Quality of ethnicity coding and missed opportunities among cancer patients with acute coronary syndrome: a national multi-registry cohort study

Abstract Background Ethnic inequalities exist in the management and outcomes of patients with acute coronary syndrome (ACS) and cancer. These patients are under-represented in clinical trials and, therefore, studied using large national registries and electronic health records (EHR). However, the quality and consistency of ethnicity reporting across these datasets remain unclear. Aim The present study examined the agreement of ethnicity coding and subsequent outcomes for all cancer patients with ACS across four national datasets, namely the National Cancer Registration and Analysis Service (NCRAS), the Myocardial Ischaemia National Audit Project (MINAP), the National Audit of Percutaneous Coronary Intervention (NAPCI), and Hospital Episode Statistics (HES; hospitalisation data). Methods All cancer patients (n=60,793) with an ACS diagnosis between 2010 and 2018 were retrospectively analysed. Three linkages were performed: NCRAS-MINAP, NCRAS-MINAP-NAPCI, and NCRAS-MINAP-HES. Based on agreement in ethnicity coding, three possible categories were created for each linkage: Concordant (same across all datasets), Discordant (different in ≥1 dataset), and Missing (in any dataset). Multivariable logistic regression models were used to examine adjusted odds of 1-year cardiac and cancer-related death for each agreement group. Results Among the NCRAS-MINAP, NCRAS-MINAP-HES, and NCRAS-MINAP-NAPCI linked cohort, more than half of patients’ ethnicities were concordant across datasets (53.8%, 53.6%, and 52.4%, respectively). The highest discordance rate was observed in NCRAS-MINAP-HES (40.8%), while the highest missingness rate was in NCRAS-MINAP (45%).(Figure 1). There was moderate agreement at most between datasets for ethnicity coding, the highest being NCRAS-HES (r=0.714), and the lowest was NCRAS-NAPCI (r=0.318).(Figure 2) The odds of 1-year cardiac death were significantly increased in the Missing group, compared with the Concordant group, across all three linkages (NCRAS-MINAP: odds ratio (OR) 1.09, 95%CI 1.02-1.15; NCRAS-MINAP-NAPCI: OR 1.69, 1.12-2.54; NCRAS-MINAP-HES: OR 1.44, 1.14-1.82). Odds of 1-year cancer death were increased in the Missing (OR 1.29, 1.23-1.36) and Discordant (OR 1.14, 1.02-1.27) groups in NCRAS-MINAP and NCRAS-MINAP-NAPCI linked datasets, respectively. Conclusion The are significant variations in agreement and completeness of ethnicity coding across four linked national datasets for cancer patients with ACS. Nearly half of the cancer patients with ACS had discordant or missing ethnicity data, even when linked to hospitalisation and PCI datasets. Our findings, while observational, demonstrate worse cardiac and cancer-related mortality among patients with missing or discordant ethnicity, which could be a sign of underlying disparities in their care. Inconsistency in ethnicity coding represents a missed opportunity to study patterns and contributors to health inequalities in this high-risk population with cancer and ACS.Agreement in linked datasets Correlation of ethnicity across datasets