Watson Health Scientific Update
At IBM Watson Health, we are focused on applying next-generation data, analytics and artificial intelligence (AI) technologies to major health and societal challenges, so that people around the world can live better, healthier and longer lives. IBM has been leading with science in health and life sciences for decades, and we are proud of our work in the areas of research, real world evidence and AI. Last year, we published over 400 pieces of peer-reviewed abstracts, posters, presentations and publications in these key areas.

As we kick off 2019, we’re proud of the most recent pieces of evidence that we published in the past quarter. Our work points to scientific progress and real-world evidence among some of the most critical aspects of health and life sciences. For example, we’ve built a tiny fingernail sensor prototype that could one day help clinicians to better track and monitor complex diseases. We’ve also explored trends impacting public health and policy issues, such as the varying impact of the opioid crisis across states and the early impact of Affordable Care Act coverage.

This report highlights our top scientific progress from the fourth quarter of 2018, and spotlights new data that was recently published in peer-reviewed medical literature and at major scientific events. It features an in-depth look into an award winning piece of evidence presented at the American Medical Informatics Association (AMIA) Annual Symposium, which views the challenges of healthcare management through a unique lens and explores how behavioral influences can affect patients.

We hope you will take the time to read through our accomplishments, and that they will spark discussion, inspiration and thought in your discussions on how we can transform health through the power of data, analytics and AI.
Meet the Scientists: Pei-Yun Sabrina Hsueh, Health Behavioral Insights Project Lead

Pei-Yun Sabrina Hsueh is a research staff member in the IBM Research Lab in Yorktown Heights, New York, and focuses specifically on health behavioral insights projects. She holds a Ph.D. in informatics (computer science) from Edinburgh University and a master’s degree from the University of California, Berkeley. Upon joining IBM Research, she decided to apply her technology and healthcare informatics skills to the healthcare space to better understand the human side of health, treatment and provider care.

There are many more factors that determine the ultimate outcome of a patient than can typically be captured during the short interactions which happen within a doctor’s office. These behavioral factors account for approximately 40 percent of the global chronic disease burden. Sabrina and her team explore how AI can be used to model human health behavior; and how these models can inform the management of multiple interacting factors that may influence a patient’s choices. Her expertise includes data analytics, AI engineering, machine learning, natural language processing and intelligent interfaces

Why Sabrina is interested in this field of work, in her own words

“There is a pressing need to better understand human interactions within healthcare, and how changing certain behaviors among doctors, caregivers and patients can dramatically transform and shape the health of individuals and broader populations,” said Sabrina. “This is where AI has a great deal of potential. Machine learning and technology can help us identify patterns and nuances in healthcare data that can give us clues into how ongoing routines and interactions with patients – such as care management – can help them to be healthier and prevent future complications.”
Behind the Data: A Real-World Evidence Approach of Care Plan Personalization Based on Differential Patient Behavioral Responses in Care Management Records, AMIA 2018
In the world of care management, evaluating patients, giving a diagnosis, and providing clear instructions for treatment and evaluating patients is the job of a team of health professionals working with a patient. However, there are multiple factors that play a role in a patient’s treatment - and ultimate outcome - outside of relatively infrequent and short visits to a doctor’s office. This is where the essential role of a care manager comes into play.

The outcomes of healthcare interactions that require patient engagement are as varied as patients’ behavioral preferences. Three patients who are prescribed the same medication treatment from their doctor may have totally different outcomes due to their level of engagement in their health. The first patient may diligently take medication, but choose to ignore the lifestyle and diet changes recommended by their healthcare team. On the other hand, the second patient may completely change their diet, but may be less than reliable when it comes to maintaining a regular exercise regimen or remembering to pick up their prescriptions. The third patient placed under the same treatment plan may have managed to adhere to both medication and lifestyle changes, but may fail to show up to routine appointments meant to monitor their progress. Care managers help to close these gaps between what is expected from the patient and what they actually do by helping patients address their unique barriers to adherence.

When left unaddressed, barriers to treatment adherence limit the effectiveness of otherwise efficacious treatments, and lead to a tremendous healthcare burden. In this study, we’ve explored how AI can improve the productivity of care managers and remove inefficiencies, and ultimately increase the volume and effectiveness at which they can help their patients to live in a way that increases their chances of a healthy outcome.

Care planning often relies upon population-based evidence. However, individuals respond differently to interventions. The end goal of this research is simple: empower care managers to drive better engagement by helping them focus their attention on patients who are more likely to respond well to different interventions. The team set out to apply machine learning to real world data from care management records to personalize care planning strategies to the individual. Individual-level care planning strategies learned from practice by the Behavioral Response Inference Framework (BRIeF) framework achieved 87.24 percent accuracy, outperforming population-level strategies and delivering more accurate intervention recommendations for goal attainment. These results may lead to increased engagement, which is a goal unto itself as it can increase a patient's sense of control and agency, and is generally a necessary precursor to improved outcomes.

This collaboration between IBM Research and IBM Watson Health was awarded a Distinguished Paper Award at the 2018 American Medical Informatics Association Annual Symposium. Research such as this is crucial to lay a foundation for a healthier and aging society. As our population becomes older, the most effective way to alleviate our healthcare burden will be to move the needle on chronic disease from treatment to prevention. Ensuring that critical players of this effort, such as care managers, are supported will be crucial to reinforcing the prevention arm of our healthcare system.
Data + Clinical Evidence Highlights – Q4 2018
Watson Health AI Offerings & Capabilities
Manuscripts and Articles

The Oncologist

Frontiers in Medicine

Pharmaceutical Medicine

IBM Journal of Research and Development

American Medical Informatics Association (AMIA) Annual Symposium 

Thought Leadership

American Journal of Roentgenology

Abstracts and Posters

American Public Health Association (APHA) 2018 Annual Meeting and Expo

Watson Health Real-World Evidence

Manuscript and Articles

Journal of the American College of Cardiology

Health Services Research


The Journal of Behavioral Health Services Research

Journal of the American College of Surgeons

Journal of the American Geriatrics Society

Journal of Occupational and Environmental Medicine 

Thought Leadership

American Journal of Public Health 

IBM Research

Manuscripts and Articles

American Medical Informatics Association Annual Symposium

American Medical Informatics Association (AMIA) Annual Symposium

American Medical Informatics Association (AMIA) Annual Symposium

American Medical Informatics Association (AMIA) Annual Symposium

American Medical Informatics Association (AMIA) Annual Symposium

BMC Bioinformatics

Academic Radiology

Scientific Reports

Innovation in Aging

BMC Health Services Research

PLoS One
Contact Us:
Stay subscribed to keep up to date on how we’re helping our clients and partners around the globe accelerate innovation and tackle some of the world’s biggest health challenges with data, analytics and AI. For additional information, or to receive the full list of studies published in Q4 2018, please contact Erin Lehr Angelini at edlehr@us.ibm.com.
IBM Watson Health| 75 Binney Street, Cambridge, MA