Determination of efficacy of nuclear medicine procedures

Nuclear medicine, a high technology field, is evaluated as to its usefulness. This paper describes the SNM study of 2023 patients comparing two methods evaluating efficacy for lung scanning (LS). Only the referring physicians determined the probabilities of the most important (MI) and most likely (ML) diagnoses and management before and after lung scanning. A logistic regression model was developed for probability of a signout diagnosis of PE. Equal patient groups tested the validity of the regression equations for the probability of PE as MI or ML. The models developed on Group I (G-I) and used on Group II (G-II) gave similar results. This shows that LS classifies PE and NOT PE categories where PE was considered both MI and ML. Entropy minimax pattern detection (EMPD) attempts prediction of signout diagnosis and management from prior patient attributes. In 2023 cases, attributes alone could not eliminate the use of LS for all patients. Comparing the two methods, the predictive values, sensitivity and specificity of each method are similar. EMPD predicts on a relatively small percent (40% before LS, 71% post LS) while the logistic equation predicts on 100% of the cases. An advantage of EMPD is that it does not requiremore » estimates of prior probability. However, LR, uses this estimate, thus incorporating intuitive knowledge not evaluated by EMPD. These methods are unique in showing that LS can direct the referring physician toward or away from anticoagulant therapy based on findings of the lung scan.« less