Themes to address from SME sessions

  • Best practices and recommendations
    • standardization of data and data structure and requirements (Karly Rowe)
      • USCDI R2 in progress
      • USPS address standardization (publication 28)
      • Carmen Smiley  to write up our position on this, which is it is complex and we do not have specific recommendations, but here is a list of ongoing/upcoming efforts to keep an eye on
    • IAL
      • do we want to make any recommendation on minimum IAL for patients that organizations should put into their agreements when implementing these solutions together? Julie Maas Catherine Schulten 
    • Logs and Audits Meena Jambulingam
      • Responder to log information on non-match scenarios for further analysis 
      • Requester to log information on no results/non-matches although they won't have the same level of information on why there wasn't a match as the responder
    • Use of Only Certain Matches flag in FHIR patient match operation Meena Jambulingam
      • use cases when this should be set to true or false
      • definition of "certain match" - what is the threshold?
      • when working with a matching service or solution, we encourage organizations to explore the definition of "confidence level" of the service and any options to configure it based on the specific datasets being handled
    • Public curated synthetic data sets for matching Meena Jambulingam
      • We acknowledge this could be a beneficial asset for the industry in the absence of a way to benchmark against real data sets with privacy preserving practices
      • Organizations such as MITRE and RAND may be best positioned to make available and maintain such a curated data set of synthetic clinical data
      • Data sets like these could be used for benchmarking various matching solutions
    • Biometrics Meena Jambulingam
      • we see a pathway from simple uses today to more interoperable scenarios. begin with human readable facial image on screen compared with patient standing in front of the practitioner all the way up to interoperable biometrics that can be stored and shared across systems.

      • Biometrics doesn’t solve data standardization/mapping problem, but creates more interoperable identity – it’s the consumer-facing aspect of all this

      • Biometrics fit in more with B2C than B2B (today).

      • B2B use cases may be able to use a broader/richer data set of biometric data such as HLA type, blood type, antibodies tested for, whether I’m immune to mumps or measles, etc.
  • Comparison of Solutions
    • Adjusted Technical and Process complexities based on feedback from Blake Hall
    • Meena Jambulingam to write up/tweak intro to solutions to indicate use cases when one or more solutions may be applied either individually or together
    • Meena Jambulingam extend FHIR patient match operation to add IAL in request
    • Call out that solutions 3 and 4 don't necessarily preclude participation in/of multiple trust networks
    • Collin Wallis - establish a NIST 63-3 "healthcare profile" - this is probably a good addition in scope to whatever happens next with FAST
  • Follow up for second opinion with rest of team 
    • One or None approach to matching responses. See notes from SMEs
      • From Eric Heflin to All Panelists: 04:06 PM - 1) human-mediated identification. In which case a lower standard of traits may be of value because the human can disambiguate. 2) if the system is automated then we may need to have either "one or none" matching as desired.
      • From Abel Kho to All Panelists: 04:07 PM - There is not a single answer here because there are different intents for why you want to match
      • From Paul Oates to All Panelists: 04:07 PM - yes, seeing some energy for a "floor" but not a consensus on whether we can set that. something we need to pursue in further discussion as a topic I think. until then, looks contextual
      • From Abel Kho to All Panelists: 04:08 PM - If in B2C someone wants to access their data or a service, they will be willing to provide high quality data or an authentication token
      • From Eric Heflin to All Panelists: 04:08 PM - @Matt, if a human is disambiguating the patient list, then we don't need a yes/no answer for the first list returned of candidate matches.
      • Solution 2: If we’re going to scale demographic matching nationally, you need to find centers of service to get it done – can’t do point to point
        • Agree, wasn’t thinking point to point or even regional – need to think about nationally
  • Solution updates
    • Solution 4

      • Meena Jambulingam solution/DLT would have to be ‘both’ public but permissioned? 

  • Open Questions
    • Will the chief architects provide some overall commentary on regulatory or other measures required to incentivize the industry to work together on interoperability? And maybe some view on how TEFCA plays into FAST solutions?
  • No labels