Quality and completeness of malignant cancer recording in United Kingdom Clinical Practice Research Datalink Aurum compared to Hospital Episode Statistics
Katrina Wilcox Hagberg, Catherine Vasilakis‐Scaramozza, Rebecca Persson, Eleanor Yelland, T. Harry Williams, Puja Myles, Susan S. Jick
Abstract
Background: Data source validation is necessary to inform suitability for use in medical research. The objective was to examine agreement of cancer diagnoses recorded in Clinical Practice Research Datalink (CPRD) Aurum compared with linked Hospital Episode Statistics (HES) data to provide information on CPRD Aurum data correctness (accuracy, validity) and completeness (presence, missingness). Methods: The source population was a 50,000 random sample of CPRD Aurum patients with HES linkage (1997–2017). Patients with cancer diagnoses recorded in either data source were selected for these analyses. Correctness was the proportion of patients with cancer recorded in CPRD Aurum with a concordant cancer diagnosis recorded in HES. Completeness was the proportion of patients with cancer recorded in HES with a concordant diagnosis in CPRD Aurum. Results: A total of 6,019 patients had a cancer diagnosis: 3,864 in CPRD Aurum, 5,545 in HES, and 3,390 in both. Correctness estimate was 87.7% and an additional 8.4% had supporting cancer codes recorded in CPRD Aurum. Completeness was 61.1% and an additional 22.7% had supporting cancer codes recorded in CPRD Aurum. Correctness and completeness estimates varied by cancer site and calendar time. Conclusions: Cancer diagnoses in CPRD Aurum were of relatively high correctness for use in medical research. Completeness varied by cancer type and calendar year and may not be sufficient for all research questions. Use of linked data may improve completeness.