Algorithms For Early Detection of Cancer Using Linked Primary Care Data
Professor Jon Emery,
The survival rate for cancer depends significantly on the stage at which the cancer is diagnosed. Detecting cancer earlier, when treatments are more effective, is therefore crucial. Since primary care is the first point of contact for most patients who are eventually diagnosed, the primary care electronic medical record can reveal signals that may indicate a cancer diagnosis down the track. The primary care cancer group is investigating the application of statistical and machine learning algorithms for the early detection of cancer, using linked primary care, hospital, and cancer registry data. Similar algorithms have been developed in the UK and shown to meet the threshold for recommending further screening. We aim to validate existing algorithms and develop novel algorithms for this purpose. Furthermore, implementation of these algorithms through the Future Health Today decision support tool will provide general practitioners with valuable information regarding which patients are at highest risk of developing cancer.
Herman Prof Jon Emery, Academic GP, University of Melbourne
Dr. Javiera Martinez Gutierrez, Academic family physician, University of Melbourne
Dr. Meena Rafiq, Research Fellow, University of Melbourne and University College, London
Dr. Alex Lee, Research Fellow, University of Melbourne
Mr. Damien McCarthy, Data Analyst, University of Melbourne
Ms. Olivia Wawryk, Data Analyst, University of Melbourne
Ms. Sophie Chima, Graduate researcher, University of Melbourne
Dr. Shaoke Lei, Research Fellow, University of Melbourne
Ms. Silja Schrader, Health Data Analyst, University of Melbourne
Dr. Brian Nicholson, Nuffield Department of Primary Care, Oxford University
Mr. Zaher Joukhadar, Melbourne Data Analytics Platform, University of Melbourne
Dr. Kristal Spreadborough, Melbourne Data Analytics Platform, University of Melbourne
Victorian Comprehensive Cancer Centre, and VCA through their fellowship program.
We aim for this research to provide a set of algorithms that are applicable to the Victorian primary care context, that have been rigorously tested, and can be implemented into a primary care setting to assist GPs identify those patients who are at greatest risk of developing cancer. The linkage of additional data sources in the future may potentially increase the accuracy and suitability of these algorithms further.