APH 1997, 55, 231-238:

The reliability of register-based patient characteristics.

M. Van Den Akker, G.H.L.M. Franssen, F. Buntinx, J.F.M. Metsemakers, and J. Knottnerus

Keywords: bias, observer variation, registries

To examine the reliability of demographical patient characteristics as routinely registered by the general practitioner (GP) in the Registration Network Family Practices, the educational level and type of living arrangement were compared with responders' answers in a postal questionnaire (N= 3 745). The degree of agreement was calculated using weighted and unweighted kappa values. The agreement for the level of education was good with an unweighted kappa value of 0.61 and a weighted kappa value of 0. 75. The kappa value for type of living arrangement was also good with 0. 73, but was considerably lower (0.43) for responders who did report any changes in living arrangement during the previous five years than for those who did not.

lt is concluded that registration of patient characteristics by the GP was fairly reliable for the characteristics studied, but reliability gets worse in case of change. Therefore items that are subject to change, such as the type of living arrangement, should preferably not be used for patient selection and should be measured again when important for a particular study.