A validation study of a diagnostic algorithm to identify patients with colorectal cancer

Project Details

A diagnostic validation analysis to create estimates of the sensitivity, specificity, positive predictive and negative predictive values of the PIT–M3D algorithm for detecting CRC.

Patron ID: PAT054

Project Lead:

Prof Jon Emery

Overview:

Haemaglobal Pty Ltd funded the VCCC Alliance Data Connect program within the Cancer in Primary Care team, to create a cut of Patron data that includes a cohort of patients diagnosed with colorectal cancer (CRC) and a larger cohort of age-sex matched controls. Haemaglobal Pty Ltd, an Australian biotechnology research company that specialises in the research and development of precision haematology tools and data driven diagnostics.

Patron primary care data was linked to the Victorian hospital Admitted Episodes Dataset (VAED) through BioGrid Australia.  Data Connect’s research identified patients with a CRC diagnosis recorded in either dataset. Each identified CRC case was then matched to ten controls from a pool of controls in Patron who are ‘active’ in the year before diagnosis based on age and sex.

The data sets were provided to Heamaglobal via a secure data transfer.  After the initial ‘training’ step, Haemaglobal Pty Ltd ran their diagnostic algorithm “PIT-M3D System” across the ‘validation’ data set to generate a report that flagged patients with a suspected diagnosis of CRC identified by the algorithm.  This report and the datasets were returned to the Data Connect research team to conduct the diagnostic validation analysis to create estimates of the sensitivity, specificity, positive predictive and negative predictive values of the PIT-M3D System for predicting CRC.