eP716 - Leveraging Real-World Data for Colorectal Cancer Care: Challenges and Opportunities

Purpose/Background: Real-world data (RWD), health data routinely collected by electronic health records, poses unique opportunities for research, quality improvement, and disease monitoring. However, RWD come with a multitude of challenges that include quality, completeness, generalizability, variability precluding aggregation across systems, and lack of data feedback to the clinical setting. We aimed to utilize a novel RWD clinical environment to define a colorectal cancer cohort, describe clinically meaningful data elements, and identify data quality issues for future quality improvement work.Methods/Interventions: Through the Learning Health Network, we queried Cerner Real World DataTM (CRWD), a national database that houses deidentified data from a diverse cross section of hospital networks within the United States. Patients with a diagnosis of colorectal cancer were extracted from 2014 to 2022 by their index date of diagnosis. Patient demographics including age, sex, and race were collected. Cancer subtypes were split into colon, anal, and rectal based on string selection within their diagnosis. Hospital networks were analyzed for contribution to cohort volume. Furthermore, data elements were analyzed for quality based on missingness and similarity.Results/Outcomes: We identified 99,618 patients at 132 hospital networks using the colorectal cancer ontology. Fifty-two percent of patients were male and 81% were White, 7.7% were Black, 2.4% were Asian, 7.3% were other, and 1.8% were unknown . The median age was 67 years with a range of 0-90 years. Specific subtypes represented in the database include cancer of the colon (n = 73,276), rectum (n = 27,873) and anus (n = 530). The largest hospital network accounted for 12,741 (13%) of patients. Ten hospital networks with the largest cohorts accounted for 50% of all patients diagnosed with colorectal cancer in this database. Birth date values were missing or inaccurate in at least 7.3% of patients and 5 patients had a negative age recorded. Additionally, patient birth dates were truncated within CRWD at age 90 to maintain patient anonymity. Therefore, patients who were approximately 81 years or older were not able to be validated for age due to first diagnosis date in 2014.Conclusion/

Discussion: Our study describes extracting a cohort of colorectal cancer patients based on disease ontology using a novel RWD set. CRWD may provide insights into patient data missed by other large cancer registries. Challenges remain in validating individual data elements and generalizability to the population. Future work involves comparisons with established cohorts of cancer patients from registries; developing methods for describing cancer outcomes and processes of care (e.g., treatment delivery); and describing cancer patterns in this national data set.