Some observations on the development of a 'scoring system' in an expert system for prediction of ovulation

As part of the construction of an expert system for the diagnosis of infertility we have developed a scoring system for the identification of conditions which can interfere with ovulation. The scores were obtained from a group of clinicians, and the arithmetic means of these scores were used in the actual system. Correlation between the scores obtained from individual clinicians was very poor (r = 0.345 +/- 0.134). Correlation improved when groups of clinicians were compared. Thus scores derived from a group of clinicians will reflect more reliably the 'general medical opinion'. Scores derived from the opinion of clinicians should be regarded with caution until probabilities or weighting factors derived from real clinical data become available.