Innovation, data science & risk in healthcare
Bern Shen MD
HISA Health Data Analytics
Brisbane 11 Oct 2017
2
Policies & interventions
Health status
Access to quality healthcare
Technology
Physical environment
Behavior
Biology
Social environment
Individual
Adapted from Healthy People 2010
Source: : Mokdad, et al. 2004. Actual causes of death in the United States, 2000. JAMA. 2004;291:1238-45.
3
See Understand Predict Control
“Illness is about learning to live with lost control.”- Arthur Frank. The Wounded Storyteller: Body, Illness & Ethics.
“It may not be dying we fear so much, but the diminished self.”- Anatole Broyard. Intoxicated by My Illness.
Analytics & algorithms are beautiful…
…but only useful if they effect benefit in the real world.
4
(DCF startup valuation)
(SVM classifier for machine learning)
5Population Health FrameworkSource: Care Continuum Alliance. Outcomes Guidelines Report, Vol. 5. Washington, DC: Care Continuum Alliance. 2010.
Translation, implementation, dissemination
• NIH National Center for Advancing Translational Sciences (2011)
• Centres in Australia including Queensland Translational Research Institute
6
7
Data science
Risk Innovation
Top risks
“Second on the list is the one we haven’t thought of, and at the very top is the one we can’t imagine.” – David Morens, US NIAID
8Source: : Mokdad, et al. 2004. Actual causes of death in the United States, 2000. JAMA. 2004;291:1238-45.
9
Adapt (anticipate?) or die
10
Identify the tree shrew
Or better, be the next apex predator
Context & Pasteur’s prepared mind
“Dans les sciences d'observation le hasard ne favorise que des esprits préparés.”-Louis Pasteur, 1854
11
Data points (analytical) → data clouds (ecological)
12
Data points (analytical) → data clouds (ecological)
13Time
Disease signalDisease driver signal
Signal strength
Disease detected
Disease anticipated
(Noise)
Poverty, social inequalityWeather, climate changeMalnutrition, famineCrowding, human/wildlife contactLand use/ecosystem changeDisaster…
“Complexity rheostat”
14
Doctor
Patient
Med
Similarly for diagnostics, devices, services…
Decision support, practice guidelines, care pathways, etc.
Adherence, health beliefs & behaviors, social determinants, etc.
15
Map of science derived from
clickstream dataBollen J, Van de Sompel H, Hagberg A, Bettencourt L, Chute R, et al. (2009) Clickstream Data Yields High-Resolution Maps of Science. PLoS ONE 4(3): e4803. doi:10.1371/journal.pone.0004803 http://127.0.0.1:8081/plosone/article?id=info:doi/10.1371/journal.pone.0004803
Geographic → genomic
16Source: http://rocs.hu-berlin.de/complex_sys_2015/resources/Presentations/C_Mueller.pdf
a
Kwang-Il Goh, and In-Geol Choi Briefings in Functional Genomics 2012;11:533-542
Human disease network graph
17
• 67% of disorders linked to
at least one other
• Giant cluster contains 516
of 1284 (40%)
18
Health risks…
19Source: Univ. of Washington IHME. http://vizhub.healthdata.org/gbd-compare/
…change over time...
20
1990 2010
Source: R Lozano, et al. 2012. Global & regional mortality from 235 causes of death for 20 age groups in 1990 & 2010. Lancet 380:2095-128.
…with age...
21Source: R Lozano, et al. 2012. Global & regional mortality from 235 causes of death for 20 age groups in 1990 & 2010. Lancet 380:2095-128.
…by gender & place…
22
…by wealth & place
23
Malaria
Neonatal sepsis
TB
HIV/AIDS
Diarrhea
Malnutrition
Meningitis
Interpersonal violence
Pancreatic cancer
Breast cancer
Dementias
Prostate cancer
Diabetes
Neck pain
Lung cancer
AnxietyDrug use
Colon cancerAlcohol use
Self-harm
Lung cancer
Alcohol use
Colon cancer
Stomach cancer
Forces of nature in Caribbean
Wealthier Poorer
Global
Liver cancer in China
Interpersonal violence in Central & South America
Technobiome
24
“A House is a machine for living in.”-Le Corbusier. Vers Une Architecture. 1923.
“… joining and separation of human and nonhuman are everyday affairs.”-Suchman. Human-Machine Reconfigurations. 2007.
Unintended consequences, ethics
“We are building a civilization that is deeply connected yet technologically insecure…
in other words, we are constructing a world that is wired for crime.”
25
Dealing with bad data
• Unintended
• Deliberate
26
Toward Precision Medicine: Building a knowledge network for biomedical research & a new taxonomy of disease. National Academies Press, 2011.
New data, new connections
27
• New data types Beyond usual text, tracings, & images
• New data sources Outside of the hospital, clinic & lab
• …create new information management challenges
28
29
Data (test) Insight (diagnose) Action (treat)
30