Inter-rater reliability of a national acute stroke register

Torunn Varmdal, Hanne Ellekjær, Hild Fjærtoft, Bent Indredavik, Stian Lydersen, Kaare Harald Bonaa

Medical quality registers are useful sources of knowledge about diseases and the health services. However, there are challenges in obtaining valid and reliable data. This study aims to assess the reliability in a national medical quality register. We randomly selected 111 patients having had a stroke in 2012. An experienced stroke nurse completed the Norwegian Stroke Register paper forms for all 111 patients by review of the medical records. We then extracted all registered data on the same patients from the Norwegian Stroke Register and calculated Cohen's kappa and Gwet's AC1 with 95 % confidence intervals for 51 nominal variables and Cohen's quadratic weighted kappa and Gwet's AC2 for three ordinal variables. For two time variables, we calculated the Intraclass Correlation Coefficient. Substantial to excellent reliability (kappa > 0.60/AC1 > 0.80) was observed for most variables related to past medical history, functional status, stroke subtype and discharge destination. Although excellent reliability was observed for time of stroke onset (ICC 0.93), this variable was hampered with a substantial amount of missing values. Some variables related to treatment and examinations in hospital displayed low levels of agreement. This applies to heart rate monitoring (kappa 0.17/AC1 0.46), swallowing test performed (kappa 0.19/AC1 0.27) and mobilized out of bed within 24 h after admission (kappa 0.04/AC1 -0.11). A majority of the variables in The Norwegian Stroke Register have substantial to excellent reliability. The problem areas seem to be the lack of completeness in the time variable indicating stroke onset and poor reliability in some variables concerning examinations and treatment received in hospital.