Liver transplant data from Melbourne’s Austin Hospital has been used to create an algorithm that can predict the outcomes of future transplants with a high degree of accuracy.

The research findings, published recently in Transplantation, have the potential to reduce liver transplant failures, morbidity and mortality while improving utilisation of a scarce resource.

Chief Investigator Dr Lawrence Lau, a surgeon and research fellow at Austin Health, said a machine-learning methodology extracted the key donor, recipient and transplant factors influencing outcomes at the hospital between 2010 and 2013.

The subsequent algorithm was 84% accurate at predicting graft failure 30 days post-transplant compared to 68% accuracy with current methods.

“This study is a proof-of-concept that machine-learning algorithms can be an invaluable tool, supporting the decision-making process for liver transplant organ allocation.”

“The benefits of being able to assess the suitability of organs in a quantitative way, and to assess how well they match a particular recipient, are huge,” he said.

Dr Lau said machine-learning algorithms predict the outcome of a new event, based on multiple interactive factors observed in previous known events.

“This approach not only considers the influence of each variable, but also looks at how the variables interact with each other in complex, interdependent ways.”

“Machine-learning algorithms are already used across a wide range of fields including search engines, agriculture, financial markets and match-making. There is so much untapped potential to apply this in medicine,” Dr Lau said.

The next step in the research is a randomised prospective trial comparing liver transplant decisions aided by machine-learning algorithms with unaided clinician-made decisions.

“At the moment there’s really no method to determine the safest and most effective way to use the scarce donor livers. It largely comes down to a surgeon’s judgement call of who we should give a particular organ to,” Dr Lau said.