Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Precise Evidence for Specific Problems
@ehekler
Dr. Eric Hekler
Arizona State University
August 4, 2016
Outline
• Motivations & perspective
• Precise solutions
• Precise evidence
• Agile science (v.2)
• Citizen-led science & P...
Motivations & Perspective
Human Genome ProjectWalking on the Moon
Penicillin
Eric Hekler, @ehekler
theamazingworldofgumball.wikia.com
http://www.gen...
http://youtu.be/QPKKQnijnsM
Flickr – just.Luc
Flickr-meanMrmustard
Behaviors explain most variability in health
Flickr – Stuck in Customs@ehekler
40
15
5
10
30
Sub-Optimal Health behaviors
...
Behavior at the center
Hovell M, Wahlgren D, Adams M. Emerging theories in health promotion practice and research. 2009;2:...
Core problem: Skeumorphisms
Schueller et al. 2013
Precise Solutions
Personal, pervasive, & powerful technologies
Flickr – Stuck in CustomsPatrick, Hekler, Estrin, Godino, Crane, Riper, & Moh...
@ehekler http://www.nih.gov/precisionmedicine/
Just in Time Adaptive Interventions
@ehekler
Just in time: State of vulnerability
Flickr - Rob Marquardt
@ehekler Nahum-Shani, Hekler, & Spruijt-Metz, (2015) Health Ps...
Just in time: State of opportunity
Flickr - Miroslav Petrasko
@ehekler Nahum-Shani, Hekler, & Spruijt-Metz, (2015) Health ...
Just in time: Receptive
Flickr-Jonathan Powell
Nahum-Shani, Hekler, & Spruijt-Metz, (2015) Health Psychology@ehekler
Adaptive: Series of “just in time” moments
@ehekler Flickr - Dave Gray
System controlled
“Giving the fish”
NSF IIS-1449751: EAGER: Defining a Dynamical Behavioral
Model to Support a Just in Tim...
Modeling behavior
Riley, Martin, Rivera, Hekler, et al. 2016; Martin, Riley, Rivera, Hekler, et al. 2014@ehekler
Three Example Individualized Computational Models via Black-Box
System ID: Goals-Expected Points-Granted Points model; B: ...
Future-oriented predictions
Hekler, et al. 2013 Health Education and Behavior@ehekler
Martin, Rivera, & Hekler Manuscript Submitted for Publication
Future-oriented decisions
@ehekler
Individual controlled
“Teaching to fish”
Eric Hekler, Jisoo Lee, Erin Walker, Winslow Burleson, Arizona State University; ...
Measure
success
towards
goal
Results
Self-experimentation
Plan
+ Implement for 1 week
@ehekler
MS Wearables 101 Course
Emil Chiauzzi, PatientsLikeMe
Eric Hekler, Arizona State University
Pronabesh DasMahapatra, Patien...
Precise Evidence
Specific Solutions
for Specific Problems
Design &
Engineering
“On Average”
Science
“On Average” Evidence
for General Probl...
Specific Solutions
for Specific Problems
Design &
Engineering
“On Average”
Science
“On Average” Evidence
for General Probl...
Specific Solutions
for Specific Problems
Design &
Engineering
“On Average”
Science
“On Average” Evidence
for General Probl...
Subjectivity matters
@ehekler
• Solving the “last mile” problem
• Requires a damn good designer
• AND(/OR?) patient empowe...
From “on average” to algorithms
@ehekler
• From generally true to true for me
• Requires acknowledging variance
“On Averag...
Role of the professional may change
@ehekler
• From solving to empowering
• Professional can support
– Education
– Tool bu...
Agile Science
friko-diamondsdesigns.blogspot.com
HealthFoo, December 2013:
https://www.youtube.com/watch?v=wY-stOXqmuw
Wat...
Agile science products
• Modules
• Computational models
• Personalization algorithms
@ehekler
Modules
Smallest, meaningful, self-contained,& repurposable
“Perfect” intervention package Components
Flickr - Paul Swanse...
Modules
@ehekler
Inputs Process Output
Proximity sensor module
@ehekler
Inputs
iBeacons
Phone
Meta-data
Process
Transform
tagged data
into a time-
stamped db
Out...
Modules
APIs
www.yelp.com@ehekler
IFTTT
http://www.ifttt.com
Modules
Templates
www.ifttt.com@ehekler
Modules
http://www.ifttt.comwww.ifttt.com@ehekler
Computational models
Riley, Martin, Rivera, Hekler, et al. 2016; Martin, Riley, Rivera, Hekler, et al. 2014@ehekler
Computational models: Ontologies
Larsen, Michie, Hekler, et el. 2016, Journal of Behavioral Medicine@ehekler
Personalization algorithms
www.netflix.com@ehekler
Martin, Rivera, & Hekler Am. Control Conference (2015)
Personalization algorithms
@ehekler
Agile Science Process v0.2
@ehekler
Agile Science
Process
Generate
Design & engineer specific solutions for
specific problems
@ehekler
Formative research
• Defining a “niche”
• Defining constraints
• Generating solutions
IDEO
“Niche” specification
IDEO: Human-Centered Design Kit
Design constraints
DesignWhat fits in time &
budget constraints
What does your
target audience
think they need?
What do th...
Generating solutions
Stanford d.School, Bootleg Bootcamp
Complexity mapping
• Finding assumptions
• Defining causal
pathways
• Defining a research
agenda
Finding assumptions via simulation
Martin, Rivera, & Hekler (In preparation)
Finding assumptions via simulation
Martin, Rivera, & Hekler (In preparation)
Causal pathways
Antecedents
Body
Movement ConsequencesContext (People, place,
time)
Time
scale
Year
Month
Day
Hour
Min
Bou...
Research agenda
Prototyping
• Testing “hunches”
• Testing assumptions
• Examining feasibility
Amy Luginbill; Samantha Quagliano; Sepideh Zohreh
S=Stop
M=Move
I= I statement; I can do it!
L=Love (positivity)
E=Exhale
...
Testing assumptions
John Harlow, Erik Johnston, Zoe Yeh@ehekler
Phoenix Proposition 104
John Harlow, Erik Johnston, Zoe Yeh@ehekler http://movephx.org/get-the-facts/maps/
Examining feasibility
https://www.youtube.com/watch?v=xy9nSnalvPc
Evaluate
Determine the “boundary conditions” on when,
where, for whom, and in what state a tool
produces its desired outco...
Linda M. Collins
The Methodology Center
Penn State
methodology.psu.edu@ehekler
Micro-randomization design
• Sequential, full factorial designs
• Randomize intervention component
• Each time we might de...
Dynamic hypotheses- “sweet spot”
Hekler (PI), Rivera (Co-PI), NSF IIS-1449751
-15
-10
-5
0
5
10
15
20
0
2000
4000
6000
800...
System identification experiments
-100
100
300
500
700
900
1100
1300
1500
0
2000
4000
6000
8000
10000
12000
14000
1 8 15 2...
Curate
Evidence-based insights for match-making of
specific solutions to specific problems
@ehekler
Ontologies
Larsen, Michie, Hekler et al. in press
Shared test-beds
@ehekler
PatientsLikeMe
@ehekler
Research Kit
https://www.apple.com/ios/whats-new/health/ http://researchkit.github.io/ http://sagebase.org/
Paco
www.pacoapp.com@ehekler
Open Humans
@ehekler
eEcosphere
@ehekler DISCLAIMER: On scientific advisory board w/ equity stakes in the company
Patient-led science @PLM
PLM is a true pioneer (as you know ;)
Specific Solutions
for Specific Problems
Design &
Engineering
“On Average”
Science
“On Average” Evidence
for General Probl...
OpenAPS
OpenAPS
End-User design, engineering, & science
• Target: empowering systematic patient “hacking”
– Disease management (e.g., MS “...
What you get?
• Insights on the “last mile” problem
• Highly marketable (?)
• Strong value back to your patients
Advocating for culture change
• Target: shifting social, ethical, methodological, and
regulatory change to embrace patient...
Specific Solutions
for Specific Problems
Design &
Engineering
“On Average”
Science
“On Average” Evidence
for General Probl...
Thanks! What can we build together?
Dr. Eric Hekler, Arizona State University
ehekler@asu.edu, @ehekler
TARGET: Precision behavior change
Individual/User
Controlled
System
Controlled
Individual/System
Balanced Control
@ehekler
Why now? Behavioral meteorology
Flickr-Bart Everson
Patrick, Riley, Estrin, Hekler, Godino, Crane, Riper, & Mohr, Manuscri...
Why now? The world needs us…
Flickr – Stuck in Customs
http://youtu.be/QPKKQnijnsM
Flickr – just.Luc
Flickr-meanMrmustard
First step…
@ehekler
Stop building “perfect”
packages…
Start building interoperable
modules
Flickr - Paul Swansen Flickr -...
Interoperable systems
@ehekler
LeadSecondary
Secondary
Secondary
SecondarySecondary
Interoperable systems
www.openmhealth.org
Ecologically-valid data streams
@ehekler
Lead Secondary
Secondary
Co-Lead
Turning “noise” into information
https://ubicomplab.cs.washington.edu/
Data standardization
@ehekler
LeadCo-Lead
Secondary
Secondary
Secondary
Data standardization
www.openmhealth.org
Precise Evidence for Specific Problems
Precise Evidence for Specific Problems
Precise Evidence for Specific Problems
Upcoming SlideShare
Loading in …5
×

Precise Evidence for Specific Problems

146 views

Published on

This talk, given to PatientsLikeMe, discusses how science can move from "on average" insights to evidence that provides answer for specific individuals.

Published in: Science
  • Be the first to comment

  • Be the first to like this

Precise Evidence for Specific Problems

  1. 1. Precise Evidence for Specific Problems @ehekler Dr. Eric Hekler Arizona State University August 4, 2016
  2. 2. Outline • Motivations & perspective • Precise solutions • Precise evidence • Agile science (v.2) • Citizen-led science & PLM @ehekler
  3. 3. Motivations & Perspective
  4. 4. Human Genome ProjectWalking on the Moon Penicillin Eric Hekler, @ehekler theamazingworldofgumball.wikia.com http://www.genome.gov/
  5. 5. http://youtu.be/QPKKQnijnsM Flickr – just.Luc Flickr-meanMrmustard
  6. 6. Behaviors explain most variability in health Flickr – Stuck in Customs@ehekler 40 15 5 10 30 Sub-Optimal Health behaviors Social Circumstances Environmental Exposures Healthcare Genetics McGinnis, et al. 2002 Health Affairs
  7. 7. Behavior at the center Hovell M, Wahlgren D, Adams M. Emerging theories in health promotion practice and research. 2009;2:347-85.@ehekler
  8. 8. Core problem: Skeumorphisms Schueller et al. 2013
  9. 9. Precise Solutions
  10. 10. Personal, pervasive, & powerful technologies Flickr – Stuck in CustomsPatrick, Hekler, Estrin, Godino, Crane, Riper, & Mohr, Riley, Manuscript in Prep@ehekler
  11. 11. @ehekler http://www.nih.gov/precisionmedicine/
  12. 12. Just in Time Adaptive Interventions @ehekler
  13. 13. Just in time: State of vulnerability Flickr - Rob Marquardt @ehekler Nahum-Shani, Hekler, & Spruijt-Metz, (2015) Health Psychology
  14. 14. Just in time: State of opportunity Flickr - Miroslav Petrasko @ehekler Nahum-Shani, Hekler, & Spruijt-Metz, (2015) Health Psychology
  15. 15. Just in time: Receptive Flickr-Jonathan Powell Nahum-Shani, Hekler, & Spruijt-Metz, (2015) Health Psychology@ehekler
  16. 16. Adaptive: Series of “just in time” moments @ehekler Flickr - Dave Gray
  17. 17. System controlled “Giving the fish” NSF IIS-1449751: EAGER: Defining a Dynamical Behavioral Model to Support a Just in Time Adaptive Intervention, PIs, Hekler & Rivera @ehekler
  18. 18. Modeling behavior Riley, Martin, Rivera, Hekler, et al. 2016; Martin, Riley, Rivera, Hekler, et al. 2014@ehekler
  19. 19. Three Example Individualized Computational Models via Black-Box System ID: Goals-Expected Points-Granted Points model; B: Predicted Busyness; S: Predicted Stress; T: Predicted Typical; W: Weekday-Weekend Modeling differences
  20. 20. Future-oriented predictions Hekler, et al. 2013 Health Education and Behavior@ehekler
  21. 21. Martin, Rivera, & Hekler Manuscript Submitted for Publication Future-oriented decisions @ehekler
  22. 22. Individual controlled “Teaching to fish” Eric Hekler, Jisoo Lee, Erin Walker, Winslow Burleson, Arizona State University; Bob Evans, Google Flickr Juhan Sonin @ehekler
  23. 23. Measure success towards goal Results Self-experimentation Plan + Implement for 1 week @ehekler
  24. 24. MS Wearables 101 Course Emil Chiauzzi, PatientsLikeMe Eric Hekler, Arizona State University Pronabesh DasMahapatra, PatientsLikeMe
  25. 25. Precise Evidence
  26. 26. Specific Solutions for Specific Problems Design & Engineering “On Average” Science “On Average” Evidence for General Problems Key Traditional pathway Emerging pathway Product Process Professional-led
  27. 27. Specific Solutions for Specific Problems Design & Engineering “On Average” Science “On Average” Evidence for General Problems Key Traditional pathway Emerging pathway Product Process Precise Evidence for Specific Problems Personalization Algorithm Science Professional-led
  28. 28. Specific Solutions for Specific Problems Design & Engineering “On Average” Science “On Average” Evidence for General Problems Key Traditional pathway Emerging pathway Product Process Precise Evidence for Specific Problems Personalization Algorithm Science Professional-led Citizen/Patient-led
  29. 29. Subjectivity matters @ehekler • Solving the “last mile” problem • Requires a damn good designer • AND(/OR?) patient empowerment . Mullainathan S. Solving social problems with a nudge. TEDIndia. 2009. http://www.ted.com/talks/sendhil_mullainathan.
  30. 30. From “on average” to algorithms @ehekler • From generally true to true for me • Requires acknowledging variance “On Average” ~50% “Personalization/Matchma king” ~35% Idiosyncratic/ Subjective ~15%
  31. 31. Role of the professional may change @ehekler • From solving to empowering • Professional can support – Education – Tool building – Communication – Curation
  32. 32. Agile Science friko-diamondsdesigns.blogspot.com HealthFoo, December 2013: https://www.youtube.com/watch?v=wY-stOXqmuw Watch this video on being a “thought leader”: https://www.youtube.com/watch?v=_ZBKX-6Gz6A
  33. 33. Agile science products • Modules • Computational models • Personalization algorithms @ehekler
  34. 34. Modules Smallest, meaningful, self-contained,& repurposable “Perfect” intervention package Components Flickr - Paul Swansen Flickr - Benjamin Esham @ehekler
  35. 35. Modules @ehekler Inputs Process Output
  36. 36. Proximity sensor module @ehekler Inputs iBeacons Phone Meta-data Process Transform tagged data into a time- stamped db Output Time- stamped csv of indoor location
  37. 37. Modules APIs www.yelp.com@ehekler
  38. 38. IFTTT http://www.ifttt.com Modules Templates www.ifttt.com@ehekler
  39. 39. Modules http://www.ifttt.comwww.ifttt.com@ehekler
  40. 40. Computational models Riley, Martin, Rivera, Hekler, et al. 2016; Martin, Riley, Rivera, Hekler, et al. 2014@ehekler
  41. 41. Computational models: Ontologies Larsen, Michie, Hekler, et el. 2016, Journal of Behavioral Medicine@ehekler
  42. 42. Personalization algorithms www.netflix.com@ehekler
  43. 43. Martin, Rivera, & Hekler Am. Control Conference (2015) Personalization algorithms @ehekler
  44. 44. Agile Science Process v0.2
  45. 45. @ehekler Agile Science Process
  46. 46. Generate Design & engineer specific solutions for specific problems @ehekler
  47. 47. Formative research • Defining a “niche” • Defining constraints • Generating solutions IDEO
  48. 48. “Niche” specification IDEO: Human-Centered Design Kit
  49. 49. Design constraints DesignWhat fits in time & budget constraints What does your target audience think they need? What do the “experts” think your target audience needs? How might policy impact what is possible? What does your target audience ACTUALLY need? What’s fundable? What’s feasible to build? What ideas can be evaluated? What’s something that can be sustained during & after the study? What can you actually build with your tech partner?
  50. 50. Generating solutions Stanford d.School, Bootleg Bootcamp
  51. 51. Complexity mapping • Finding assumptions • Defining causal pathways • Defining a research agenda
  52. 52. Finding assumptions via simulation Martin, Rivera, & Hekler (In preparation)
  53. 53. Finding assumptions via simulation Martin, Rivera, & Hekler (In preparation)
  54. 54. Causal pathways Antecedents Body Movement ConsequencesContext (People, place, time) Time scale Year Month Day Hour Min Bouts of MVPA Min/day MVPADaily min/day goal of MVPA Cardiovascular Fitness (vO2) Self- Management Skills Self-Identity as an “exerciser” Atherosclerotic Plaque Prevention
  55. 55. Research agenda
  56. 56. Prototyping • Testing “hunches” • Testing assumptions • Examining feasibility
  57. 57. Amy Luginbill; Samantha Quagliano; Sepideh Zohreh S=Stop M=Move I= I statement; I can do it! L=Love (positivity) E=Exhale SMS: “If you are stressed today, try one of the following options, Deep breathing, Stretching, get up move around.” MOBILECAR MAIDSERVICES GREEN CLEAN Prototype 1: S.M.I.L.E. Prototype 2: Facial Wave Prototype 3: SMS Intervention Prototype 4: De-stress your carTesting hunches @ehekler
  58. 58. Testing assumptions John Harlow, Erik Johnston, Zoe Yeh@ehekler
  59. 59. Phoenix Proposition 104 John Harlow, Erik Johnston, Zoe Yeh@ehekler http://movephx.org/get-the-facts/maps/
  60. 60. Examining feasibility https://www.youtube.com/watch?v=xy9nSnalvPc
  61. 61. Evaluate Determine the “boundary conditions” on when, where, for whom, and in what state a tool produces its desired outcome. @ehekler
  62. 62. Linda M. Collins The Methodology Center Penn State methodology.psu.edu@ehekler
  63. 63. Micro-randomization design • Sequential, full factorial designs • Randomize intervention component • Each time we might deliver component • Multiple components can be randomized • Randomized 100s or 1000s of times Klasnja, Hekler, Shiffman, Boruvka, Almirall, Tewari, Murphy, Health Psych, 2015@ehekler
  64. 64. Dynamic hypotheses- “sweet spot” Hekler (PI), Rivera (Co-PI), NSF IIS-1449751 -15 -10 -5 0 5 10 15 20 0 2000 4000 6000 8000 10000 12000 14000 AveChangeSelfEffficacy ActualDailySteps Recommended Goal Actual Steps Δ Self-Efficacy @ehekler
  65. 65. System identification experiments -100 100 300 500 700 900 1100 1300 1500 0 2000 4000 6000 8000 10000 12000 14000 1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 Points Stepsperday Days Points Provided (100, 300, 500) Fictionalized actual steps per day Daily step goal ((Baseline Median) to (Baseline Median+100% Baseline Median)) NSF IIS-1449751: Defining a Dynamical Behavioral Model to Support a Just in Time Adaptive Intervention, PIs, Hekler & Rivera@ehekler
  66. 66. Curate Evidence-based insights for match-making of specific solutions to specific problems @ehekler
  67. 67. Ontologies Larsen, Michie, Hekler et al. in press
  68. 68. Shared test-beds @ehekler
  69. 69. PatientsLikeMe @ehekler
  70. 70. Research Kit https://www.apple.com/ios/whats-new/health/ http://researchkit.github.io/ http://sagebase.org/
  71. 71. Paco www.pacoapp.com@ehekler
  72. 72. Open Humans @ehekler
  73. 73. eEcosphere @ehekler DISCLAIMER: On scientific advisory board w/ equity stakes in the company
  74. 74. Patient-led science @PLM
  75. 75. PLM is a true pioneer (as you know ;)
  76. 76. Specific Solutions for Specific Problems Design & Engineering “On Average” Science “On Average” Evidence for General Problems Key Traditional pathway Emerging pathway Product Process Precise Evidence for Specific Problems Personalization Algorithm Science Professional-led Citizen/Patient-led
  77. 77. OpenAPS
  78. 78. OpenAPS
  79. 79. End-User design, engineering, & science • Target: empowering systematic patient “hacking” – Disease management (e.g., MS “sweet spot” study) – Next gen drugs (e.g., Lithium study v2.0) – Next gen medical devices (e.g., OpenAPS) • Courses on patient-led design, engineering & science • End-user programming tools (e.g., Paco) to empower patient-led design, engineering & science
  80. 80. What you get? • Insights on the “last mile” problem • Highly marketable (?) • Strong value back to your patients
  81. 81. Advocating for culture change • Target: shifting social, ethical, methodological, and regulatory change to embrace patient-led design, engineering, and science • Devise a pathway through the FDA – OpenAPS • Build communication pathways between patient- innovators and professionals – OpenAPS
  82. 82. Specific Solutions for Specific Problems Design & Engineering “On Average” Science “On Average” Evidence for General Problems Key Traditional pathway Emerging pathway Product Process Precise Evidence for Specific Problems Personalization Algorithm Science Professional-led Citizen/Patient-led
  83. 83. Thanks! What can we build together? Dr. Eric Hekler, Arizona State University ehekler@asu.edu, @ehekler
  84. 84. TARGET: Precision behavior change Individual/User Controlled System Controlled Individual/System Balanced Control @ehekler
  85. 85. Why now? Behavioral meteorology Flickr-Bart Everson Patrick, Riley, Estrin, Hekler, Godino, Crane, Riper, & Mohr, Manuscript in Prep@ehekler
  86. 86. Why now? The world needs us… Flickr – Stuck in Customs http://youtu.be/QPKKQnijnsM Flickr – just.Luc Flickr-meanMrmustard
  87. 87. First step… @ehekler Stop building “perfect” packages… Start building interoperable modules Flickr - Paul Swansen Flickr - Benjamin Esham www.agilescience.org
  88. 88. Interoperable systems @ehekler LeadSecondary Secondary Secondary SecondarySecondary
  89. 89. Interoperable systems www.openmhealth.org
  90. 90. Ecologically-valid data streams @ehekler Lead Secondary Secondary Co-Lead
  91. 91. Turning “noise” into information https://ubicomplab.cs.washington.edu/
  92. 92. Data standardization @ehekler LeadCo-Lead Secondary Secondary Secondary
  93. 93. Data standardization www.openmhealth.org

×