Potential serious bias in National Clinical Databases with low degree of reported follow-up

We identified reasons for the low follow-up rate in the Danish Knee ligament Reconstruction Register (DKRR) and evaluated its influence on the data quality. All 946 primary ACL-reconstructed patients in the Capital Region of Denmark during 2012 were identified in the databases of 8 participating hospitals. We studied the patient files and compared them to figures reported to the DKRR. 92.5% of the operated patients was registered in DKRR. The 1-year follow-up rate reported to DKRR was 33.4%, and 14.5% filled in patient reported outcomes (KOOS and Tegner) at 1 year. Only 65% had actually been invited for follow-up, but among the patients who had been invited 91% were seen. 41% of existing follow-up data was not reported. Contemporary technology and structured motivation should be introduced to increase validity of data in national clinical databases. Follow-up >90% in the DKRR is realistic if patents are invited and reported. The unreported data is potentially a serious bias. It is suggested that data from clinics with low follow-up should not be used in studies involving outcomes based on national databases because of risk of bias.