Key points are not available for this paper at this time.
A computer-assisted statistical method of logistic discrimination is described for the diagnosis of keratoconjunctivitis sicca. Ten symptoms and seven signs of keratoconjunetivitis sicca in 40 patients with keratoconjunctivitis sicca and 37 patients with no clinically obvious ocular pathology who acted as controls formed the basis of the computer diagnosis. As a test, the statistical method was applied to 34 further patients, 17 of whom had keratoconjunctivitis sicca. Of the latter, two patients were incorrectly diagnosed and five were given queried diagnosis. The remaining 27 patients were diagnosed correctly. Study of the inter-observer error in eliciting the symptoms and signs of keratoconjunctivitis sicca by an ophthalmologist and two physicians indicated a high degree of concordance in eliciting symptoms, but a low concordance in eliciting signs. When calculation of diagnosis was based on analysis of the 10 symptoms only, in the 34 patients, 17 of whom had keratoconjunctivitis sicca, the results were better than when analysis was based on the 10 symptoms and seven signs. Four patients with no keratoconjunctivitis sicca were allocated a query diagnosis: none of the patients were allocated incorrect diagnoses. The method is suitable for use by physicians in screening patients with rheumatoid arthritis for keratoconjunctivitis sicca, and has the attraction that once the scoring system has been established it is possible to calculate the diagnostic score for a patient using simple arithmetic or a desk calculator.
Anderson et al. (Sat,) studied this question.