Q&A with Susan A. Murphy, Harvard University

Dr. Murphy is a preeminent data scientist working in Precision Medicine and focused on developing data analysis methods and experimental designs to improve real time multi-stage decision-making in mobile health. She is particularly focused on methods and algorithms that can be employed on wearable devices, to deliver individually tailored treatments. She developed the sequential, multiple assignment, randomized trial (SMART). SMART designs provide scientists with the empirical tools to build adaptive interventions, treatment rules that dictate whether, how, and when to alter treatment for patients. The decision rules employ variables such as patient response, risk, burden, adherence, and preference. SMARTs are currently being used to build better treatments for a broad range of health problems including cocaine abuse, depression, alcohol abuse, ADHD, autism, and bipolar disorder. Read her full bio.

PMWC 2018 Michigan takes place June 6-7, 2018.

Q&A with Susan A. Murphy

Q: Your lab develops experimental trial designs and data analysis methods for increasing the usefulness of mobile health interventions. Can you please give examples how these trials are being used in the areas of substance abuse, physical activity, or helping people quit smoking?

A: We are involved in a number of mobile health trials that employ our experimental designs. Check out the website : https://methodology.psu.edu/ra/adap-inter/mrt-projects#proj! At this website you will find schematics of mobile health trials, most of which we are involved in. These include SARA, a first trial aimed at learning how we, along with Dr. M. Walton, Dr. P. Klasnja and Dr. I. Nahum-Shani can keep young adults, who are at high risk of substance use, engaged in data collection. This is a first step toward providing just-in-time interventions to reduce their substance use. We are also involved in a smoking cessation study with Dr. B. Spring and Dr. S. Kumar in which people who are trying to quit smoking wear a variety of sensors. These sensors are used to detect when the person is physiologically stressed –in this study we are using micro-randomization to learn whether it is useful to remind the person to practice their stress reduction exercises at these stressed times.

Q: What is the Sequential Multiple Assignment Randomized Trial (SMART) model that you developed and how it can be used by clinicians to tailor personalized interventions?

A: Whereas the micro-randomized trial is all about learning when and where it is useful for a wearable or smartphone to provide supportive interventions to you, the SMART is all about learning when and where it is useful for clinicians and other health personnel to provide treatments. Here the treatments can be medications or behavioral counseling or other types of interventions provided by clinical staff.

Q: How the Just-in-time adaptive interventions (JITAI) concept is applied in patient care?

A: JITAIs operationalize how a smartphone or other mobile device, including wearables can be used to extend care beyond the clinic. For example we are working with Dr. P. Klasnja to develop a JITAI for enhancing care after bariatric surgery. Bari-Fit is a quality improvement micro-randomized study recently conducted at Kaiser Permanente in Seattle. A schematic of this study can be found at https://methodology.psu.edu/ra/adap-inter/mrt-projects#proj This JITAI, once finalized, can be used to help people avoid regaining weight following bariatric surgery.

Q: What do you see as the greatest challenges facing the field of statistics in the coming years?

A: Statistics is undergoing very rapid change. The kinds of data that we can collect now-a-days is very complex, multi-time scale and multilevel ranging from many real-time streams of sensor data collected by wearables or other sensors placed in our environment all the way to genetic data and proteomics data. The old principles used by statistics and data scientists, to extract information from data and to provide measures of the confidence that one can attach to any statement involving data, need to be generalized or even replaced in this new world. It is a very exciting time to be a Statistician !