatory variables amenable to structured data capture. However, medical knowledge about cancer ontology has not been methodically investigated. Methods: Online survey with 4 multiple-choice questions that illustrate in clinical cases the most important concepts and terminologies related to cancer diagnosis, treatment, and patient outcome. The questionnaire was shared with ~ 300 medical oncologists from our institution through mobile messaging app in December 2022. Physicians signed electronic informed consent for anonymized data analysis. Results: In total, 203 oncologists agreed to participate and 152 completed the survey. Most oncologists (75%) correctly define “therapy intent” (palliative versus curative) and corresponding “treatment line”. Many (51%) do not understand that “disease stage” is static (defined at the time of diagnosis) and does not change to stage IV if patient presents with relapsed/recurrent disease. A significant proportion of oncologists (66%) wrongly define “response status” to a given therapy as stable/progressive disease in a patient that had major clinical/radiological response but presents with unconfirmed tumor marker increase and no clear change in bone metastases by imaging. However, most oncologists (88%) can correctly define “reason for therapy discontinuation” as “planned interruption” in the end of adjuvant chemo-immunotherapy. When stratifying physicians according to self-defined knowledge of disease ontology, we found that 27% of those with moderate/deep understanding of mCODE scored 4 correct answers in the questionnaire as compared to 8% with 4 correct answers if mCODE knowledge was null or vague (P value < 0.001, odds ratio 5.1, CI95% 1.8-15.8). Conclusions: Overall understanding of cancer-specific data dictionaries is low among community oncologists. With the new EHR system we are nudging physicians not to update the “stage at diagnosis” status at the time of disease relapse/recurrence by creating a “current cancer status” element with mandatory completion (linked to prescriptions/ pharmacy claims). Also, “response status” as assessed by physicians must be cross-checked with other data sources (tumor markers, imaging results, physician notes and reason for therapy discontinuation) given inconsistent definition of “clinical benefit versus progressive disease” among physicians. In addition, we will constantly educate oncologists and monitor their EHR data entry behavior for successful execution of RWD projects.