Innovator Spotlight Q&A Series: Sudeshna Das, PhD
Q: Tell us about your innovation/work in the telehealth space, and the challenge(s) you were trying to solve.
SD: Alzheimer’s disease and related dementias (AD/ADRD) are under-recognized in the community, under-diagnosed by healthcare professionals, and under-coded in claims data, and as many as 50% of AD/ADRD patients may be undiagnosed. Underdiagnosis leads to missed opportunities to intervene in correctable conditions, failure to address safety risks like driving or medication errors, higher risk for medical procedures, and substantially inaccurate forecasting of future needs for public health officials. Thus, tools to identify signs of cognitive impairment are urgently required for timely intervention.
Passively collected real-world data, such as electronic health records, provide such an opportunity. With my colleagues at the MADRC, we have been developing deep learning algorithms to detect signs of cognitive impairment using structured (diagnosis, medications, healthcare utilization patterns, such as frequency of emergency department visits, and number of refill requests) and unstructured data (clinical notes) from electronic health records.
I hope to contribute in any way I can and make a difference in someone’s mother’s life—and I’ll feel I have made a difference in mine.
Our deep learning algorithm detects about 75% of missed diagnoses and has an AUC of 0.94 [indicating strong diagnostic accuracy]. We demonstrated that natural language processing of clinical notes provides the most information for detecting cognitive impairment. We are collaborating with the Dementia Care Collaborative at MGH to prospectively test our algorithm on Mass General Brigham primary care patients who are 65 years or older and who do not have an AD/ADRD diagnosis.
In addition to electronic health records, we are also investigating whether heart rate, activity, and sleep data from sensor streams, such as an Apple Watch or Fitbit, contain signals of cognitive decline. Such studies hold promise for developing low-cost, easily accessible Apps in real-world settings to address the global burden of disease.
Q: The process can be long and challenging, but also rewarding. What inspired you to begin this journey?
SD: I was first drawn to ADRD research because of the amazing work at our Alzheimer center, MADRC, led by Dr. Bradley T. Hyman. I was fascinated by the ongoing longitudinal studies of individuals with AD alongside those with normal cognition, the research on brains from our Brain Bank, and the myriad of technologies to measure and model the disease.
My research took on a particular personal relevance when my own mother started developing signs of cognitive impairment. I hope to contribute in any way I can and make a difference in someone’s mother’s life—and I’ll feel I have made a difference in mine.
Q: Where are you in the innovation cycle (i.e., early-stage commercialization)?
SD: We are very early on in our innovation cycle—deep in the development and validation phase. Much work remains to fully test our algorithm and ensure that it doesn’t suffer from biases in the training data. We are applying innovative active learning strategies to sample cases from diverse demographic and clinical backgrounds to build algorithms that are generalizable to a wide population.
Q: What internal resources have been most helpful to you?
SD: We had support from both the MADRC leadership and the Department of Neurology to fund this work. We are also extremely grateful to the CIDH and the Center for Faculty Development for publicizing our work and spreading the word.
Q: If you could give one piece of advice to another Innovator in the Mass General Brigham network, what would it be?
SD: Dream big: aim for the sky and you will at least reach the treetops. Imagine what you want to do and who you want to be. Have a plan to get there: a 5-year plan, a 3-year plan, a 1-year plan, and work backwards to today and now. Let failures be your teaching moments and move ahead with resilience.
And finally, always think about WHAT you are trying to do and WHY you are doing it. The HOW can always be figured out and is the easy part!