Skip to content

Latest commit

 

History

History
195 lines (177 loc) · 16.1 KB

agenda.md

File metadata and controls

195 lines (177 loc) · 16.1 KB

DIF Healthcare Special-Interest Group Running Agenda & Notes

hackmd-github-sync-badge

{ mailing list | zoom link }

February 24, 2021 - ??

February 10, 2021 - Note new time: 7am Pacific, 4pm CET - FHIR and OCA (with some talk about CCI& VCI)

  • Guests: Burak Serdar (Cloud Privacy Labs) and John Walker (Linux Foundation Public Health/Covid Credentials Initiative)
  • Recording

Topics discussed:

  • Layman's history of and intro to FHIR, SMART on FHIR, and LH7 protocols
  • Description of the fundamental technical problem: getting secure, trustworthy medical records out of an EHR system, via FHIR, into a maximally portable VC that's just as trustworthy (while not the system of record for that patient)
  • Two approaches prototyped by the same task force at ToIP:
    • VCI: FHIR --> LD --> OCA --> VC format of choice (depicted in minutes)
    • Burak's more end-to-end solution, allowing trust authorities to specify the conversion from FHIR to VC Schemata of their design via OCA
  • Useful Links:
Minutes

John's CCI/LFPH prototype:

Burak's prototypes for extending OCA to specify FHIR extraction:

  • John Walker
    • JSON --> JSON-LD --> OCA
  • Burak Serdar
    • OCA originally conceived as a data CAPTURE mechanism, but I have been involved for a while applying it to privacy use cases
  • John's overview of FHIR
    • protocol for HL7 (4th iteration of it)
    • JSON/JSON Schema-based
    • EPIC and Cerner (sp?) are actually more installed in the wild; FHIR still a minority pipe-separated messaging structure
    • EHR and Health Info Exchanges use it as best-practice messaging protocol
      • SMART on FHIR (boston group and Josh Mandell)- OAuth2 + JSON interface and transport for FHIR
    • Our work starts from the assumption that personal health info is accessible via SMART on FHIR
      • extracting and exchanging data in
    • As with any open community, there's tons of development happening here; one strategy that we have decided on is to leverage validation and data structure definitions, and not reinvent any of this or create new opacity
  • VCI
    • proposed architecture SMART on FHIR endpoint with FHIR endpoint
  • Data Mechanics
    • OCA consumes the data and applies overlays and hashes and locks it to be tamperproof
  • Burak's overview of [OCA work] and FHIR work at ToIP
  • John - LD versus JWT - We're working on FHRI --> LD --> JWTs
  • Q and A
    • Programmer could convert to and from LD the programmer could do on the server side? Does OCA enable client-side ZKP?
      • 2 use cases:
      1. EHR --> LD (apply your own schemata, apply ZKP) --> credentials (you'd need an agent before it goes to a wallet); there's value in making a generalized approach for transparently and accurately get health records into an LD format and out of "FHIRland" - tamperproofed
      2. ??
      • Burak:
      1. OCA can be presentation layer on extracted FHIR data
      2. But there's another use case where OCA can help translate server-to-server
      • John: I might download my own health records and put them in a POD, or a PDS-- how do I get my health records into my own controlled storage;
    • John: Once you pull data out of EHR, it's read-only (that's the authoritative system of record)-- we just want to give people control over their own copy of that
      • what about the deltas? That's further down the roadmap
    • Burak: Complementary approach: extending OCA to "Projection" - more "end to end", allows governing authorities to define not just the LD Schema for issuance of creds but also the FHIR conversation end to end
    • Schema work - Ontology happening (early days; Mattr leading it?)
  • Stephan - FHIR has no idea of doctor's DID and patient's DID
    • Stephan: Josh's VCI is poking in that direction: ISO unification of business processes could hold EHR and hospitals to best-practices of switching their ID records to SSI envelopes and identities and tooling...
    • Stephan: Vaccination might be too shiny of an object to go after-- if this isn't done by June, will it go anywhere? What about other use-cases? What about health proofs for life insurance? Do you have other use-cases in mind for self-managed health proofs?
      • John: We're not exploring them very actively, but we're open to them... (follow-up session!)
    • Contacts: bserdar at computer dot org and johnw dot cci at lfph dot io

January 27, 2021 - Note new time: 7am Pacific, 4pm CET

  • Recording
  • Overview of the problem space (smart contracts and trust ecosystems)
    • B2B: DSCSA (Spherity!); Allied Clinicial Access (?), Federated Learning Coordination (OpenMined?)
    • B2C/E: Vaccination Proofs; Ecosystem-Consumer; Clinician Credentialling; eligibility/proof of coverage
    • B2G: Covid19 reporting
  • What's missing?
    • Auth (whence siop?); lightweight API in a sound JS library would accelerate B2B exploration;
  • Q&A: Tom Jones: patient matching as a blocker left off this list; Foster bill; CA blockchain alliance; Kantara work & Jim is also involved with a Medicaid group

List of topics and guests 2020:

Meetings

Minutes
  • Self introduction:
    • Consultant for the Scottish govt - advise startups and facilitate business thinking
    • JLinc - Business side
    • Very active in MyData -
    • IIW
  • Intro - Data grabs & Surveillance Capitalism - Most but not all
    • Confidential data - there are business niches where people DON'T want as much data as possible!
  • MyData Commons
    • Blog post - Covid as trial by fire for our data ecosystem - we need global citizenship management process, but we don't have the infra for that
    • Thesis early on: We need a commons, won't have it soon, but maybe in time for Covid-20
    • April-may-june - MyData team in a Covid Hackathon - Personal data store for each individual - used JLinc for SSI and a SalesForce instance (had both on hand)
    • Lessons from multiple hackathons
  • Massive data problem- how can public health authorities get data about the asymptomatic, the healthy, and the mildly symptomatic? How can you get people to donate helpful data at scale?
    • Symptom-tracking apps (i.e. DataYogi.me,
    • Dummy data+SalesForce —> population-level analytics
    • Integration of wallets and credentials (Digital Wallet JLinc integration) - just a prototype, not taken to market
  • MyData alignment/interop
    • MyData - working on shared Schema for Operators group - a dictionary that can be consumed as LD, schema.org, OCA
  • Last 6 months have taught me that
    • 3 separate clients all coming to : current clients - 1 health, 2 in less sensitive areas
      • leave data on device
      • get ML models on depersonalized/anonymized
      • openMined - analyze data without it leaving the subject's device
    • "Notice & Consent doesn't scale" —John Wunderlich (JLinc, IIW)
    • realtime co-management of data
    • multi-directional, multi-stakeholder CRUD on all data
    • "personal data logistics"
  • Q and A
    • Bernard: Data quality in Covid Commons?
      • Iain: Medics designed the system (HL7/FHIR)
      • Iain: Self-Managed data tends to have lower incentive problems - individual managing own data and only-once-ing incentivizes honesty
    • Bob: How does patient control their data sharing?
      • Iain: SISA (written to ledger/immutable store!) —> contract law as enforcement rather than technical control of data
      • Iain: JLinc's early work a chain of confidentially
    • Bob: Accord project and CommonAccord? Attempting to take common legal language and standardize on it, make it more machine-readable (IEEE WG on machine-readable contracts as well)
    • Iain: Individual can come to the info-sharing counterparty as a peer ;
      • data quality and compliance are shot in status quo
      • FAANG only people making money in current system
    • Juan: PDS?
      • Iain: Covid is driving home the problem of access to and control of each patient's healthcare records; UK example: high-level summary records are owned by NHS and shared automatically to all pharmacies; My contention is that the core record should be available to patient, AT LEAST through their covid app and build a bridge between their individual records and make better data available to public health records
        • In statute, the NHS record is mine
        • Children's "blue folders" (maternity records) and "red books" (birth - 5 records) are digital and shared; when they were in paper, it was under mother's control, but in being digitized, shared data
        • MyData "DIDs for kids" didn't go very far, too scary to medical industry interlocutors
    • Bernard: I want to understand role of ML in here
      • This example was from self-asserted diet study— OpenMined was involved to help diet data be studied
    • Bob: 4% of US medications are misdirected!
  • Recording 21 October ) 21 October 2020 - Joost Flach & Mark van der Waal (Triall.io, NL) & Maarten Boender (Sphereon, NL) on decentralized clinical trials
  • Recording 7 October - Georg Jürgens (Spherity GmbH) and Robert Celeste (Center for Supply Chain Studies, USA) on VC use-cases in B2B pharmaceutical ecosystems, including Spherity's Authorized Trading Partner trial with SAP and Novartis.
Minutes
  • Attendance/ Introductions
    • Najib Rehman (??)
      • working with a data integrator to support swiss pharma clients on CTs; DLT to speed things up a bit
    • Willy (??)
    • Leah Houston
    • Bernard
    • Catherine
  • Robert's presentation - scale: billions of transactions in a 3-hour overnight crunch period - trial generating valuable material and guidance for industry governance processes - interaction of technical, operations, and compliance/governance groups - API and Schemata —> starting point for standardization
    • "This is an industry that runs on standards"
      • GS1 for txn info and ids,
      • Everyone obvi wants to look with the W3C standards
        • all the examples in the spec are very consumer-based; bias and mental model https://s3-us-west-2.amazonaws.com/secure.notion-static.com/c3cd24cf-2a3c-4e0c-9ffd-ccc02ef75187/Untitled.png
  • "Challenges"
    • Challenge #1: Scale
    • Standards and misfit b/w W3C as-is today and these use-cases
    • Challenge #2: Retention req (6 year record-keeping obligation) —> binding VCs and VPs need to be archival/auditable for 6 years!
      • VP ephemerality (in VC spec)
      • Lesson#1 Identifier ←→ Company name binding needs to be durable/archival
    • Challenge #3: How does Company B know that A was handling credentials properly and according to industry and/or regulatory rules, if they just get a log or a recipt?
      • Logging of credential exchanges? reusing credentials falsely, replay attacks, etc?
      • Out of scope for now, but we're starting to look at the possibility of a third-party compliance auditor or guarantor? Do they attach additional VCs?
  • security risks
    • drug cartel made 25 shell corps to hide infiltration in a drug supply chain
      • mitigations of legal identity issues
      • drug cartels have a big R&D budgets for impersonation, counterfeiting, etc— way more tech savvy than many supply chain actors such as smalltown pharmacies!
  • Q and A
    • Najid: FMD (I used to work for Pfizer so I know the DCSCA and serialization!); what are your thoughts on B2B2P and P identity?
    • Orphan disease area? Teaching hospital ... Therapies and vectors?
    • Volume and complex stakeholders...
    • Standards around labeling (tagging blood in loop); 200K-1.3m$ therapies! the liabilities (and stakes) are massive there
    • How do we get everybody on board? Some people are incentivized for intransparency
    • Bob: One of the pilots in the FDA program was a very expensive personalized drug: the roundtrip; temp control in transit; temperature indicator behind a bar code, so that bar code changed
    • Georg: Novartis is looking into this for Kamriah; validated (manual) process for now