November 13, 2020

How an optimization algorithm can help Ontario detect opportunities for better cancer care

Drs. Katharina Forster, Timothy Chan and Claire Holloway.

Research team develops a Google maps-like algorithm to pinpoint when cancer patients may diverge from the standard course of treatment

Every cancer patient’s experience is unique but there are standard sequences of steps that help patients and their care teams navigate through screening, diagnosis, treatment and monitoring. These steps are published in pathway maps but are these maps followed in practice? Researchers supported by OICR’s Health Services Research Network, led by Drs. Timothy Chan and Claire Holloway, are working to answer that question.

Chan and collaborators at Ontario Health have developed new methods to measure the difference between a standard clinical pathway map and the actual care that a patient receives in practice. They leveraged real-world health data from Ontario patients to develop these methods, which could potentially be used to identify targets for quality-improvement initiatives.

“Pathway maps help optimize patient survival, healthcare costs and wait times at a population level,” says Holloway, co-principal investigator of the project and Provincial Clinical Lead of Disease Pathway Management (DPM) at Ontario Health.

“We have now derived a way to measure the alignment between actual care and the care described in a pathway map, analogous to measuring how a driver’s route differs from the Google Maps-suggested route,” says Chan, co-principal investigator of the project, Professor at the University of Toronto and Canada Research Chair in Novel Optimization and Analytics in Health.

To address this challenge, the team based their algorithm on an inverse optimization framework, a type of framework used to solve problems across a variety of disciplines, including telecommunications routing, medical radiation therapy planning, and investment portfolio management.

The research team first applied their methods to stage III colon cancer patient data and is now applying their methods to breast cancer care. The ultimate goal would be to use these methods across different cancer sites and potentially different diseases to help promote and implement best practices along the care continuum in Ontario’s healthcare system.

“We’re proud to apply our framework at a large scale to help provide meaningful quantitative measures of system efficiency and variation,” says Chan. “It’s exciting to see that these methods could allow Ontario Health to monitor and evaluate complex practice patterns at a population level.”

“Variations between a patient’s experience and the standard clinical pathway map isn’t necessarily a bad thing but it may prompt us to investigate further,” says Dr. Katharina Forster, Team Lead of DPM at Ontario Health. “We can look into why, when and where the variation is occurring.  In this way these new methods and tools are allowing us to generate hypotheses about the causes of variation so we can better understand our care practices, make data-driven decisions and ultimately improve our cancer care system.”

“Ultimately, we’re looking to measure, monitor and improve our system across the province,” says Holloway. “Our rich data in Ontario and our capabilities in machine learning are outstanding. Thanks to OICR, we can bring these disciplines together to make a positive impact on our health system.”

The Health Services Research Network is co-funded by OICR and Cancer Care Ontario, now part of Ontario Health.