The Patient-Centered Outcomes Research Institute (PCORI) has announced a $1.05 million award to Fusheng Wang, professor in the departments of Biomedical Informatics and Computer Science, and his team, including Richard Rosenthal, MD, professor and addiction psychiatrist from the Department of Psychiatry. Their research is focused on machine learning models to predict patient outcomes.
“Deep learning is revolutionizing risk prediction models in health care by providing unprecedented accuracy and insights. As we integrate these advanced techniques into our systems, building credibility and trust through rigorous research and transparent processes is essential,” said Samir Das, professor and chair of the Department of Computer Science. “By demonstrating the robustness, reliability and explainability of these models, we not only enhance the quality of patient care but also foster confidence in the transformative power of machine learning.”
PCORI recently approved funding awards for studies on improving methods in research focused around patients. Wang’s research on using machine learning to predict patient risk was one of the studies granted an award.
Wang’s research focuses on creating models to predict how likely patients are to develop opioid use disorder and opioid overdose. Every individual patient varies in opioid risk. Using a prediction model, Wang and his team seek to develop a tool for clinicians to foresee patient risk.
This machine model is designed to help predict patient outcomes, which clinicians can use to influence their course of treatment. The machine learning model pulls from patient records to make a prediction. What is revolutionary about Wang’s research is what he calls the “stakeholder-in-the-loop approach,” where clinicians can provide feedback to the prediction model to make the output more accurate. This approach makes the machine learning model more human-centric.
“I think probably the most important contribution is the stakeholder-in-the-loop approach,” said Wang. “Stakeholders, including clinicians and patients, will participate in the full cycle of model design, development and evaluation. I think for the health care domain, that’s really something missing. We don’t see anybody doing something systematically like us. I think if we can, through our study, if we can provide a framework, the lessons learned can be very useful for others to adopt a similar methodology.”
One of the challenges that Wang and his team have to overcome is that patient data is very complex with lots of clinical variables contributing unclearly towards predicting risk. This is another reason why the stakeholder-in-the-loop approach is important because it gives clinicians a chance to add their clinical knowledge.
Wang's team aims at making this complex model user-friendly so that clinicians not only understand what it says but also contribute their own knowledge into it. The output then has to be concise, easy-to-understand and effectively communicated to patients.
“A doctor wants all information as quickly as possible while being comprehensive,” said Wang.“If the machine learning model generates a prediction then we need a precise summary about why such risk exists.”
What sets apart this research project involves collaboration among diverse experts.“The project brings in patient partners; clinicians; computer scientists; researchers; community representatives from New York State Office Mental Health & Suffolk County Dept Health,” said Rosenthal.
In long term goals,Wang hopes expanding beyond current focus,to other diseases like heart ailments eventually implementing within clinical settings,e.g., emergency departments testing fullest extent.
— Angelina Livigni