Substantial accuracy of fracture classification in the Swedish Fracture Register: Evaluation of AO/OTA-classification in 152 ankle fractures.

INTRODUCTION Since 2011 the Swedish Fracture Register (SFR) has collected information on fracture epidemiology, treatment and outcome. More than 112 000 fractures, including 11 600 ankle fractures, have been registered so far. The accuracy of the classification is crucial when using the register in future research and quality improvement of care. In this study we examine the degree of accuracy of the AO/OTA-classification in which ankle fractures are recorded in the SFR. METHODS 152 randomly selected ankle fractures registered in the SFR were classified independently by a reference group of three orthopedic surgeons on two occasions. The agreed AO/OTA-classification of the reference group was regarded as the gold standard classification for each case. The originally recorded classification in the SFR was subsequently compared with the classification of the reference group and accuracy calculated. RESULTS The agreement between the classification in the SFR and of the reference group was 74%, corresponding to kappa 0.67, 95% CI (0.58-0.76) for AO/OTA group level. An agreement of 88% equivalent to kappa 0.77, 95% CI (0.67-0.87) was noted for AO/OTA type level. The agreement corresponds to substantial according to Landis and Koch. The kappa value of inter- and intraobserver agreement ranged from 0.67 to 0.93. CONCLUSIONS The study results show substantial agreement between classifications made in the SFR and gold standard classification. The finding is equivalent to or higher than in previous studies. Consequently, we conclude that classifications of ankle fractures in the SFR are accurate and valid.