Medical diagnostic testing models for multifactorial disease classifications

In this note we extend the logistic regression receiver operating characteristic (ROC) analysis framework to diagnosing disease that can be classified in a 2 × 2 factorial setting. Similar to the standard ROC curve analysis these new models accommodate prediction of cross-classifications for treatment success and diagnosis of disease. A didactic example is given for jointly diagnosing malaria positive and respiratory distress in children in Ghana.