Evaluation of different methods used in applying Bayes' rule to medical decision making

The computer is often used as a tool for medical decision-making. One of the common methods used in computer diagnosis consists of using Bayes' equation to determine the probability of a patient having a disease. One step in this procedure consists of determining the probability distribution of each laboratory test. This distribution is usually expressed in the form of a histogram, in order to obtain the probability values need for Bayes' equation. In order to reduce errors introduced by using histograms, a method has been developed to obtain a continuous curve. This methods uses a mathematical model to obtain a curve approximating the distribution curve. An evaluation of this method has been conducted using two separate studies. The overall results indicate that the model does provide a good representation of the laboratory data. The results for the continuous curve method are comparable with the other methods, however, and do not show a consistent improvement. The two studies used in testing the different methods have also been evaluated to determine their usefulness in a clinical setting. Both studies consist of methods of evaluating laboratory data. The first study is a method of diagnosing myocardial infarction and produces results that indicate it may be useful in a clinical setting. The second study determines the reliability of an abnormal creatinine result. In this case the results may not be reliable enough for clinical use.