Presentation (CTI)

OBJECTIVES: To assess whether the main drivers of healthcare costs associated with fragility fractures estimated from CPRD-HES data mapped to the OMOP Common Data Model are comparable to those identified from the source datasets.
METHODS: We conducted a retrospective cohort study of women aged >50 years with a first fragility fracture. Analyses were performed using source CPRD-HES data from England and their OMOP-mapped version based on two strategies: OMOP-standard (standardised concepts only) and OMOP-source (including retained source variables). Cost drivers were analysed for total healthcare costs and separately for primary care, inpatient, outpatient, and emergency care costs. Outcomes with <5% zero costs were analysed using generalised linear models with Gamma distribution and log link; otherwise, they were analysed using two-part models comprising logistic regression for healthcare utilisation and Gamma GLMs among users only. Models were adjusted for fracture site, ethnicity, age at fracture, cardiovascular disease, heart failure, rheumatoid arthritis, and osteoporosis diagnosed within the previous two years. Average marginal effects were estimated for total costs from the Gamma model using counterfactual predictions across the full sample.
RESULTS: The study included 23,106 patients in CPRD-HES and 22,900 in OMOP-CDM. Patient characteristics were highly comparable across datasets. The main cost drivers identified in CPRD-HES were largely reproduced in both OMOP analyses. Non-hip fractures were associated with ~£8,000-£12,000 lower annual costs per patient than hip fractures, whereas cardiovascular disease, rheumatoid arthritis, and a 10-year increase in age were each associated with ~£1,500-£2,700 higher annual costs per patient-year. Utilisation odds ratios, cost ratios, and average marginal effects were highly concordant between CPRD-HES and OMOP-source analyses, although OMOP-standard showed attenuated associations.
CONCLUSIONS: The main drivers of fragility fracture-related healthcare costs and their marginal effects were highly comparable between OMOP-mapped and CPRD-HES source data, supporting the use of OMOP-CDM for health economic research.