Development of a risk-prediction model

A quantitative risk prediction model is to be constructed. We need to determine if the available data will fit an existing model and validate the results or if a new statistical model is required. Each case will be allocated into one of three categories (low, moderate and high risk). This stratification must have clinical validity and utility. The cut-offs for the stratification will be established based primarily on clinical utility and on the availability of the data. The cut-offs will be optimized to achieve optimum AUC, NPV, PPV, sensitivity and specificity values. It is expected that a minimal number of variables (2 maximum) will be used in the risk-prediction model for both over fitting and clinical action reasons. This prediction algorithm will allow proteocyte to have in hand an exploratory model to continue its work in developing a robust test that can have clinical utility in stratifying premalignat oral lesions according to the level of risk of progressing to cancer.

Faculty Supervisor:

Lehana Thabane

Student:

Partner:

Proteocyte Diagnostics

Discipline:

Mathematics

Sector:

Professional, scientific and technical services

University:

McMaster University

Program: