1. Metadata Capital: Simulating the Predictive Value of Self-Generated Health Information (SGHI)
IEEE Big Data 2014
Jane Greenberg, Adrian Ogletree
CCI/Drexel University,
Metadata Research Center
Angela P. Murillo, Thomas P. Caruso, Herbie Huang,
University of North Carolina at Chapel Hill

2. Metadata Capital Metadata duplication is inefficient, tedious
Metadata as an asset, a product
Reuse of good quality metadata increase value of initial investment
Goals: Discover and advance application of methods for quantifying the cost and value of metadata over time; raise dialog …
M.C – incremental
Successive growth rates
Metadata duplication is inefficient, tedious
An economic concept (Weber, 1905; Smith’s, 1776)
Business and operations (net gains or losses)
Finances, goods and services, and public needs
Intellectual capital (Marr, 2005)
Social capital
a tangible result, value can increase, or…
Modified Capital-sigma notation n
R + ∑ ai = R + a1 + a2 +a3 + …an
i=1

4. Modified Capital-sigma notation
R + ∑ ai = R + a1 + a2 +a3 + …an
i=1
R = value of the metadata record
i= number of usages
a = incremental increase in value
n = maximum number of reuse
Cost / value
Reuse 

5. Modified Capital-sigma notation
DFC April 2013 NSF Review
Modified Capital-sigma notation
Cost / value
Robust metadata reuse a1to a24
Reuse of metadata

6. Cycles… Successive growth rates N ∑ ic = Θ (nc +1) i=1
What about successive growth rate tied to a concept? A concept can be
in ~ vernacular to canonical
fall by the wayside, less popular
out (deprecated)

7. SGHI

9. The Metadata Capital Initiative ~ MetaDataCAPT’L ~
Explore methods for quantifying metadata cost and value over time. Metadata capital targets metadata as an asset containing contextual knowledge about data content
Environments
Ontology development in collaboration with the National Institute for Environmental Health Sciences (NIEHS).
Self-generated health information (SGHI) monitoring daily activity in collaboration with the Research Triangle Institute (RTI).

10. Advance nascent work on “metadata capital” for data science
Discover and advance the application of methods for quantifying the cost and value of metadata over time; raise dialog
Advance nascent work on “metadata capital” for data science
Actively engage with the NCDS community
The Team
3. Connect NCDS metadata efforts w/the Research Data Alliance

11. Self Generated Health Information (SGHI)
“SGHI or Self-Generated Health Information is information generated by mobile health (mHealth apps), wearables and smartwatches.  Quantified Self efforts tend to provide examples of the use of this information.” (T. Caruso)

12. facets Data Vendors/brands BodyMedia DailyMil FatSecret Fitbit Fitbug
Validic API Category
“simple, standardized connection between healthcare companies and mobile health...” 
Vendors/brands
BodyMedia
DailyMil
FatSecret
Fitbit
Fitbug
Fleetly
Garmin
Glooko
iHealth
Jawbone Up
ManageBGL
MapMyFitness
Moveable
MovesApp
MyGlucoHealth
Nike+
Omron
RunKeeper
Strava
VitaDock
Withings
Fitness
Routine
Nutrition
Sleep
Weight
Diabetes
Biometrics
Tobacco Cessation

13. metadata Most popular timestamp type start_time distance duration
calories
utc_offset
metadata
Total Fields Referenced (FitBit), toward SGHIx
X Availale: 39
P (Pending): 3
NA (not available): 42
(Caruso & Ogletree)

14. Conclusion…other Valuation Approaches
Market cap of Facebook per user: $40 – $300
Revenues per record per user: $4 – $7 per year
Facebook
Experian
Market prices of personal data:
$0.50 for street address
$2.00 for date of birth
$8 for social security number
$3 for driver’s license number
$35 for military record
SOURCE: OECD. Exploring the Economics of Personal Data: A Survey of Methodologies for Measuring Monetary Value. OECD Digital Economy Papers. Office for Economic Cooperation and Development Publishing, 2013.

16. h
In the Fitbit data scenario, if a patient’s exercise data and environmental quality data can be combined with asthma condition data, we will get a better prediction of the way in which asthma evolves. Prediction error in the model when including the Fitbit data

17. Behavioral changes vs. asthma condition evolution.

18. Limitations
Modified capital-sigma is only one dimensional; all metadata properties/concept are not equal
Also, we know cost/value relationship is not 1:1.
Metadata is only as good as your data
not always true
What about successive growth rate may be the way to go

19. Concluding remarks Interest….traction
Limitations: bad data, cost/value, more metadata
We should care about cost
Metadata capital can contextualize the discussion, provide a foundation
Generic formula for further research
Proof

20. The Team / acknowledgments
Tom Caruso, Health Information
Liaison Research Associate, UNC-SILS/RTI
Self-generated Health Information (SGHI)
Rebecca Boyles, Data Scientist, NIEHS
Common Core Vocabulary
Jane Greenberg, SILS/UNC, MRC
Herbie Huang, Ph.D. student, Economics Dep. UNC
Austin Mathews, BSIS student, SILS/UNC, MRC
Angela Murillo, Ph.D. student, ,SILS/UNC, MRC
Adrian Ogletree, MSIS student, SILS/UNC, MRC
Erik Scott, Senior Software Dev./RENCI (Renaissance Computing Institute)