Identifying clusters of raters with a common notion of diagnosing erosive tooth wear: a step towards improving the accuracy of diagnostic procedures.

BACKGROUND

Heterogeneous results are to be expected when multiple raters diagnose whether the dentine of a tooth with erosive tooth wear (ETW) is exposed or not. Identification of notions (fundamental concepts and understanding) about the diagnostic problem shared by groups of raters can be helpful to develop guidelines and to optimize teaching and calibration procedures. We aim to illustrate how clusters of raters with a common notion can be identified and how first insights about the notions can be obtained.

METHODS

This investigation is based on a former study in which 49 tooth surfaces affected by ETW were rated visually by 61 raters (23 scientists, 18 university dentists, 20 dental students) in terms of dentine exposed or not. The true status was determined histologically. Gender, age, professional experience, and specialization of the raters were documented. An algorithm was used to search for clusters of raters with high agreement in their ratings suggesting a common notion. The clusters identified were examined with respect to various aspects.

RESULTS

Four clusters of raters with high agreement could be found. The ratings of the raters in the cluster with the lowest diagnostic accuracy showed the highest correlation with the degree of tissue loss and the background tooth color, whereas the correlation with tissue loss was least in the cluster with highest diagnostic accuracy. The 15 raters of the latter cluster covered both students and dentists with or without specialization in erosion/cariology and/or long experience. This suggests that similar conceptual understanding of ETW can exist independent of professional experience.

CONCLUSIONS

The described methodology is useful to identify clusters of raters with a common notion about a specific diagnostic problem. The cluster-specific notions can be further examined based on existing study data or by group-based interviews of the raters of a cluster. This methodology allows investigators to learn more about useful or useless cues in diagnostic decision-making. This information can facilitate development or enhancement of guidelines on diagnostic decision-making.