Mobile app helps people recover from substance abuse

What are digital interventions? They provide support to individuals in their everyday life. Examples of such interventions are: suggestions on how to be more active in one’s current setting; different types of reminders to perform self-care behaviors such as taking the prescribed medications; motivational messages, and reinforcements of positive behaviors. Interventions may be delivered via a wearable or other smart device. According to researcher Susan A. Murphy, digital interventions can also be used to augment clinical care. It is therefore critical to learn whether and in which settings it is most effective to deliver the interventions.

05. 10. 2022 08.00 hod.
Od: Redakcia Našej univerzity

How can “micro-randomized trials” be used to improve digital health interventions? „For example, a person who is recovering from drug use might follow regular counseling sessions. In addition, the person might be provided a mobile health application that provides helpful information/tips whenever, between sessions, the person might need such help. The app might include a locator which will help the person find the nearest Addicts Anonymous meeting in both time and space. The app might contain information/ideas for managing cravings covered in the counseling session that the person can refer to between sessions,“ Susan A. Murphy explains.

Health care outside the clinics

Digital intervention can also help those who either don’t have access to in-person care, or don’t want it. Digital intervention could use sensor data from one’s smartphone or wristband (step count, patterns of typing on the smartphone, location, proximity to where the person used to take/buy drugs and so on) to learn useful times to deliver suggestions to individuals or to connect the individual to a peer mentor. „In some cases the digital intervention might be used by the individual to track their behaviors/physiological state and then share the information with clinical staff. In summary, digital interventions are one way to extend care outside of a clinic setting,“ adds Susan Murphy.

Data science and health psychology

Data science can be very helpful in health psychology. According to Susan Murphy, it can be used to design experiments to collect information that health psychologists use to inform the construction of more effective digital, and other types of interventions. The micro-randomized trial method which she developed can also be used in constructing digital interventions. It randomizes each individual at each of the many times. At each of these times it may be more or less effective for a smart device to reach out to the individual and deliver behavioral treatment (e.g. ideas for managing cravings, motivational content, connection to others, information about where to obtain help, assistance in setting goals for the next day and so on). 

Susan A. Murphy
is Professor of Statistics, Radcliffe Alumnae Professor at the Harvard Radcliffe Institute and Professor of Computer Science at the Harvard John A. Paulson School of Engineering and Applied Sciences. Her research focuses on improving sequential, individualized, health-related decision-making; in particular, clinical trial design and data analysis to inform the development of just-in-time, adaptive digital health interventions. Her lab works on online learning algorithms for developing personalized mobile health interventions. She developed the micro-randomized trial method used in constructing digital health interventions. This trial design is in use across a broad range of health-related areas. She is a 2013 MacArthur Fellow, a member of the National Academy of Sciences and the National Academy of Medicine, two of the three US National Academies. She was one of the keynote speakers at European Health Psychology Society conference in Bratislava. 

„These trials provide data to ascertain which type of behavioral treatment is most effective at what time and in which settings. Further, the data from these trials can be used to investigate positive, as well as negative, delayed effects of the treatment (negative: burden or habituation; positive: forming healthy habits, increasing resilience),” says Susan A. Murphy.

Artificial intelligence method

In addition, Susan A. Murphy also uses an artificial intelligence method, "reinforcement learning", to personalize digital health interventions. Artificial intelligence algorithms, one of which is “reinforcement learning,” can be used to continually learn/revise which behavioral treatments are most effective for an individual at a given time. „The idea is that as the individual interacts with the digital intervention, the artificial intelligence algorithm will dynamically, using the accruing data on the individual, personalize the delivery of behavioral treatments to each individual. This may involve the use of predictions/detections of moments of high risk (e.g. high risk of substance use relapse) and moments at which the individual might be least burdened by the delivery of a behavioral treatment.“

Any negatives?

Of course, there are some negatives in the use of artificial intelligence. Susan Murphy states three examples. First, if the digital intervention involves the use of online artificial intelligence algorithms, then it is critical that these algorithms operate in a stable and reliable way so as to not discourage an individual who is struggling with managing a chronic disorder. Second, if behavioral health science and the AI algorithm are not carefully coordinated, less confident risk predictions might be incorrectly communicated to the individual. Third, an incorrect deployment of digital interventions may aggravate health disparities.

„For example suppose a digital intervention for helping people manage cardiovascular problems is only developed using input/data from middle-income Caucasian men. In this case, the intervention may be much less effective in other types of individuals. The behavioral suggestions may be inappropriate for individuals with different ethnic backgrounds or for individuals who have lower incomes and thus potentially strongly competing needs and goals,“ explains Professor Murphy

Radka Rosenbergová