(116) Feasibility of a Federated Network Analysis Using Real-World Data Mapped to OMOP Common Data Model to Estimate Healthcare Resource Utilisation and Costs of Imminent Subsequent Fracture

Background: Fragility fractures increase healthcare use but their impact is poorly understood. Real-world data (RWD) mapped to the OMOP-CDM enables access to standardised international health records, however feasibility of capturing data required to analyse resource use and costs remains underexplored.

Objectives: To assess the suitability of OMOP-mapped RWD from six European countries to estimate the healthcare resource use and costs of imminent subsequent fractures in postmenopausal women with fragility fractures in primary and secondary care.

Methods: Data from primary care databases from the United Kingdom (CPRD), Germany (IQVIA DA), the Netherlands (IPCI), France (LPD), Italy (LPD) and Spain (SIDIAP) mapped to the OMOP-CDM were used. Primary care resource use was characterised based on reported encounters, staff roles, and types of visits as available. Consultation with data providers was needed to clarify the nature and limitations of the source data. Local unit costs were applied. Where databases included secondary care data (CPRD, SIDIAP), hospitalisations were characterised by diagnoses, procedures and length of stay, and costs estimated using bespoke simplified local HRG/DRG algorithms. A core R script was written and adapted to each dataset before deployment.

Results: Granularity of resource use variables varied by dataset. Primary care resource use and costs were successfully analysed for all databases, reporting number of visits and costs per person per year for each cohort. Usage and cost by staff role (e.g. GP, nurse) or type of visit (e.g. clinic, home) were available for some datasets. In CPRD, we used values as recorded in the source data to identify staff roles. In IQVIA DA, France LPD and SIDIAP, source values were not usable hence standardised specialty concept IDs were extracted. Italy LPD categorised all encounters under a single specialty, whereas in IPCI no roles were identifiable. Visit type was available in France LPD, SIDIAP and IPCI. For secondary care, CPRD reported mapped diagnostic and procedure codes for all hospitalisations, whilst in SIDIAP only primary diagnosis and up to 10 procedures were available. These were reported for each cohort and costs estimated based on simplified algorithms and unit costs.

Conclusions: This study demonstrates the feasibility of performing health economics analyses in the OMOP-CDM environment using electronic health care records from administrative datasets. We tested this successfully by estimating healthcare resource use and costs of imminent subsequent fractures. Such analyses appear to require database-specific script given differences in source data.