(P-193) Quality and Completeness of Diagnostic Codes for Myocardial Infarction in CPRD Aurum

Background: The Clinical Practice Research Datalink (CPRD) now provides data from CPRD Aurum, a new medical record database. Understanding the validity and completeness of diagnoses recorded in any new data source is critical for making decisions about suitability and validity for medical research purposes.

Objectives: To examine data source agreement describing myocardial infarction (MI) diagnoses recorded in CPRD Aurum compared with linked Hospital Episode Statistics (HES) APC data . This comparison provides information on CPRD Aurum data ‘correctness’ (e.g. accuracy, validity) and ‘completeness’ (e.g. presence, missingness).

Methods: Patients with acute MI diagnoses recorded in either data source were selected from a random sample of 50,000 patients in CPRD Aurum with HES linkage (1997-2017). We calculated the proportion of acute MI cases recorded in CPRD Aurum who also had a MI diagnosis recorded in HES (correctness). Among all acute MI diagnoses recorded in HES, we calculated the proportion also present in CPRD Aurum (completeness). Lastly, among MI diagnoses recorded in only one data source, we describe the likelihood that the MI was a true case based on supporting codes in either data source.

Results: There were 1260 patients with acute MI recorded in CPRD Aurum. Of these, 986 (78%) were confirmed in HES within 90 days and another 123 (10%) were confirmed in HES but with an unconfirmed date (e.g. HES code for “old myocardial infarction” at a later date); the overall correctness was 88%. 80 (6%) CPRD Aurum-coded MIs not found in HES had weak or no other evidence of MI in either data source. Some of these could have been rule-out diagnoses or miscodes in CPRD Aurum. A further 35 (3%) CPRD Aurum-coded MIs not found in HES lacked evidence due to patient death soon after the MI. There were 1434 patients with acute MI recorded in HES. Of these, 965 (67%) were recorded in CPRD Aurum within 90 days and 9 (1%) with an incorrect date; overall completeness of 68%. Of 1082 MIs recorded in a primary HES diagnosis field, 813 (75%) were recorded in CPRD Aurum. 131 (9%) primary HES-coded MIs were not found in CPRD Aurum but had strong evidence of being a true MI. These are likely true MI cases missing from CPRD Aurum. Another 85 (6%) died after HES-coded MI but had no CPRD Aurum-coded MI. Many are likely true cases that could be confirmed with death registry data.

Conclusions: Findings suggest that the ‘correctness’ of MI diagnoses recorded in CPRD Aurum compared to HES is high. Researchers should consider study design choices (e.g. case control) or the addition of linked data (HES and death registry) to address incomplete capture of MI cases in CPRD Aurum.