Prospective Trial of Computer Program for the Diagnosis of Thyroid Disorders

ABSTRACT A computer program, developed to help in the diagnosis of thyroid disorders, has been operated routinely in a general thyroid clinic. Diagnostic probabilities were computed from relevant signs, symptoms and test results using Bayes' theorem. In a prospective trial with 1,200 consecutive new patients entering the clinic the program performed very satisfactorily in classifying metabolic status (99.1% correct). In deciding about other thyroid disorders before biopsy neither the physicians (41.4% correct) nor the computer (34.5% correct) were very successful. Better tests and statistics, are expected to improve these results. In this (Sydney, Australia) clinic population the estimated percent incidence of thyroid disorders was: hypothyroidism 2.7%; hyperthyroidism 10.0%; subacute thyroiditis 0.92%; chronic thyroiditis, non-specific 0.38%, Hashimoto's 2.63%; thyroid cancer 1.00%, abnormal synthesis 0.08%. For 19/2, the mean iodine intake for female patients in Sydney was computed to be 242 μg/day.