IntroductionHealth care patient records have been digitalised the past twenty years, and registries have been automated. Missing registrations are common, and can result in selection bias. ObjectiveTo assess the prevalence and characteristics of missed registrations in a Dutch regional trauma registry. MethodsAn automatically generated trauma registry export was done for ten out of eleven hospitals in trauma region Southwest Netherlands, between June 1 and August 31, 2020. Second, lists were checked for being falsely flagged as ‘non-trauma’. Finally, a list was generated with trauma tick box flagged as ‘trauma’ but were not automatically in the export due to administrative errors. Automated and missed registration datasets were compared on patient characteristics and logistic regression models were run with random intercepts and missed registration as outcome variable on the complete dataset. ResultsA total of 2,230 automated registrations and 175 (7.3 %) missed registrations were included for the Dutch National Trauma Registry, ranging from 1 to 14 % between participating hospitals. Patients of the missed registration dataset had characteristics of a higher level of care, compared with patients of automated registrations. Level of trauma care (level II OR 0.464 95 % CI 0.328–0.666, p < 0.001; level III OR 0.179 95 % CI 0.092–0.325, p < 0.001), major trauma (OR 2.928 95 % CI 1.792–4.65, p < 0.001), ICU admission (OR 2.337 95 % CI 1.792–4.650, p < 0.001), and surgery (OR 1.871 95 % CI 1.371–2.570, p < 0.001) were potential predictors for missed registrations in multivariate logistic regression analysis. ConclusionMissed registrations occur frequently and the rate of missed registrations differs greatly between hospitals. Automated and missed registration datasets display differences related to patients requiring more intensive care, which held for the major trauma subset. Checking for missed registrations is time consuming, automated registration lists need a human touch for validation and to be complete.
Read full abstract