Abstract 6233: Registries in Coronary Artery Disease: Assessing or Biasing Real World Data?
As many registries lack quality control, we assessed the potential selection bias due to incomplete inclusion in MASCARA registry. This was a prospective registry of consecutive patients with acute coronary syndrome (ACS) in 50 Spanish hospitals, designed to assess the impact of guidelines on practice patterns and clinical outcomes in 2004–2005 in Spain. All eligible patients for MASCARA study who had not been prospectively included (missing patients) were retrospectively identified by cross-checking the MASCARA database with the hospital episode statistics obtained in 18 centers. The clinical records of the missing patients were carefully reviewed. In-hospital mortality and the prevalence of 22 prognostic variables in the missing patients (n=1.439) were compared with those in the not-missing patients (n=3.265). The potential selection bias was quantified estimating the association between the 22 variables and the in-hospital mortality by comparing the initial relative risks (RRs) (i.e. using only not-missing patients) with the actual final RRs (i.e. after including missing patients). The prevalence of the 22 prognostic variables consistently differed between the missing and not-missing population, and 19 reached statistical significance. In general, the missing population had greater risk and received less treatment strategies than the not-missing patients. In-hospital mortality was 3 times higher in the missing population [ST-elevation ACS: 13.67% (95%CI 10.9–16.8) vs. 4.56% (3.1–6.8); non-ST-elevation ACS: 6.57% (4.8–8.3) vs. 2.65% (1.6–3.9)]. Initial RRs were substantially different from the final actual RRs (>10%) in 12 variables (relative difference 11.13%–33.95%): in 10 variables the initial RRs had underestimated and in 2 had overestimated the actual RRs. In prospective registries, missing patients have greater risk than the included population. This can lead to a serious underestimation of actual mortality. Therefore, results of registries without quality control of inclusion should be cautiously considered.