Machine Learning and Health Outcomes in Cancer Care Delivery Research

Monday, May 16 & Tuesday, May 17, 2022

Read the report (PDF, 3.2 MB)

Overview

The National Cancer Institute conducted a Machine Learning and Health Outcomes in Cancer Care Delivery Research virtual workshop on May 16-17, 2022 focused on research in machine learning, cancer care delivery, and health outcomes. The goals of the workshop were to identify scientific gaps, barriers, and opportunities in this emerging area of cancer-focused healthcare research.

Background

The application of machine learning and natural language processing on electronic health record data has resulted in new approaches and data sources used in healthcare delivery to improve cancer care quality and patient health outcomes. These methods allow researchers to:

Goals

The goal of this workshop was to highlight current cancer-focused research in health outcome risk prediction, symptom trajectories, natural language processing methods, and point-of-care decision support. Panel sessions identified the strengths and potential of machine learning methods, barriers, and opportunities specific to cancer care delivery research.

Agenda & Recordings

Title Speaker(s)
Day 1: May 16, 2022

Welcome and Workshop Overview

video icon Watch this sessionExternal Web Site Policy

Katrina Goddard, Ph.D.
Division Director
Division of Cancer Control and Population Sciences (DCCPS)
NCI

Roxanne E. Jensen, Ph.D.
Outcomes Research Branch
Healthcare Delivery Research Program
DCCPS
NCI

Keynote Presentation: Driving Health Equity through Representative AI and Inclusive Innovation

video icon Watch this sessionExternal Web Site Policy

Abby Sears, M.H.A.
President & Chief Executive Officer
OCHIN

Session 1: Landscape of NCI-funded Research

The inclusion of machine learning methods in healthcare delivery research has rapidly increased, resulting in increased grant NCI grant application. This session will review extramural trends in NCI research grant applications over the last five years proposing machine learning methods, an electronic health record-based data structure, and cancer healthcare delivery research focus. It will highlight emerging trends in research topics, methods, populations, and outcomes.

video icon Watch this sessionExternal Web Site Policy

Moderator:

Ngoc Tran, M.D., M.S.
Surveillance Informatics Branch
Surveillance Research Program
DCCPS
NCI

Panelist:

Roxanne E. Jensen, Ph.D.
Outcomes Research Branch
Healthcare Delivery Research Program
DCCPS
NCI

Session 2: Applying Machine Learning to Identify High-Risk Patients & Outcomes

Machine Learning holds great promise to leverage rich digital data collected during routine clinical practice to identify high-risk patients and predict diverse health outcomes. In particular, prediction models are of great interest in clinical settings, where preventing adverse outcomes, such as severe treatment side effects or early detection of aggressive cancers, may allow for more informed, early, and effective clinical intervention. Despite the potential of machine learning to advance cancer research and clinical practice, there are challenges limiting its use. This panel will include presentations from experts in the application of “real world” machine learning examples including the identification of clinical patients at high risk for lethal cancer and the development of algorithms to predict severe treatment-related adverse events among cancer patients.

video icon Watch this sessionExternal Web Site Policy

Moderator:

Andrew Freedman, Ph.D.
Clinical and Translational Epidemiology Branch
Epidemiology and Genomics Research Program
DCCPS
NCI

Panelists:

  • Amrita Basu, Ph.D.
    University of California
    San Francisco
  • Chris Sander, Ph.D.
    Dana-Farber Cancer Institute
  • Sherry-Ann Brown, M.D., Ph.D., FACC, FAHA
    Medical College of Wisconsin
  • Shreyas Lakhtakia
    Flatiron Health

Session 3: Machine Learning Approaches to Improve Cancer Patient Symptom Monitoring & Health Outcomes

Innovations in machine learning, especially when coupled with novel or underutilized data sources, have the potential to transform symptom monitoring of cancer patients and to improve health outcomes. Despite the enormous potential of machine learning, there are challenges limiting its use and adoption within the cancer research and clinical communities. This panel will include presentations from experts in the application of machine learning related to the capture and evaluation of patient-generated health data, including patient-reported outcomes. Panelists will also discuss opportunities and challenges with utilizing machine learning for symptom monitoring and health outcome prediction.

video icon Watch this sessionExternal Web Site Policy

Moderator:

  • Ashley Wilder Smith, Ph.D., M.P.H.
    Outcomes Research Branch
    Healthcare Delivery Research Program
    DCCPS
    NCI
  • Dana L. Wolff-Hughes, Ph.D.
    Risks Factor Assessment Branch
    Epidemiology and Genomics Research Program
    DCCPS
    NCI

Panelists:

  • Chris Sidey-Gibbons, Ph.D.
    MD Anderson Cancer Center
  • Dana Rollison, Ph.D.
    Moffitt Cancer Center
  • I-Chan Huang, Ph.D.
    St. Jude Children's Research Hospital
  • Jukka-Pekka “JP” Onnela, D.Sc.
    Harvard T.H. Chan School of Public Health

Day 1 Wrap Up

video icon Watch this sessionExternal Web Site Policy

Roxanne E. Jensen, Ph.D.
Outcomes Research Branch
Healthcare Delivery Research Program
DCCPS
NCI

Day 2: May 17, 2022

Session 4: Algorithmic Bias & Ethics

This session will explore algorithmic bias and other ethical issues raised by the use of machine learning in cancer and health outcomes research. The speakers will discuss the sources and implications of algorithmic bias and the meaning of algorithmic fairness; identify other potential harms of applying machine learning algorithms in outcomes research and clinical care; and explore ways of preventing and mitigating these harms. The overall goal of this session is to identify key knowledge gaps that need to be addressed in future research.

video icon Watch this sessionExternal Web Site Policy

Moderator:

Paul Han, M.D., M.A., M.P.H.
Office of the Associate Director
Behavioral Research Program
DCCPS
NCI

Patient Perspective: Kimberly Richardson

Panelists:

  • Alex London, Ph.D.
    Carnegie Mellon University
  • David Magnus, Ph.D.
    Stanford University
  • Emma Pierson, Ph.D.
    Jacobs Technion-Cornell Institute at Cornell Tech
  • Tina Hernandez-Boussard, Ph.D., M.P.H.
    Stanford University School of Medicine

Session 5: Unstructured Data: Drawing Meaning Across Domains to Inform Healthcare

As a patient moves through their cancer care journey, they create a wealth of information, much of it in the form of unstructured data. The ability to create meaning from this data represents endless possibilities in better understanding a patient’s journey, both inside and outside the formal healthcare system, and potential to improve care and outcomes for these patients. This panel will explore the variety of domains and the machine learning methodologies used to add meaning to unstructured information that may impact cancer care delivery and health outcome.

video icon Watch this sessionExternal Web Site Policy

Moderators:

Elizabeth Hsu, Ph.D., M.P.H.
Surveillance Informatics Branch
Surveillance Research Program
DCCPS
NCI

Nicole Senft Everson, Ph.D.
Health Communication and Informatics Research Branch
Behavioral Research Program
DCCPS
NCI

Patient Perspective: Kimberly Richardson

Panelists:

  • Guergana Savova, Ph.D.
    Boston Children’s Hospital
  • Jessie Chin, Ph.D.
    University of Illinois at Urbana-Champaign
  • Meliha Yetisgen, Ph.D.
    University of Washington
  • Zhijun Yin, Ph.D., M.S.
    Vanderbilt University Medical Center

Session 6: Machine Learning-Based Decision Support at Point of Care

In this session, we gathered a group of leaders in machine learning-based clinical decision support and cancer care delivery. This panel will provide valuable insights on improving the use of machine learning with clinical decision support tools and how it may be incorporated in clinical workflow to improve cancer care. The panel will provide examples of projects and studies reflecting on what principles and frameworks can be used in research and practice.

video icon Watch this sessionExternal Web Site Policy

Moderator:

Bryan Kim, Ph.D.
Outcomes Research Branch
Healthcare Delivery Research Program
DCCPS
NCI

Patient Perspective: Kimberly Richardson

Panelists:

  • Edwin Lomotan, M.D.
    Agency for Healthcare Research and Quality
  • Kensaku Kawamoto, M.D., Ph.D., M.H.S.
    University of Utah
  • Raj Ratwani, Ph.D.
    MedStar Health Research Institute
  • Ravi Parikh, M.D., M.P.P.
    University of Pennsylvania

Day 2 Wrap Up

video icon Watch this sessionExternal Web Site Policy

Roxanne E. Jensen, Ph.D.
Outcomes Research Branch
Healthcare Delivery Research Program
DCCPS
NCI

For further information or questions about the meeting, contact NCIHDRP@mail.nih.gov

Last Updated: 10 Jan, 2023