Cut point selection to discretize continuous diagnostic tests: an application for diabetes disease

Emerson Barili; Jorge alberto Achcar

No Detail Found!

Cut point selection to discretize continuous diagnostic tests: an application for diabetes disease

Author(s)

Emerson Barili Medical School, Social Medicine, University of S˜ao Paulo Ribeir˜ao Preto, S˜ao Paulo, BRAZIL

Jorge alberto Achcar Medical School, Social Medicine, University of S˜ao Paulo Ribeir˜ao Preto, S˜ao Paulo, BRAZIL

Abstract

Diagnostic tests are usually measured in a continuous scale but discretized to an outcome such as an indicator of the disease if the continuous test value is larger (smaller) than a fixed threshold or no disease if the continuous test value is smaller (larger) than the specified threshold. The choice of cut points to transform the continuous covariates in binary covariates usually is arbitrary or based on medical experts. In this study, we propose under a Bayesian inference approach a method to find optimal cut points for the covariates afecting the occurrence or not of a specified disease assuming a logistic regression model in presence of many covariates. With the obtained cut points, we also estimate sensitivities and specificities for the obtained diagnostic tests. We also present a Bayesian discrimination criterion to compare the proposed model with unknown fixed cut points and considering a model assuming fixed arbitrary cut points. An application considering a diabetes data set in India illustrates the proposed methodology.

Publication Details

Page(s) 43-58

DOI DOI not available

Published Journal: Journal of Applied Probability and Statistics, Volume: 19, Issue: 3, Year: 2024

Request the full text from PASTIC

Keywords

Bayesian inference, Logistic Regression, and phrases, diagnostic tests, MCMC methods, cut point, diabetes in India

Citations

Downloads

Views