[Openid-specs-heart] Considering profiling resource sets/scopes for the "before first visit" sharing scenario

Mon Jun 27 16:05:51 UTC 2016

Hmm.

The SMART on FHIR work invented OAuth scopes, which we subsequently put
into our HEART Profile for FHIR Scopes
<http://openid.bitbucket.org/HEART/openid-heart-fhir-oauth2.html> draft
spec. These are not just useful scopes for "explanations, examples, and
test-bench" purposes, they are meant as real, normative scopes that give
real interoperable shape to the security and consent properties of
OAuth-protected FHIR APIs. We may or may not yet have picked the right
scopes, but pick them we did. (And our charter calls for us to define spec
profiles.)

The very same calculus is true for UMA, only more so. Each data element
already has a confidentiality level, orthogonal to "what it is" (and here
we're talking about the common clinical data set and virtual clipboard bits
of information, as noted/linked in the use case). I'm assuming that
"preferred language" and "smoking status" and "medication allergies" and
whatnot might possibly have different security classifications, while
obviously having different semantics and structure.

The reason why it's "more so" for UMA is that resource owners don't just
react to OAuth requests for authorization all in one go, but they have
finer-grained ways of interacting, and so there are more UX implications.
For example, you could go back later and change your sharing settings
differentially for different people.

If we don't choose to give any guidance at all about resource set and scope
boundaries, we could do that, but I'm pretty confident that it would be
difficult to satisfy a minimal set of "sharing scenario 1" expectations on
the part of the doctor's office, and Alice's experience would be
extra-confusing too.

It's fine if HL7 or other bodies don't give us existing semantic buckets to
go on. We could go finer-grained too. We should consider each of PGHD
<https://www.healthit.gov/policy-researchers-implementers/patient-generated-health-data>,
the Common Clinical Data Set
<https://www.healthit.gov/sites/default/files/commonclinicaldataset_ml_11-4-15.pdf>,
and Virtual Clipboard
<http://healthitanalytics.com/news/virtual-clipboard-to-tackle-patient-matching-interoperability/>
(the last of which really seems to be US-centric, with its
financial/insurance data focus) in context.

As an aside, Adrian, I'm sorry, but I'm trying to add real value here with
every message. I didn't just change the subject line. Our very job is to
profile specs. "Profiling privacy" has no meaning, but "profiling UMA"
actually does. Since I've documented many UMA use cases and designed
several UMA POCs, I hope you'll allow me the courtesy here of doing
something that's more than "intent" -- it's a structured exercise.

*Eve Maler*ForgeRock Office of the CTO | VP Innovation & Emerging Technology
Cell +1 425.345.6756 | Skype: xmlgrrl | Twitter: @xmlgrrl
*ForgeRock Summits and UnSummits* are coming to
<http://summits.forgerock.com/> *Sydney, London, and Paris!*

On Mon, Jun 27, 2016 at 6:58 AM, John Moehrke <johnmoehrke at gmail.com> wrote:

> Hi Adrian,
>
> I would agree HEART shouldn't be 'profiling', as in defining formal
> constraints, on FHIR resources available nor scopes available...  I agree
> with you that these things are discoverable in FHIR using the FHIR
> conformance mechanisms, there is no need to constrain by HEART rules.
>
> Were we really profiling these? I thought we were looking for an
> instructive set to use in explanations, examples, and test-bench.
>
> so, what we might have here is a different understanding of the word
> profiling. Formal profiling is to place constraints that can't be violated.
> I think this is what Adrian is pointing out and rightfully getting worried
> about.  Where as some people will use the word profiling in a more informal
> way, more as a sample use-case. We should use the word profiling only in
> the formal sense. This formal meaning does cause future trouble when life
> changes, but then you just define a new profile... However that doesn't
> mean we should formally constrain something that is not necessary to
> formally constrain, when we really only want to put out a useful subset.
>
> John
>
> John Moehrke
> Principal Engineering Architect: Standards - Interoperability, Privacy,
> and Security
> CyberPrivacy – Enabling authorized communications while respecting Privacy
> M +1 920-564-2067
> JohnMoehrke at gmail.com
> https://www.linkedin.com/in/johnmoehrke
> https://healthcaresecprivacy.blogspot.com
> "Quis custodiet ipsos custodes?" ("Who watches the watchers?")
>
> On Mon, Jun 27, 2016 at 8:47 AM, Adrian Gropper <agropper at healthurl.com>
> wrote:
>
>> Hi John and Debbie,
>>
>> Let's stipulate the universality of the confidentiality classifications.
>> My question remains: What are we profiling and why?
>>
>> To put this in specific FHIR and OAuth terms:
>>
>> - Would a FHIR resource server offer different patient-level resource
>> types to an authorization server controlled by a patient vs. one controlled
>> by a licensed practitioner?
>>
>> - Is there an objection to offering the full set of FHIR patient-level
>> resource types for an AS to apply policy to whether or not a particular RS
>> chooses to also bundle resource-types based on confidentiality
>> classifications or any other business logic they choose?
>>
>> Adrian
>>
>> On Mon, Jun 27, 2016 at 9:45 AM, Debbie Bucci <debbucci at gmail.com> wrote:
>>
>>> I agree with something useful.
>>>
>>> Aren't we profiling common resource sets and scopes for a FHIR RS;
>>> Something that could be use generically by all but perhaps  extended if
>>> need be?
>>>
>>> I  think FHIR RS shares resource sets/scopes it supports to UMA AS.
>>> Period.  No hints further (c on Adrians list) User sets preferences for
>>> sharing on those resource sets.  If a release seems like a logical RS set -
>>> then let's add to the list.
>>>
>>>
>>> On Jun 27, 2016 8:49 AM, "John Moehrke" <johnmoehrke at gmail.com> wrote:
>>>
>>>> Hi Adrian,
>>>>
>>>> Small but important clarification. The confidentiality classifications
>>>> I gave are universal in healthcare. They are used in all forms of HL7,
>>>> including FHIR and CDA. They are used in DICOM. They are used in IHE
>>>> XDS/XCA and Direct. There is even a mapping to 13606 equivalent risk scale.
>>>> So by starting with this group you are not picking something special for
>>>> FHIR, but you are picking something that works with FHIR and other
>>>> exchanges. In all of these cases this value-set is built right into the
>>>> core of each of these standards.  We could use a smaller subset, like older
>>>> CDA did.  Note this set also supports when data has been de-identified and
>>>> thus moved to a lower risk, where lower risk is an assessment and an
>>>> appropriate movement down the scale. So it even works where
>>>> differential-privacy might move the data down to "M" or "L".
>>>>
>>>> Are we tying to determine the FINAL list of scopes? Or just a starter
>>>> set that seems to be useful? I was hoping we were just looking for
>>>> something useful. I understand that our need is to enable interoperability,
>>>> as the scopes do need to be agreed to by all parties. Thus this starter set
>>>> just gets us back to focusing on the HEART specifics. Scopes are infinitely
>>>> expandable so the starter set can be enhanced or even replaced (deprecated)
>>>> later.
>>>>
>>>>
>>>> John
>>>>
>>>> John Moehrke
>>>> Principal Engineering Architect: Standards - Interoperability, Privacy,
>>>> and Security
>>>> CyberPrivacy – Enabling authorized communications while respecting
>>>> Privacy
>>>> M +1 920-564-2067
>>>> JohnMoehrke at gmail.com
>>>> https://www.linkedin.com/in/johnmoehrke
>>>> https://healthcaresecprivacy.blogspot.com
>>>> "Quis custodiet ipsos custodes?" ("Who watches the watchers?")
>>>>
>>>> On Sun, Jun 26, 2016 at 11:03 PM, Adrian Gropper <
>>>> agropper at healthurl.com> wrote:
>>>>
>>>>> Thank you John for making things clear. Changing the subject line of
>>>>> my "Is HEART Profiling Privacy?" thread does not solve the problem but I do
>>>>> appreciate and respect Eve's intent.
>>>>>
>>>>> I'm not sure the OpenID Connect analogy is useful for HEART or for UMA
>>>>> in general.
>>>>>
>>>>> I claim that building on FHIR Confidentiality Classification is beyond
>>>>> HEART's charter. Patient Privacy Rights would love to participate in that
>>>>> discussion with whomever in HL7 wants to join in but it's not easy to see
>>>>> who else would be at that table and I don't see anyone clamoring to charter
>>>>> that group. For reference, I've attache a nice study of what might be
>>>>> involved in such an exercise.
>>>>>
>>>>> I hope UMA does not force HEART to profile privacy. As John makes
>>>>> clear, FHIR resource types are not being designed to make privacy
>>>>> classifications or patient understanding of privacy issues particularly
>>>>> easy or hard. I suspect the same thing is true outside of healthcare.
>>>>>
>>>>> My suggestion does not require that HEART profiles privacy or make use
>>>>> of confidentiality classification. A HEART resource server (a) can publish
>>>>> the list of resource types it supports, and
>>>>> (b) can post to a specific HEART AS which resource types are or are
>>>>> not available for a particular resource registration, and maybe
>>>>> (c) can optionally post explanations, hints, resource type groupings,
>>>>> or examples to the specific HEART AS. This would enable a RS to send the
>>>>> equivalent of their current Release of Information form to the AS.
>>>>>
>>>>> Note that (c) is optional. Whether (c) is there or not, FHIR supports
>>>>> whatever granularity it supports and there's no reason an AS should not
>>>>> choose to take advantage of that.
>>>>>
>>>>> Therefore (a) and (b) together allow any RS to register with any AS
>>>>> and it's up to the AS to map the resource types into policy choices for the
>>>>> resource owner. Rather than HEART profiling privacy, i suggest we allow ASs
>>>>> to compete on how to present the privacy issues to the resource owner.
>>>>>
>>>>> Adrian
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On Sun, Jun 26, 2016 at 9:21 PM, John Moehrke <johnmoehrke at gmail.com>
>>>>> wrote:
>>>>>
>>>>>> Hi Eve,
>>>>>>
>>>>>> It would seem that we would indeed have this figured out. But reality
>>>>>> is that we never quite talk about it so bluntly, mostly because classifying
>>>>>> healthcare data is so very hard. Further the various ways that would be
>>>>>> logical don't fall nicely upon the logical ways that data can be segmented.
>>>>>> That is the way that people would like to segment their data is according
>>>>>> to how sensitive it is; yet this doesn't align with FHIR resources, or
>>>>>> departments, or function, etc...  Plus data changes sensitivity based on
>>>>>> current medical knowledge (HIV was very sensitive a few years ago, today it
>>>>>> tends to trend closer to normal data).
>>>>>>
>>>>>> So the best I have to offer is the _confidentiality classifications
>>>>>> we have in the security-tags: It is a scale that centers around Normal -
>>>>>> that group of data that everyone agrees is the typical healthcare data
>>>>>> (mathematically normal). Think of it as a risk scale of 6 different risk
>>>>>> quantum. All data MUST have one of these values assigned to it.
>>>>>>    http://hl7-fhir.github.io/v3/ConfidentialityClassification/vs.html
>>>>>>
>>>>>> so U, L, M, N, R, V --- definitions are on the page above.
>>>>>>
>>>>>> Is this a good solution? NO! but it does exist today.
>>>>>>
>>>>>> More on my blog article
>>>>>>
>>>>>> https://healthcaresecprivacy.blogspot.com/2015/07/how-to-set-confidentialitycode.html
>>>>>> or
>>>>>>
>>>>>> https://healthcaresecprivacy.blogspot.com/2016/01/fhir-oauth-scope.html
>>>>>>
>>>>>> John
>>>>>>
>>>>>> John Moehrke
>>>>>> Principal Engineering Architect: Standards - Interoperability,
>>>>>> Privacy, and Security
>>>>>> CyberPrivacy – Enabling authorized communications while respecting
>>>>>> Privacy
>>>>>> M +1 920-564-2067
>>>>>> JohnMoehrke at gmail.com
>>>>>> https://www.linkedin.com/in/johnmoehrke
>>>>>> https://healthcaresecprivacy.blogspot.com
>>>>>> "Quis custodiet ipsos custodes?" ("Who watches the watchers?")
>>>>>>
>>>>>> On Sun, Jun 26, 2016 at 11:18 AM, Eve Maler <eve.maler at forgerock.com>
>>>>>> wrote:
>>>>>>
>>>>>>> This message is intended to give shape to our profiling efforts. I
>>>>>>> apologize in advance if I "step in it" due to my lack of HL7 or FHIR
>>>>>>> knowledge. :-)
>>>>>>>
>>>>>>> Given our first sharing scenario
>>>>>>> <https://bitbucket.org/openid/heart/wiki/Alice_Shares_with_Physicians_and_Others_UMA_FHIR> (in
>>>>>>> the latest use case) of providing basic data before a first visit, are
>>>>>>> there various subsets or supersets of such data, e.g., in different
>>>>>>> jurisdictions, that Alice would be expected to provide, or one obvious and
>>>>>>> universally understood data set? I'm guessing that payment data is one
>>>>>>> subset of data that doesn't apply in some jurisdictions, whereas quite a
>>>>>>> lot of "medically related data" would generally be considered desirable to
>>>>>>> share if available.
>>>>>>>
>>>>>>> What I'm going for here is: If you could have a system of "keywords"
>>>>>>> representing virtual clipboard-type data sets and how you want them shared,
>>>>>>> what's the smallest arrangement of keywords that would do the trick?
>>>>>>>
>>>>>>> By (very rough) analogy, look at how OpenID Connect standardizes
>>>>>>> client requests for claims using scope values
>>>>>>> <http://openid.net/specs/openid-connect-core-1_0.html#ScopeClaims>.
>>>>>>> It rolls up multiple identity claims under single scope names, so that
>>>>>>> "openid" gives the client full access at the UserInfo endpoint, "profile"
>>>>>>> gets access to 14 specific claims, "email" gets access to two claims, etc.
>>>>>>>
>>>>>>> UMA gives us two dimensions to play with, so it could be like this:
>>>>>>>
>>>>>>> Potential resource sets to access (I'm totally making this up! for
>>>>>>> all I know, HL7 figured this out 10 years ago and it doesn't look anything
>>>>>>> like this...):
>>>>>>>
>>>>>>>    - BasicVirtualClipboardData
>>>>>>>    - FinancialVirtualClipboardData (or a unique version per
>>>>>>>    jurisdiction?)
>>>>>>>    - AllVirtualClipboardData (an aggregation of the two of them)
>>>>>>>
>>>>>>> Potential scopes of action across them (same for each one? it
>>>>>>> doesn't have to be that way, and other resource sets might have
>>>>>>> deidentified-read options or something...):
>>>>>>>
>>>>>>>    - read
>>>>>>>    - write
>>>>>>>
>>>>>>> Note that the scope work already begun in our OAuth FHIR profile,
>>>>>>> coming from the SMART on FHIR work, gives us ideas for things like a "read"
>>>>>>> verb and a "write" verb, but for right now, we can think outside the box if
>>>>>>> that's the right thing to do.
>>>>>>>
>>>>>>>
>>>>>>> *Eve Maler*ForgeRock Office of the CTO | VP Innovation & Emerging
>>>>>>> Technology
>>>>>>> Cell +1 425.345.6756 | Skype: xmlgrrl | Twitter: @xmlgrrl
>>>>>>> *ForgeRock Summits and UnSummits* are coming to
>>>>>>> <http://summits.forgerock.com/> *Sydney, London, and Paris!*
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Openid-specs-heart mailing list
>>>>>>> Openid-specs-heart at lists.openid.net
>>>>>>> http://lists.openid.net/mailman/listinfo/openid-specs-heart
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> Openid-specs-heart mailing list
>>>>>> Openid-specs-heart at lists.openid.net
>>>>>> http://lists.openid.net/mailman/listinfo/openid-specs-heart
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>>
>>>>> Adrian Gropper MD
>>>>>
>>>>> PROTECT YOUR FUTURE - RESTORE Health Privacy!
>>>>> HELP us fight for the right to control personal health data.
>>>>> DONATE: http://patientprivacyrights.org/donate-2/
>>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Openid-specs-heart mailing list
>>>> Openid-specs-heart at lists.openid.net
>>>> http://lists.openid.net/mailman/listinfo/openid-specs-heart
>>>>
>>>>
>>> _______________________________________________
>>> Openid-specs-heart mailing list
>>> Openid-specs-heart at lists.openid.net
>>> http://lists.openid.net/mailman/listinfo/openid-specs-heart
>>>
>>>
>>
>>
>> --
>>
>> Adrian Gropper MD
>>
>> PROTECT YOUR FUTURE - RESTORE Health Privacy!
>> HELP us fight for the right to control personal health data.
>> DONATE: http://patientprivacyrights.org/donate-2/
>>
>
>
> _______________________________________________
> Openid-specs-heart mailing list
> Openid-specs-heart at lists.openid.net
> http://lists.openid.net/mailman/listinfo/openid-specs-heart
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openid.net/pipermail/openid-specs-heart/attachments/20160627/1e333a2e/attachment.html>