Machine Learning and Patterns of Primary Care Utilisation in Cancer Diagnosis and Outcomes

The Victorian Comprehensive Cancer Centre-funded Data-Driven Research Program was the first in Australia to link large-scale primary care and hospital data for the purpose of enabling increased capabilities in health services research. Until now, traditional analytical and statistical methods have been utilised to examine and map patterns in primary care attendances and how these are associated with cancer diagnoses, treatment and outcomes. These are generally based on the existing evidence base to test prevailing hypotheses and apply these to the local context.

This project seeks to apply novel analytical and machine learning methods to the large-scale linked data sources in order to generate new hypotheses and further characterise how people with cancer engage with primary care services. Specifically, this project would aim to identify patterns in various aspects of primary cancer attendances prior to a definitive cancer diagnosis (as identified in linked data sources).

These could include prescribing, test requests/results, semi-coded fields such as ‘reason for encounter’ and co-morbidities/other conditions captured in primary care management systems.

The project would identify a specific cancer type (i.e. Upper Gastrointestinal) with sufficient linkages between hospital diagnosis and primary care data and apply tools and algorithms to detect patterns present within specific timeframes of diagnosis or treatment.