Cancer patients undergoing stem cell transplants face a long recovery requiring drugs that debilitate side effects and support. It is a challenging experience, indicating that over 70% of patients are not accommodating a drug regimen.
Statistician Susan Murphy spends her days helping people suffering from such challenging illnesses. Professor Mallinckrodt, professor of statistics and computer science and assistant faculty at Kempner Institute, and her team, meet healthcare needs through mobile apps rather than medicine.
Murphy's lab specializes in creating sophisticated calculation instructions known as Renforcement Learning Algorithms. This will form the technical backbone of next-generation programs and help you stick to regular toothbrushing, such as drug protocols.
And if this sounds like one of these ubiquitous apps that track steps or count calories, think about it again.
“If you've ever downloaded health apps, they tend to be pretty ridiculous,” Murphy said. “For example, you'll get a physical activity app, sprain your ankles, and keep ordering you to go for a walk.”
“If you've ever downloaded health apps, they tend to be pretty ridiculous.”
Susan Murphy
Using advances in artificial intelligence and sensing technology, to move beyond one size intervention, lab apps can meet real-time personalization, psychological rewards, and sometimes leverage social networks to help users stick to their goals.
This approach is called “just-in-time adaptive intervention” because it aims to provide support at the right time by registering changing needs and contexts.
Currently, Murphy Lab works with software engineers, cancer clinicians and behavioral scientists to develop apps for STEM cell transplant patients and their primary caregivers, usually parents.
Health care usually requires the involvement of others, especially for the most sick. For example, up to 73% of family care partners are primarily responsible for managing cancer-related drugs.
The researchers are in the early stages of developing the algorithm, and this year they will be deployed in a first round of clinical trials with collaborators from the University of Michigan and Northwestern University. This study indication for HCT will focus on adolescent and young adults who received stem cell transplants 14 weeks after surgery.
The algorithm will sequentially inform decisions, such as when to send motivational prompts to patients and whether to send messages and reminders to both patients and caregivers. This application includes a word guessing game that promotes social support and collaboration between patients and caregivers.
“We assume that patients can perform better and manage drugs when improving their relationship with their caregivers,” says Harvard Postdoctoral PhD Xu, who leads the development of the HCT algorithm.
The app employs augmented machine learning, in which the software “learses” from previous interactions. For example, rather than simply sending preset reminders about drugs, the algorithm adjusts timing and content according to the case that is most useful to the patient. This will make the notification less likely to be considered irrelevant or untimed and ultimately be habitually ignored.
“We learn the best ways to interact with each patient using algorithms,” Xu said.
“We'll learn the best ways to interact with each patient using algorithms.”
Ziping XU
Murphy Lab deploys algorithmic expertise across other domains. Along with collaborators at the University of Michigan, they recently piloted a program called Miwaves, targeting young adults who are abusing cannabis.
Like Adapts in the HCT app, Miwaves aims to continuously learn from interactions with each patient, improve decision rules and reduce daily intake.
The lab has also become a few years of a project called Oralytics. Recently, we have completed a 10-week randomized trial to help patients improve their delivery of push notifications to help them adhere to the toothbrush protocol.
The first Oralytics clinical trial included around 70 participants who received the mobile app using a wireless-enabled toothbrush that sent data to Procter and his gambling team collaborators.
Graduate student Anna Li Trella, who led the Oralytics project through the first trial, said the data collected recently helped the team develop ways to better handle troubling issues like missing data and software errors.
“There are many constraints to implementing algorithms in real life,” Torrella said. “We have conducted our first exam so the algorithms can make improvements to collect better data and help with better learning.”
Murphy believes her lab is creating practical pocket coaches that can help people get to where they want to go.
“Very few people can afford to buy a human coach. In fact, some people may not want such intensive human interaction,” Murphy said. “That's where these digital support ideas come into play.”
