[Openid-specs-heart] Resource/Scope Proposal #2
Salyards, Kenneth (SAMHSA/OPPI)
Kenneth.Salyards at SAMHSA.hhs.gov
Wed Aug 17 18:06:01 UTC 2016
First, I would like to thank Linda and Vivek for their proposals.
My first comment is that when specifications are based on artifacts (FHIR Resources) that will change over time it makes it very difficult to constantly update resource sets to accommodate new resources.
My second comment is to my knowledge no production systems using FHIR exist. FHIR is going into STU3 ballot in September. So the availability of FHIR data in the form of resources does not exist and may not for the foreseeable future. I know, I know; FHIR is going to be everywhere and everyone will be using FHIR! Of course, it’s inevitable! (yes, I’m being facetious ☺).
My third comment is it appears HEART is going down the path of requiring patients to define resource sets associated with FHIR resources. I thought HEART is supposed to be user/patient centered. Having to take a patient through a process to define resource sets on sharing FHIR resources seems to be the epitome of not user friendly. Having been in the data sharing patient facing space for some 15 years, I can say that most patients do not want to know or try to understand the technology underlying their need to share or not share their medical information. Also, in my opinion asking a patient to decide to share or not share information based on a limited set of data does the patient and the whole community a disservice.
My last comment for this email (whew!); OAuth and OpenID Connect are about authentication and authorization for an identity. UMA adds the user perspective with a trivial example of printing a photo (http://kantarainitiative.org/confluence/display/uma/Introduction+to+UMA) – Dare I say it: Health information is a complex domain and in the case of FHIR resources the same health information is represented differently across multiple FHIR resources (e.g., medications, problems, labs, procedures, care plan) which contain information a patient may not want to share with a particular provider. A patient defines a resource set of read for all of the resources except problems thinking that she is not sharing information that identifies a particular condition to the intended recipient. However, any of the permitted resources contain information related to the treatment of this problem.
In closing, to the patient we can say: here is a nifty little app that we will walk you through on how to have control over sharing your health information; it really doesn’t do anything for you, however we will pretend it does so you think you are in control of how you share information with your providers because the implementers didn’t want to do anything complicated because they felt the app uptake would be better if it didn’t do anything. ☹ Ken
From: Openid-specs-heart [mailto:openid-specs-heart-bounces at lists.openid.net] On Behalf Of John Moehrke
Sent: Wednesday, August 17, 2016 9:56 AM
To: Nancy Lush; HEART List
Subject: Re: [Openid-specs-heart] Resource/Scope Proposal #2
didn't realize this was only to Nancy.-- so reflecting to the list.
Hi Nancy,
I suggest this - multi-generation plan - approach simply because I know that the vectors through the healthcare-privacy space are close to infinite. I think we should pick the battles we win, one at a time. Getting a UMA system in place for healthcare would be a huge improvement. Even if it had next to no flexibility in scopes. The concept of "User Managed Access" is so powerful to enable a patient to allow others (human or computer) to access their data and thus improve their life. We are deadlocked on details that can be improved later. By writing down the items we are deferring, we create a backlog to pull from in future iterations. I recognize however that privacy advocates (I count myself one) have heard this multi-generation approach before and they have scars to show that only the first generation ever gets accomplished. I sympathize, and say to them that this is even more proof that the first generation should be under-powered. I will repeat, getting UMA in place is very powerful, even with just Read/Write as scope. We have alot of trust and infrastructure issues with just this. I don't want to stop there, hopefully this is clear by my email an blog posts. However I am frustrated at the stalemate we have wrapped around something that is not the most important improvement overall. (Note this i also why SMART scopes are so simplistic, as use of OAuth is more important than the scopes we will invent in the coming decade)
John
John Moehrke
Principal Engineering Architect: Standards - Interoperability, Privacy, and Security
CyberPrivacy – Enabling authorized communications while respecting Privacy
M +1 920-564-2067<tel:%2B1%20920-564-2067>
JohnMoehrke at gmail.com<mailto:JohnMoehrke at gmail.com>
https://www.linkedin.com/in/johnmoehrke
https://healthcaresecprivacy.blogspot.com
"Quis custodiet ipsos custodes?" ("Who watches the watchers?")
On Tue, Aug 16, 2016 at 6:15 PM, Nancy Lush <nlush at lgisoftware.com<mailto:nlush at lgisoftware.com>> wrote:
Interesting. I had not thought about that third case.
So when you say ‘I don't think we should try to put this kind of thing in a scope; yet’ – Should this team define cases that we would like to solve, but which will not be solved in a version 1?
-Nancy
From: John Moehrke [mailto:johnmoehrke at gmail.com<mailto:johnmoehrke at gmail.com>]
Sent: Tuesday, August 16, 2016 7:03 PM
To: Nancy Lush <nlush at lgisoftware.com<mailto:nlush at lgisoftware.com>>
Cc: HEART List <openid-specs-heart at lists.openid.net<mailto:openid-specs-heart at lists.openid.net>>
Subject: Re: [Openid-specs-heart] Resource/Scope Proposal #2
I want to clear up what seems to have gotten confused.
There are a few places where date-ranges are needed.
1. How long is this authorization statement valid --- This is a not-before, not-after. This is typically short-term specific to an authorization decision. OAuth includes this in the token
2. How long is this policy statement valid -- when I give rules to the UMA AS, how long do I want this to last. This is around the authorization rules. For example, I am authorizing doctor Bob to have access to my data for 24 months. This is commonly seen as an expiration of a consent. This is what the period element in the FHIR consent is for. UMA might have internal values for this, but no need to expose this. Clearly (1) would need to be shorter than (2) when the deadlines approach.
3. The period for which a rule applies. For example I authorize access to my data, except NOT the data created during 1998. Or I authorize access to Joe any data that was created during 1998.
It is this third one that gets involved in a scope discussion. I don't think we should try to put this kind of thing in a scope; yet. It is however likely the top priority vector for rules. Often times a patient knows a timeframe that they wish the world to forget.
John
John Moehrke
Principal Engineering Architect: Standards - Interoperability, Privacy, and Security
CyberPrivacy – Enabling authorized communications while respecting Privacy
M +1 920-564-2067<tel:%2B1%20920-564-2067>
JohnMoehrke at gmail.com<mailto:JohnMoehrke at gmail.com>
https://www.linkedin.com/in/johnmoehrke
https://healthcaresecprivacy.blogspot.com
"Quis custodiet ipsos custodes?" ("Who watches the watchers?")
On Tue, Aug 16, 2016 at 4:15 PM, Nancy Lush <nlush at lgisoftware.com<mailto:nlush at lgisoftware.com>> wrote:
HEART proposal Example 2
(I apologize in advance for such a long email. Also, sorry for the delay as I was away on vacation and catching up.)
Justin proposed that team members propose suggested resources, scopes and what they mean. The objective is to suggest positive solutions. None will be totally ‘correct’ but that is OK because together these suggestions can be combined and adjusted to reach the ‘correct’ solution. Based on that, I will create another stab as an alternative.
1. First, since HEART is based on FHIR, we should not expect any changes in FHIR, at least for the 1.0 version. The resources should be the FHIR resources. I am stating this as a discrete point. I think we all agree, but I would like to know that we do indeed concur.
2. In an attempt to simplify, my original list of resources was the list of FHIR resources that are implemented by the Argonaut group. Since this is the set that most of the current servers have implemented, we could expect them to be supported first. (I have no objection to the list being the complete list of FHIR resources, if that is preferred by the group.) This subset list also happens to correspond to the Common Clinical Data set. (I am suggesting that we start with this limited set in the assumption that we could implement HEART as soon as it is specified.)
• Patient demographics (Known as ‘patient’ per FHIR)
• Allergies
• Problems & Health concerns (Conditions)
• Vital Signs (Category of Observations)
• Labs
• Smoking Status
• Care Team (Some vendors have Care Plan of which Care Team is a subset.)
• Medications
• Immunizations
• Goals
---- this next subset of resources could expand on the above list and are a continuation of the Argonaut implementation list.
• UDI (Device)
• Procedures
• Plan of Treatment
• Assessment
3. The chart below is simply to demonstrate to those less familiar with FHIR how this might work. All of these results are per patient.
Common Name
FHIR Resource
Category if applies
Other filter
Patient demographics
Patient
Allergies
AllergyIntolerance
Problems and Health Concerns
Condition
Vital Signs
Observation
Vital-signs
Lab Results
DiagnosticReport
Lab
Smoking Status
Observation
Code=72166-2
Care Team
CarePlan
CareTeam
Medications
MedicationStatement
Medications
MedicationOrder
Immunizations
Immunization
Goals
Goal
UDI
Device
Procedures
Procedure
Plan of Treatment
(Still being defined in re-sprint)
Assessment
(Still being defined in re-sprint)
There are different filters for each resource type. These should not be addressed by HEART in terms of having our patient, Alice, provide varying permissions at that level.
The reasons for limiting the resources to a list like this is because they are already implemented and tested for many servers and because they are a great first step. If we can do this in HEART for starters we will have achieved a lot!
4. Terminology: This group often uses different terminology which causes confusion. Eve has suggested several times that we focus on clarifying terminology and has begun that effort in her use case document. One term that has bothered me in the past is the term ‘Resource Set’. At the HL7 conference, Josh clarified this for me. My current understanding is that UMA uses ‘Resource Set’ for the same purpose as FHIR uses ‘Resource Type’, or as is often shortened in FHIR as ‘Resource’. So in FHIR, if we request medications for the patient Alice, we refer to the resource ‘Medications’, while UMA refers to this as a ‘Resource Set’ for the patient Alice. (The result would be a list of medications for Alice.) If this understanding is currently incorrect, can someone please correct it?
5. Grouping of resources: Let’s assume that we have agreed on the list of FHIR resources. I propose that we allow Alice to select which of the list she desires to share, with the addition of choices ‘all’ or ‘none’. The requests for the resources by the client would still be made individually for each resource. (To me the notion of ‘resource set’ implied that we would define a set of individual resources – say by some category. I would advise against using such a notion as it only complicates and does not add any benefits.)
6. Scopes
a. Read/Write
i. In last week’s meeting, Justin referenced the current scope stream: individual, bulk, read, write. Since we are specifying HEART, I would assume this is always what the patient is specifying. So for HEART, the scope of ‘bulk’ is not relevant. Further we have scopes of Read, Write, or *. I suggest that we allow Alice to specify Read, Write, and have the ability to specify both. These settings can be set either per resource or for all resources.
Two points should be noted:
ii. Current implementations are supporting ‘Read’ and do not yet support ‘Write’. We may be fine with starting with just ‘Read’ for version 1.0.
iii. Since Alice is defining what she is willing to share – wouldn’t that be the ‘Read’ case? I can imagine that we will see implementations where Alice controls her RS and in that case could specify ‘Write’ permissions. For the immediate future, I would be surprised if the EMR would allow Alice to give permissions to Dr. Bob to write to an EMR. There may be a case where Alice can give Dr. Bob permission to write to her PHR.
b. Dates
i. My understanding from yesterday’s conversation was that we do not want to include dates in the scopes on resources, per HEART. There are date filters available for some resources, but that is within the FHIR query specification, an outside of the HEART spec.
ii. Eve raised the objective of having expiration dates associated with the permissions granted, but that functionality does not apply to resource date scopes
iii. So ‘dates’ are not considered as a scope per this discussion.
c. Confidentiality codes: I was the culprit that initially raised the case of Alice not wanting to share her HIV condition, but only sharing her non-HIV condition. I believe most of us would like to provide such a feature in HEART, but since that original discussion I am now of the opinion that we should NOT include this feature in version 1.0.
i. Confidentiality codes were suggested as a potential solution to this issue. It has been stated that some additional ‘magic’ needs to occur to make this viable. Since it is not currently supported in current FHIR implementation, adding this to HEART in version 1.0 could derail the effort.
ii. Confidentiality codes are defined in FHIR as tags.
iii. Confidentiality codes are not implemented in most of the current servers. Even if they were, there is not a consistent method to consistently code values across servers. This could lead to inconsistent results and be worse that not providing the feature at all.
iv. If Alice desired to not share her ‘HIV’ condition, we would need to add a scope to the resource request asking the server to withhold data that matches the requested confidentiality settings. This scope is not currently defined in FHIR and we should not be attempting to change FHIR as part of our HEART 1.0 specification.
v. If I were running a RS, I certainly would not be willing to send Alice’s HIV condition to the client tagged as confidential and expect the client to not share it – instead I would want to not send the data at all – thus the requirement to add a new scope – which we should definitely avoid.
This document is also added as an attachment.
-Nancy
Nancy Lush
nancy.lush at lgisoftware.com<mailto:nancy.lush at lgisoftware.com>
Lush Group, Inc
Office: (401) 423-9111<tel:%28401%29%20423-9111>
28 Narragansett Ave
PO Box 651
www.lgisoftware.com<http://www.lgisoftware.com>
Cell:(401) 965-9347<tel:%28401%29%20965-9347>
Jamestown, RI 02835
[LGI_logo_small]
_______________________________________________
Openid-specs-heart mailing list
Openid-specs-heart at lists.openid.net<mailto: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/20160817/36a14288/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.gif
Type: image/gif
Size: 3006 bytes
Desc: image001.gif
URL: <http://lists.openid.net/pipermail/openid-specs-heart/attachments/20160817/36a14288/attachment-0001.gif>
More information about the Openid-specs-heart
mailing list