Discriminating Three or More Ordinal Outcomes with Continuous Measurements by Combining Cumulative Logit Regression and ROC Curve Analysis

• Activity Number: 500
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Rey DeCastro*
• Companies: CDC
• Keywords: cumulative logit regression ; clinical chemistry ; diagnostic test ; proportional odds ; receiver operating characteristic curve

Abstract

Standard receiver operating characteristic (ROC) curve analysis estimates a cutpoint or threshold for a continuous measurement that discriminates two discrete outcomes. This paper describes a procedure for estimating continuous cutpoints discriminating three or more ordinal outcomes. The procedure comprises a two-stage, semiparametric approach combining conventional cumulative logit regression with a cumulative version of ROC curve analysis. Performance is evaluated under simulation (assuming proportional and non-proportional odds) and demonstrated with biomarker data from the National Health and Nutrition Examination Survey to discriminate three ordinal tobacco smoke exposure categories.