Demonstrating the data integrity of routinely collected healthcare systems data for clinical trials (DEDICaTe): A proof-of-concept study

Abstract

Introduction/aims: Healthcare systems data (also known as real-world or routinely collected health data) could transform the conduct of clinical trials. Demonstrating integrity and provenance of these data is critical for clinical trials, to enable their use where appropriate and avoid duplication using scarce trial resources. Building on previous work, this proof-of-concept study used a data intelligence tool, the "Central Metastore," to provide metadata and lineage information of nationally held data. Methods: The feasibility of NHS England's Central Metastore to capture detailed records of the origins, processes, and methods that produce four datasets was assessed. These were England's Hospital Episode Statistics (Admitted Patient Care, Outpatients, Critical Care) and the Civil Registration of Deaths (England and Wales). The process comprised: information gathering; information ingestion using the tool; and auto-generation of lineage diagrams/content to show data integrity. A guidance document to standardise this process was developed. Results/Discussion: The tool can ingest, store and display data provenance in sufficient detail to support trust and transparency in using these datasets for trials. The slowest step was information gathering from multiple sources, so consistency in record-keeping is essential.

Keywords: clinical trials; data integrity; data provenance; data quality; healthcare systems data; metadata; routinely collected health data.

Figure 1. The elements of the metadata model captured in NHS England’s Central Metastore, reproduced from the study’s operating manual/guidance.^,

Figure 2

Summary business lineage view of the Admitted Patient Care (APC), Outpatients (OP), and Critical Care (CC) datasets of Hospital Episode Statistics (HES). The data journey from left (submission from hospitals using business rules via an XML schema) to right (production stage where derivations and processing rules are applied) to form the final releasable HES schema containing APC, OP and CC tables in the data access environment (far right).

Figure 3

Field level lineage of the Civil Registration of Deaths (CRD) with an example of derivation to confirm NHS number. The data journey moves from left (submission via Message Exchange for Social Care and Health) to right (production stage, then to releasable in the data access environment).

Figure 4. Example of data item details captured in the Central Metastore at NHS England (diagnosis code, Hospital Episode Statistics Admitted Patient Care).