A COMPUTER PROGRAM FOR DIAGNOSIS OF THYROID DISEASE

Abstract

A program, with flow-charts, for the diagnosis of thyroid disease based on experience gained with 1,379 cases of suspected thyroid disease has been written in Fortran II (for the IBM 709) and is available on request.* It makes use of a conditional probability model of medical diagnosis and is believed to imitate the thought processes of a physician experienced in thyroid disease. First, it evaluates the probability of a functional thyroid abnormality based on signs and symptoms alone; it then evaluates the probability of a functional abnormality based on the laboratory data, and finally considers all data together. Laboratory values and the probability values are routinely searched for some 10-15 specific combinations (such as those produced by acute thyroiditis, congenital deficiency of thyroid binding globulin, masked hyperthyroidism, etc.). If a particular test may give a conclusive diagnosis, it is requested and, if present, a specific diagnosis is printed out. After the diagnosis has been made, all values are entered back into the basic data matrix so that the subroutine changes the basic data matrix after each diagnosis. In this way, the program may be said to "accumulate experience." In the last 500 cases of suspected thyroid disease examined by the program at the University of Florida College of Medicine, 14 mistakes in the diagnosis of the thyrometabolic status were made (2.8 per cent) of which 7 could be corrected by easily written subroutines. We believe that the program is now ready for general use by other institutions and that it is in a form which can be easily modified for correction mistakes. It is only by the use of this program in a large number of institutions that the necessary number of rare cases can be encountered and appropriate additions to the program made, in order to render the program truly useful to the physician.