Intracerebral hemorrhage: positive predictive value of diagnosis codes in two nationwide Danish registries

Clin Epidemiol. 2018 Aug 7;10:941-948. doi: 10.2147/CLEP.S167576. eCollection 2018.


Purpose: The purpose of this study is to establish the validity of intracerebral hemorrhage (ICH) diagnoses in the Danish Stroke Registry (DSR) and the Danish National Patient Registry (DNPR).

Patients and methods: We estimated the positive predictive value (PPV) of ICH diagnoses for a sample of 500 patients from the DSR (patients recorded under ICH diagnosis) and DNPR (International Classification of Diseases, version 10, code I61) during 2010-2015, using discharge summaries and brain imaging reports (minimal data). We estimated PPVs for any ICH (a-ICH) and spontaneous ICH (s-ICH) alone. Furthermore, we assessed PPVs according to whether patients were recorded in both or only one of the registries. Finally, in a subsample with ICH diagnoses with access to full medical records and original imaging studies (extensive data, n=100), we compared s-ICH diagnosis and hemorrhage location after use of extensive vs minimal data.

Results: In the DSR, the PPVs were 94% (95% CI, 91%-96%) for a-ICH and 85% (95% CI, 81%-88%) for s-ICH. In the DNPR, the PPVs were 88% (95% CI, 84%-91%) for a-ICH and 75% (95% CI, 70%-79%) for s-ICH. PPVs for s-ICH for patients recorded in both registries, DSR only, and DNPR only were 86% (95% CI, 82-99), 80% (95%CI, 71-87), and 49% (95%CI, 39-59), respectively. Evaluation of extensive vs minimal data verified s-ICH diagnosis in 98% and hemorrhage location in 94%.

Conclusion: The validity of a-ICH diagnoses in DSR and DNPR is sufficiently high to support their use in epidemiologic studies. For s-ICH, validity was high in DSR. In DNPR, s-ICH validity was lower, markedly so for the small subgroup of patients only recorded in this registry. Minimal data including discharge summaries and brain imaging reports were feasible and valid for identifying ICH location.

Keywords: epidemiology; intracranial hemorrhage; register-based research; stroke.