[Openid-specs-heart] HEART Scope Design Proposal #1

Wed Aug 10 22:00:37 UTC 2016

Thank you John for pointing us to a specific document so that we can figure
out how the UMA AS and UMA scopes might interact with the HL7 work. Looking
at the spec I can ask a few specific (sic) questions which I have numbered
below:

(1) To Ken's question, the NYP ROI Authorization Form is at
https://dl.dropboxusercontent.com/u/8909568/NYP%20authorization.pdf Is it
fair to say that the FHIR Consent resource
http://hl7-fhir.github.io/consent.html is designed to capture the content
of the executed NYP ROI Form in a legally accountable way, regardless of
how actual sharing will be enforced?

(2) I don't see any particular way to express date ranges in the Consent
spec. Maybe this is to be handled through extensions. If the purpose of
standardizing the Consent resource is interoperability, then it seems to me
that a lot of profiling will need to be done and it's not clear who will do
that profiling. I believe HL7 FHIR has an 80% rule. Does the Consent spec
require profiling to reach the 80% level? (Being able to encode the NYP ROI
Form would meet the 80% threshold for me.) If profiling is needed to reach
80%, who will do that profiling?

(3) The Privacy Consent Directive
http://hl7-fhir.github.io/consent.html#6.4.1.1 introduces the term
Substitute Decision Maker which is not mentioned anywhere else in the
document. Could the UMA AS be the Substitute Decision Maker this section is
referencing? If so, then point us to a spec for Substitute Decision Maker
and we will make progress on the relationship between FHIR and HEART.

I agree with John that (c) Both is the preferred solution. (b) has nothing
to do with UMA and HEART. (a) is unrealistic, impractical, and undesirable
because it assumes a massive bureaucracy that protects all the institutions
and all of the patients in all circumstances. This kind of bureaucracy has
been elusive (see the UK NHS care.data news for an example of trying to do
this in a single national system) and review the history of NHIN for the US
equivalent. Option (c) is already supported by the UMA spec because it
allows the UMA resource server to impose whatever jurisdictional and
business rules it chooses regardless of what the UMA AS authorization token
and associated scopes may say.

John is clear on the user experience aspects of the discussion. Whether an
UMA AS chooses to implement FHIR Consent in order to improve the user
experience may not be relevant to the HEART scope discussion. Authorization
Servers can compete on their user experience without harming
interoperability for FHIR endpoints labeled HEART. The interoperability
rule (be liberal in what you accept and strict in what you send out)
applies.

(4) The sine-qua-non for interoperability is that any Client labeled HEART
and any Server labeled HEART must work together under any UMA standard AS.
Some UMA AS will be FHIR-aware and others will not. Does anyone interpret
the HEART charter as implying that the AS is somehow healthcare specific?

As for John's suggestion that we get UMA in the FHIRe door first before we
deal with improvements, let's revisit the date or page range issue as part
of the relationship between UMA and FHIR as orthogonal standards that we
are all in favor of after we make some progress on questions (1) to (4).

Adrian

On Wed, Aug 10, 2016 at 4:35 PM, John Moehrke <johnmoehrke at gmail.com> wrote:

> Adrian,
>
> A Consent Directive is one of the formal terms used for results of a
> Privacy Consent act/ceremony. Lots of terms for the same thing. There is
> nothing new being spoken about here. Note these are defined on the FHIR
> Consent Resource page
>   http://hl7-fhir.github.io/consent.html
>
> In FHIR ballot about to start (STU3) the concept is managed by the FHIR
> Consent Resource, rather than a profile on contract. This was done to
> simplify and focus. ANYONE who is an HL7 member, make sure you signup for
> the ballot NOW. Outsiders can also participate.
>   www.hl7.org/ctl.cfm?action=ballots.home&ballot_cycle_id=541
>
> The FHIR Consent Resource is only a mechanism for capturing the consent
> meaning. It is not a mechanism for decision making or enforcement. It is
> also not a mechanism for interacting with the patient (UI).  The FHIR
> Consent resource is just a resource. It is not a service. It is not an
> operational environment. It is not an application. Something else would be
> the method of intervention upon requests for data about a subject, deciding
> if it is authorized, and enforcing it.
>
> Where UMA is a method of making decisions and enforcement. UMA leaves the
> capturing of the rules that it enforces as an implementation detail of the
> AS. Nore does UMA define a static format for the rules.
>
> Thus one could:
> a) Use UMA alone, leaving up to UI of the UMA AS to capture the rules,
> persist the consent experience, etc... This means that UMA AS is used to
> protect all accesses to protected data about the subject.
> b) Use FHIR Consent to capture the results of the consent ceremony; and
> all data holding servers would look to the FHIR Consent rules when they
> themselves decide if a request for data is authorized.
> c) Use FHIR Consent to capture the results of the consent ceremony; and
> use UMA to make data request authorization decisions.
>
> So, you could use one, or the other, or both. There is no need for us to
> constrain one or the other. But we should be aware of all. Clearly the best
> combination from a standards perspective is (c); but it is also the most
> complex UMA AS.
>
> Neither of these is dependent on Scopes. Scopes are an optimization
> factor. The better aligned Scopes are to the consent rules, the fewer UMA
> authorization decisions need to be made. If Scopes align well, then one
> decision authorizing a broader scope value can be re-used in many different
> transactions by an application against a resource sever. If the scopes
> don't align well, then authorization decisions need to be on a request by
> request basis. This is just an optimization. It does not inhibit the rules
> that the UMA can enforce. I believe I am right, so if I am wrong I hope
> someone from OAuth/UMA community can illuminate my mistake.
>
> This is why I recommend we just pick some scopes for now. Similar to what
> SMART did, they just picked something that was easy to pick.  The real
> magic that we are enabling is to have UMA involved, and thus getting "User"
> "Managed" "Access" (Aka Patient Centric Authorizations). Once we get UMA in
> this position, the UMA AS can be improved.
>
> 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 Wed, Aug 10, 2016 at 1:57 PM, Adrian Gropper <agropper at healthurl.com>
> wrote:
>
>> What is a consent directive?
>>
>> What is the relationship of whatever it is to FHIR or any other standard
>> data model?
>>
>> How does a consent directive interact with OAuth, UMA or OpenID Connect?
>>
>> Do you have an example of how a consent directive maps into, replaces, or
>> changes the NYP ROI authorization form?
>>
>> Thanks,
>>
>> Adrian
>>
>>
>>
>> On Wed, Aug 10, 2016 at 2:49 PM, Salyards, Kenneth (SAMHSA/OPPI) <
>> Kenneth.Salyards at samhsa.hhs.gov> wrote:
>>
>>> Fine grained authorization should be dealt with using a patients’
>>> consent directive. FHIR DSTU2 supports consent directive as part of the
>>> contract resource. In DSTU3, as I understand the consent directive will not
>>> be part of contract. This can work seamlessly within the UMA resource
>>> server construct. Ken
>>>
>>>
>>>
>>> *From:* Openid-specs-heart [mailto:openid-specs-heart-bou
>>> nces at lists.openid.net] *On Behalf Of *Vivek Biswas
>>> *Sent:* Wednesday, August 10, 2016 1:50 PM
>>> *To:* Adrian Gropper; openid-specs-heart at lists.openid.net
>>> *Cc:* Josh Mandel; Grahame Grieve
>>>
>>> *Subject:* Re: [Openid-specs-heart] HEART Scope Design Proposal #1
>>>
>>>
>>>
>>> Hi Adrian,
>>>
>>>
>>>
>>> The scopes are meant to do coarse-grained authorization and not fine
>>> grained authorization.
>>>
>>>
>>>
>>> And hence, this one of the reason why scopes are static string.
>>>
>>>
>>>
>>> If one want to perform fine grained authorization, then its based on lot
>>> of contextual information like you mention, date-range, geo-location,
>>> etc....
>>>
>>>
>>>
>>> The contextual information can reside in the Request Payload, in
>>> database, within the access token (if JWT) etc. All these contextual
>>> information associated with the request can be trusted only if the access
>>> token is valid.
>>>
>>>
>>>
>>> There can be a policy associated with context which can help us to
>>> decide if the request should be authorized or not.
>>>
>>>
>>>
>>> So, scopes can help you do first level of coarse grain authorization
>>> which will yield you an access token. Once an access token is valid, grab
>>> the contextual information from the payload, header, access token etc. Find
>>> a policy associated with the contextual object and than perform fine level
>>> authorization based on the policy.
>>>
>>>
>>>
>>>
>>>
>>> Regards
>>>
>>> Vivek Biswas, CISSP
>>>
>>> Consulting Member @Oracle
>>>
>>>
>>>
>>>
>>>
>>>
>>> ------------------------------
>>>
>>> *From:* Adrian Gropper <agropper at healthurl.com>
>>> *To:* "openid-specs-heart at lists.openid.net" <
>>> openid-specs-heart at lists.openid.net>
>>> *Cc:* Justin Richer <jricher at mit.edu>; Vivek Biswas <
>>> vivek_biswas at yahoo.com>; Josh Mandel <jmandel at gmail.com>; Grahame
>>> Grieve <grahame at healthintersections.com.au>
>>> *Sent:* Tuesday, August 9, 2016 12:57 PM
>>> *Subject:* Re: [Openid-specs-heart] HEART Scope Design Proposal #1
>>>
>>>
>>>
>>> There are many reasons why we need to find a way:
>>>
>>>    1. I have never seen a release of information form that did not have
>>>    a date range. Has anyone? If we say HEART can't do that we're calling into
>>>    question the legitimacy of HEART.
>>>    2. We have a lot of experience with 75-page CCDAs with the current
>>>    standards. Many patients have huge health records and it is
>>>    resource-intensive to get 74-pages of information you already had in order
>>>    to find the change from the last encounter. The cost is not just on the
>>>    sender but on the recipient as well. This is probably the top complaint of
>>>    the current interop methods in my medical society.
>>>    3. I don't think the HEART charter set a particular limit on how
>>>    much of FHIR we would manage. It's not in our charter to treat effective
>>>    provider-to-provider exchange as out-of-scope. Once we say that HEART will
>>>    not support the full patient-level feature set of FHIR, where do we draw
>>>    the line?
>>>    4. UMA is not OAuth. The goals of the two standards are very
>>>    different. UMA and HEART are patient-centered. If we restrict HEART to a
>>>    small fraction of the longitudinal health records or patient-centered
>>>    health rerecords transactions (because most of the traffic stays in the
>>>    batch or institution-to-institution domain) then where do we host the
>>>    discussion on a future for patient-centered records?
>>>    5. There is no logical reason for HEART to not support date ranges
>>>    once FHIR has that capability. In some jurisdictions, this would be seen as
>>>    reducing patient's right of access and subject to legal challenge.
>>>
>>> I've added Josh and Grahame to this thread. Let's try and find the
>>> solution to this problem even if it involves changing UMA, FHIR or both.
>>>
>>> Adrian
>>>
>>>
>>>
>>> On Tue, Aug 9, 2016 at 12:31 PM, Vivek Biswas <vivek_biswas at yahoo.com>
>>> wrote:
>>>
>>> I agree with Justin, that scopes are static string vs scope string be
>>> parameterized. Most of the OAuth server support static string.
>>>
>>>
>>>
>>> It will get challenging to implement a profile which requires dynamic
>>> string which in turn will hamper the adoption of HEART.
>>>
>>>
>>>
>>> We are trying to add contextual information into the scope which are not
>>> what scopes are intended for.
>>>
>>>
>>> Regards
>>>
>>> Vivek Biswas, CISSP
>>>
>>> Consulting member @Oracle
>>>
>>>
>>> On Aug 9, 2016, at 9:17 AM, Justin Richer <jricher at mit.edu> wrote:
>>>
>>> A problem I can see with this is that the “date” field in particular
>>> moves this from something that’s expressible as a table of known strings
>>> (like in the appendix of the current spec) to something that’s dynamically
>>> parameterized with a potentially infinite set of values. This is a problem
>>> in practice as many OAuth implementations treat all scopes as static
>>> strings, and will do things like limit set set of scopes a client is
>>> allowed to ask for based on a set of registered strings. That’s not
>>> possible with a field like “date” anymore, since the value is filled in at
>>> runtime. We tried to have parameterized scopes in BB+ (and even implemented
>>> it) but it was generally thought to be too complicated, and required
>>> special tooling at the authorizations server.
>>>
>>>
>>>
>>> If at all possible, I’d like HEART scopes to not require such special
>>> processing and support at the AS.
>>>
>>>
>>>
>>>  — Justin
>>>
>>>
>>>
>>> On Aug 8, 2016, at 9:08 PM, Adrian Gropper <agropper at healthurl.com>
>>> wrote:
>>>
>>>
>>>
>>> Scope Design Proposal #1 aims to support a typical ROI authorization
>>> <https://dl.dropboxusercontent.com/u/8909568/NYP%20authorization.pdf>.
>>>
>>> The structure of SDP1 is: patient/date
>>> <http://hl7.org/fhir/search.html#date>/confidentialitycl ass
>>> <http://hl7-fhir.github.io/v3/ConfidentialityClassification/vs.html>/
>>> resource <http://hl7.org/fhir/resourcelist.html>/edit
>>>
>>>
>>> The logic is as follows:
>>>
>>>    - /patient because this applies to only one patient at a time. The
>>>    patient ID is local to the resource server.
>>>    - /date <http://hl7.org/fhir/search.html#date> is defined by FHIR
>>>    and can be a range. Putting it at the highest level in the hierarchy (if a
>>>    scope hierarchy is useful) allows for efficiency in clients requesting
>>>    updates and reduces the cost to the resource server
>>>    - /confidentialityclass
>>>    <http://hl7-fhir.github.io/v3/ConfidentialityClassification/vs.html>
>>>    filters for resources at or below the specified value. Resources that do
>>>    not have a confidentiality class are considered N - Normal. It is up to the
>>>    resource server to provide jurisdictictionally appropriate policies and
>>>    user interfaces for setting confidentiality class other than N on
>>>    particular resources.
>>>    - /resource <http://hl7.org/fhir/resourcelist.html> is any resource
>>>    listed in the particular version of the FHIR specification
>>>    - /edit is "read", "write", "any" operation by the client on the
>>>    resource
>>>
>>> A client might request a scope for immunizations for patient 23 as:
>>>
>>> [ "date=le2016-8-8","conclass=N" , "resource=Immunization", "edit=read" ]
>>>
>>> on a resource registered as: [base]/Patient/23/ with a reference to HEART scopes SDP1
>>> that would tell both the AS and anyone else that dereferenced HEART/SDP1 that the RS would
>>> process specific scope strings for date, confidentialityclass, resource, and edit as described above.
>>>
>>> The AS would be free to present SDP1 to the RO any way that it chose.
>>>
>>> A resource server like NYP that wanted to offer registration for sensitive data like Mental Health Treatment
>>> would register another resource: [base]/Patient/23/ MentalHealthTx with whatever scopes it offered.
>>> These additional resources would not be specified by HEART.
>>>
>>> Any particular RS could choose to support Confidentiality Class, additional resources, neither, or both.
>>>
>>> It would be up to the AS to combine HEART SDP1 resources and additional resources and scopes offered by the RS into whatever policy setting experience it wanted.
>>>
>>> With this scheme, an AS might offer Alice a policy setting for
>>> Observation = R - Restricted even on a resource server that did not support
>>> confidentiality class. This would cause all client requests for
>>> Observations to fail because the RS would be forced to treat unlabeled
>>> resources as N and the authorization tokens for observations were R. The RS
>>> could:
>>> (a) ignore this problem and treat it as an AS bug,
>>> (b) implement confidentiality classification, or
>>> (c) offer additional resources and scopes so that the AS could tell
>>> Alice that Observations did not include those related to mental health in
>>> the hope that Alice would set Observation = N and deal with mental health
>>> as a separate part of her policy.
>>>
>>> I believe that, to a first approximation, SDP1 would capture the
>>> functionality of most sharing use-cases without forcing the RS to do
>>> anything more than it is jurisdictionally already required to do.
>>>
>>> Adrian
>>>
>>>
>>>
>>>
>>> --
>>>
>>>
>>>
>>> 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/
>>> <http://patientprivacyrights.org/donate-2/>
>>>
>>> ______________________________ _________________
>>> Openid-specs-heart mailing list
>>> Openid-specs-heart at lists. openid.net
>>> <Openid-specs-heart at lists.openid.net>
>>> http://lists.openid.net/ mailman/listinfo/openid-specs- heart
>>> <http://lists.openid.net/mailman/listinfo/openid-specs-heart>
>>>
>>>
>>>
>>> ______________________________ _________________
>>> Openid-specs-heart mailing list
>>> Openid-specs-heart at lists. openid.net
>>> <Openid-specs-heart at lists.openid.net>
>>> http://lists.openid.net/ mailman/listinfo/openid-specs- heart
>>> <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/
>>>
>>>
>>>
>>
>>
>>
>> --
>>
>> 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
>>
>>
>

-- 

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/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openid.net/pipermail/openid-specs-heart/attachments/20160810/7cbabfc5/attachment-0001.html>