Computerized coding of medical problem statements and the Baylor Coder

Due to the narrow range of medical problem statements that regularly appear in medical practice, a modified table look-up procedure can successfully code medical problem statements. Medical problem statements are noun phrases that must fit into rather arbitrary classification schemes. Principles from morphology, semantics, syntax, and quantification linguistics provide some guidelines for designing a coder. The KODIAC coder illustrates how 90% accuracy can be achieved with modified table look-up. Mechanisms of the Baylor coder include compression of words and many table options. This author has collected data to assess the accuracy of the Baylor coder and to assess the difficulty of modifying the Baylor table. The Baylor coder is only 79% accurate but has only 1,400 rows in its table, while the KODIAC table has 6,000 rows. At the moment, the cost of improving the Baylor coder is difficult to compare to the cost of obtaining an effective KODIAC-like coder.