Fuzzy logic and logistic regression in the decision making for parathyroid scintigraphy study

Abstract

OBJECTIVE: To develop and compare two mathematical models, the first one based on logistic regression and the second one on fuzzy sets theory, aiming at defining a laboratory testing-based measure of indication for submitting patients to parathyroid scintigraphy. METHODS: One-hundred and ninety-four patients with serum calcium and parathyroid hormone available were identified from the data registry of parathyroid scintigraphy of a diagnostic laboratory in São Paulo, Southern Brazil, in the period between January 2000 and December 2004. The logistic regression model was developed using SPSS and the fuzzy model was developed using MatLab software programs. The performances of both models were compared using ROC curves. RESULTS: The performance of both models were statistically different (p=0.026). The area under the ROC curves were 0.862 (95% CI: 0.811-0.913) for the logistic regression model and 0.887 (95% CI: 0.840-0.933) for the fuzzy model. The latter had the advantage of allowing to making decisions based on parathyroid hormone information within a non-discriminating range of calcium values. CONCLUSIONS: The mathematical model based on fuzzy sets theory seemed to be more useful than the logistic model in the decision making for scintigraphy indication. However, inferences can be made only regarding model comparison and not for parathyroid scintigraphy itself since the data analyzed was not representative of any population.