Published: Journal of Analytical Toxicology, Volume 22, Number 6, October 1998, pp.487-492.

An Evaluation of the Role of ROC Plots in the Prediction of Heroin Use from Total Codeine and Total Morphine Concentrations in Urine
Carol L. O’Neal and Alphonse Poklis

A diagnostic system to predict the presence of 6-acetylmorphine (6AM) in opiate-positive urines was recently proposed. A twofold criterion based on the total morphine concentration and the total codeine to total morphine concentration ratio was identified. Using relative operating characteristic (ROC) analysis, it was determined that the diagnostic system had a sensitivity of 92%, a specificity of 79%, and an overall accuracy of 73%. We applied similar decision criteria to a study population of 125 opiate-positive urines collected from criminal justice clients of a West Coast reference laboratory. ROC analysis on this population produced very different results: a sensitivity of 91%, a specificity of 49%, and an accuracy of 45%. These data illustrate the importance of choosing a representative study population without any selection biases that may compromise the validity of the accuracy measure. The ROC plot is an important tool for assessing a clinical testÕs performance, but in order for toxicologists and Medical Review Officers to benefit from the diagnostic test results, they must also know the predictive value (based on test accuracy and the prevalence of heroin use) of the test results. They need to know how well the test predicts the presence of 6AM, and therefore, the illicit use of heroin, in the population of interest whether it be workplace, criminal justice, hospital emergency department clients, or a combination of all populations.

Reproduction of editorial content of this journal is prohibited without publisher’s permission.

| Subscribe | Current Issue | Back Issues | Search | Advertise | Other Publications |