(SLA-063) Standardising Clinical Practice Research Datalink data linked to hospital admissions and cancer registry data in prostate cancer

Background: The Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) is increasingly used to transform observational health data to standard format to facilitate interoperability in a network study. Representation of routinely collected cancer data in observational research is complex and requires integration of histology, biomarkers, and treatment regimen information.

Objectives: 1. To document the process of transforming primary care data linked to hospital and cancer registry data
2. To characterise the transformed dataset

Methods: Male patients aged ≥ 18 years with a first prostate cancer diagnosis between 01 Jan 2010 to 31 Dec 2022 were selected from the Clinical Practice Research Datalink GOLD and Aurum, primary care databases in the UK, then linked to hospital admissions, mortality data from Office for National Statistics, and national cancer registry (CR) data (up to 31 Dec 2018). We included patients with prostate cancer diagnosis confirmed in CR, with ≥ 1 year of registration at a GP practice.
Cancer diagnoses in the form of [Histology]/[behaviour]-[Topography] were mapped to the CDM Condition by using ICDO3 vocabulary. Cancer modifiers (e.g., stage, grade) were mapped to the CDM Measurement. Cancer treatments were mapped to the CDM Procedure. All cancer diagnoses, modifiers, and treatments were additionally mapped to the CDM Episode and Episode_Event to represent the patient’s diagnosis and treatment trajectory. Source data field names and values were stored in the corresponding source reference variables in the CDM tables. We assessed data and mapping quality using Kahn’s framework, which applies validation and verification checks according to conformance, completeness, and plausibility. We described patient characteristics in terms of demographics and prevalence of comorbidities and comedications.

Results: The mapped dataset consists of 69,109 males; median age 71 years; 91% White, 4% Black, and 2% Asian. Prevalent comorbidities and medications were hypertension (52%), low back pain (30%), erectile dysfunction (20%); and lipid lowering agents (46%), drugs for acid-related disorders (34%), and for benign prostate hyperplasia (24%), respectively. There were 52,933 patients with cancer staging information recorded as best staging assessed by the cancer registry, that could be mapped to OMOP CDM. Of these, most had Stage 1 prostate cancer (34%), with similar proportions across Stages 2 (21%), 3 (23%) and 4 (22%). A total of 4,684 data quality checks were performed, of which 98% passed overall, with conformance (99%), completeness (99%), and plausibility (96%) satisfied quality assessment.

Conclusions: The linkage of multiple datasets provides complementary information for observational research in cancer patients such as longitudinal comorbid conditions and tumour characteristics. Standardising these data to a CDM satisfied quality assessment.