Background: The Global Assessment of Function (GAF) scale is widely used in adult psychiatric practice and research but it has not often been used in learning disability, which is inherently more complex.
Aims: To evaluate the reliability of GAF in the assessment of learning disability.
Method: GAF reliability was tested by simultaneous multiple rating of unselected case vignettes (n=19-25) from health professionals of different disciplines, under controlled conditions. Analysis of reliability was made with the intraclass correlation coefficient (R(1)) with separate assessments to determine rater bias and individual performance of raters.
Results: The results of three data-sets showed generally poor overall levels of agreement, with R(1) levels of 0.35 and 0.28 and somewhat better levels for current GAF scores (R(1)=0.49). However, a subset of raters was identified that achieved much higher levels (R(1)=0.54 to 0.74).
Conclusions: The GAF, in its current format, is not reliable enough to be used in the routine assessment of learning disability. A subgroup of raters, however, have ratings that are, by current biostatistical criteria, sufficiently reliable.