The document discusses the evolution of data management and its implications in healthcare, emphasizing the importance of understanding patient trajectories for improved outcomes. Key insights include the transition from data generation to data analysis and the emergence of population health analytics and eHealth biomarkers. The application of machine learning in healthcare aims to enhance decision support, monitor patient conditions, and facilitate preventive measures.

1. RWD, EHRs, PROs.
Using data to inform the
patient trajectory and
experience
Warren A. Kibbe, Ph.D.
Professor, Biostatistics & Bioinformatics
Chief Data Officer, Duke Cancer Institute
warren.kibbe@duke.edu
@wakibbe
#PredictiveModeling
#ConnectedHealth
#PopulationScience
Special thanks to Mike Hogarth, UCSD and NCI for some of the slides

2. Take homes
• Data generation is no longer the
bottleneck – data management,
analysis, reasoning are
• Technology is changing, ubiquitous,
connected. Platform for new insights
• Data science is everywhere
• Move from observation to prediction to
intervention & prevention
• Understanding the patient context, the
patient trajectory, is key

3. Data is pervasive

4. Understanding Patient Trajectories
We are all guilty of
looking under the lamp
post

5. Understanding Patient Trajectories
We are now measuring
a few more points
along the path

6. Understanding Patient Trajectories
But we really need to be able see the whole road

7. Data sources are evolving

8. Types of “Real-World Data”

9. Mobile sensors to enhance monitoring of
effects of new therapies

10. Emergence of ‘eHealth biomarkers’

11. Still validating wearables
https://doi.org/10.1038/s41746-020-0226-6

12. Still validating wearables
https://doi.org/10.1038/s41746-020-0226-6

13. Emergence of “population health
analytics” to measure quality
• Requires very similar infrastructure, tools, and data to
outcomes/pragmatic research with RWD

14. Access to data has changed-Epic

15. From Apple Health App
Note the FHIR JSON source in the third panel. Cool!

16. Issues in Healthcare
• Evidence is not consistently
accessible and structured
• Outcomes are not connected to care
• Patient trajectories are not calculated
or easily accessible

17. Population Health
• More data is ‘digital first’ every day
• Decision aids are needed
• Good UX and responsive computing
and analytics are critical for
improving health outcomes

18. Data Science Hype
Data is the new oil - NOT

19. Data Science Hype
Using data
increases its value

20. Machine Learning and Data Analytics are
mainstream
0
2000
4000
6000
8000
10000
12000
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
Publications in PubMed, 3/23/2020
Machine Learning Deep Learning

21. Applying Machine Learning
• Image analysis at scale, with well
characterized error, uncertainty and confidence
• Support for Molecular Tumor Boards, linking
mutations with evidence for intervention
• Decision Support such as ‘This patient has
melanoma with BRAF V600E mutations.
Consider a targeted therapy like Trametinib
or Vemurafenib’
• Identification of high risk patients, like
‘Your patient is at high risk for re-admittance.
Patients with APACHE scores, MI, and pain
medications have a x in y probability of
remittance based on 2014-2018 data.

22. Applying Machine Learning
• Using –omic and functional data to
understand complex biological
networks involved in homeostasis,
disease, health in the context of life
and healthcare

23. Applying Machine Learning
• Using sensors, IoT devices to
understand and intervene
individually at a national scale before
an acute episode
– Opportunities in prevention, monitoring
for adverse events in patients being
given therapy, behavior and improving
survivorship

24. Questions?
Warren Kibbe, Ph.D.
warren.kibbe@duke.edu
@wakibbe

25. As of January 2019, there are
an estimated
16,900,000
cancer survivors in the U. S.
From https://cancercontrol.cancer.gov/ocs/statistics/statistics.html ,
based on Bluethmann SM, Mariotto AB, Rowland, JH. Anticipating the
''Silver Tsunami'': Prevalence Trajectories and Comorbidity Burden among
Older Cancer Survivors in the United States. Cancer Epidemiol Biomarkers
Prev. (2016) 25:1029-1036

26. In 2030, there will be an
estimated
22,000,000
cancer survivors in the U. S.
From https://doi.org/10.3322/caac.21565 Miller KD, Nogueira L, et al,
Cancer Treatment and Survivorship Statistics, 2019. CA Cancer J Clin
(2019) 0:1-23
Survival, incidence, and all-cause mortality rates were assumed to be
constant from 2016 through 2030.