A Plea for Distributed Collaboration

This "Plea" was orginally published by Charlie McCay and Rene Spronk on 2007-01-02. It is a paper related to the ongoing HL7 reorganization which suggests a way forward that is compatible with the recommendations of the strategic task force, and would deliver clarity moving forwards.

Introduction

The HL7 reorganisation has been in gestation for a couple of years, with the work of PIC and ORC feeding into the Strategic Review Task Force, and subsequently into the Transition Oversight Team.

The following paper suggests a way forwards that is compatible with the recommendations of the strategic task force, addresses the issues as identified in the original problem statement, and would deliver clarity moving forwards.

In short, this paper recommends that

• HL7 moves forward as a facilitator for the creation of standards by a self-organizing body of volunteers instead of a move towards a top-down directed model, and that
• the role of HL7.org be limited to facilitating the process of harmonizing the realm-level models. All realm-level issues (development of realm specifications, marketing, MOUs, etc.) should be taken care of by the realms.

Current Status

HL7 has a long history of grassroots collaboration, with a self-organizing structure. HL7 Inc. and the affiliates serve as organizational structures that have the sole task to support the ongoing collaboration process.

Increasingly, localization efforts and the creation of implementation guides takes place at the Realm level. Implementers care most about realm level issues.

Roughly 35% of members are based in Northern America, 45% in Europe, 15% Asia/Pacific, 5% elsewhere. This is based on the latest available membership numbers as provided by the affiliates, using a 3:1 ratio for the number of individuals associated with organizational members versus an individual membership. These numbers are provided in this paper as a reminder of the geographic distribution of the membership. [See reports by the Marketing Committee for details of the geographic distribution statistics]. The current makeup of HL7 is rather different from these figures; suggesting an imbalance.

Options

It is apparent from all the work done investigating the state of HL7.org and the affiliates that the time has come to make a decision on the strategic way forward. There are a number of options available:-

1. Do Nothing
2. Consortium Model
3. Distributed Collaboration
4. Fully Directed

Each has a number of Pros & Cons which are discussed below. This document proposes that option 3 is the most viable with the greatest benefits to all parties.

Distributed Collaboration

HL7 Provides the Framework

HL7 should focus on providing the best environment where standards can be developed. The business of maintaining this open standards development framework (i.e. the infrastructure necessary for the creation of standards) should be kept separate from the business of using the framework to develop and maintain standards.

If HL7 wishes to run a Specification Development Service in parallel with maintaining the framework, this should be done at arms length if the trust and engagement of others using the framework is to be maintained. This has been done successfully by the project offices in the Netherlands and Germany.

The alternative is to move to a fully directed model, where all specifications development is centrally managed. This is not recommended as discussed in the “Alternatives” section below.

1. Project Coordination
   • The HL7.org would provide Project Coordination, via a Project Office This would be in an advisory capacity only.
2. Workload Management
   • Currently, the decision as to what should be balloted is made by the members of the technical committees. It is proposed that this remain the formal process, with the HL7 project office providing advice on when there is overlapping work, but leaving it to the committees to best determine how to resolve those conflicts. The project office would also maintain visibility for expected progress of specifications through the approvals process, but would not directly determine committee agendas.
   • This would build on the existing strengths of the HL7 organisation, and its relationships with content developers (“volunteers”) and user organisations.
3. Full lifecycle visibility
   • The HL7 project office and the affiliates should track the uptake of its standards, and the demand for new standards, in a more methodical way.
   • This sort of market research would be extremely valuable to anyone engaging with HL7 at the affiliate or global level, and would help to guide those inside and outside the organisation to the HL7 products that are relevant to them. This should be done at realm and global level in a way that supports aggregation of the information.
4. Quality of Ballots
   • HL7’s commonplace practice of using ballots both for community validation and quality control is inherently flawed. We need to some level of quality review in advance of ballot. A negative feedback loop should be present in addition to the ballot process. Options include:
   1. The project coordinator (section 4.1.1) could (in a pure advisory capacity) inform a group that their material is of poor quality.
   2. Akin to the Slashdot mechanism: a group of 3 volunteers is asked, for 1 single ballot cycle only (to lower the burden on those volunteers, and to avoid misuse of being in the reviewer role) to review the material and to inform a group (in a pure advisory capacity) that their material is of poor quality. The next ballot cycle, 3 other volunteers form the quality control group. Slashdot states "Being a moderator [limited to 3 days on Slashdot] is like jury duty - You never know when you're gonna have to do it, and when you get it, you only do it for a little bit." The committee could heed the advise and withdraw (or improve) the ballot. If the committee still decides to go to ballot with it (because they feel the quality control process was flawed), the quality advisory could be included in the ballot announcement - which alerts the balloters that something could be amiss with the material ensuring that it will receive appropriate negative feedback if indeed the quality is below par.
5. Ballot Fatigue
   • Regardless of the option one chooses (in terms of the strategic way to move forward): ballot fatique may occur in any of them. The changes proposed in this paper have a number of side effects that are likely to lower the risk of ballot fatique.
   • "For a work item to go forwards at the global level it would require support from a number of countries.." (section 4.3.1) - This will lower the number of international ballots, which has the (side effect) of lowering the risk of ballot fatique. It will increase the number of realm ballots (or whatever process a realm chooses to have or approving realm specifications and proposals) - ballots within a realm tend to have different reviewers and balloters than international specs, spreading the load on individual volunteers.
   • Increasing the quality of the individual ballots also helps to decrease the burden on the balloters: it lowers the overall number of ballot documents. This can be achieved either by a quality control mechanism as described in the previous section, or by balloting a specifications/proposals within a realm prior to brining it forward to the global organization.

Substantial HL7US affiliate established

It is proposed that the governance and running of HL7US and HL7.org should be totally split and the consequential substantial change to the organisation embraced. This would allow HL7US to focus clearly and unambiguously on the development of product for the US market, with the global realm being reserved for those specifications that need to be harmonised at that level.

One of the challenges for the organization right now is that the pressure to produce "international" specifications frequently leads to production of frameworks rather than implementable standards, and this puts the U.S. (who has no separate forum to make those frameworks implementable) in the position of seeing the v3 work-products as being unimplementable. The introduction of a strong U.S. affiliate would alleviate this issue.

In the medium term the HL7.org secretariat should not be co-located with the HL7US secretariat. The economies of scale that result from having the HL7.org and HL7US secretariat in one organisation do not justify the resulting lack of clarity and focus.

Impact

• HL7 Work Product Planning
  • Only those items for which there is a business case for work at the global level will obtain the sufficient input to develop specifications.
  • For a work item to go forwards at the global level it would require support from a number of countries, and should normally be targeted as an ISO standard as well. Ballot comments suggesting that the specification should be dealt with at the realm level would be expected for products that do not have global relevance.
  • By establishing some criteria for work done at the international level (i.e. a workitem has to concern an inherently non-realm extensible artefact -e.g. RIM- or has to have the support of –for instance- at least 3 realms) we'll push most realm issues down to the realm level. If a realm would like to discuss problems they encounter they can always do so on an informal basis, or by creating a universal variant proposal of their realm artefacts.
  • The global organization to be a forum for the development of frameworks to ensure there's a consistent underlying structure to realm-created specifications to allow for subsequent harmonization and/or mapping. This covers realm-independent content such as the RIM, datatypes, structural vocabulary, ITSs, conformance rules, methodology, etc.

• Limited Resources for HL7.org
  • HL7.org would be a lean organisation that would take a percentage (e.g. 20%) of affiliate income. On the basis of 2005 revenue figures, 20% of membership revenues equals USD 480K.
  • HL7.org would create an opportunity for global sponsorships (e.g. at USD 50K), with privileges associated it, which do not include special voting privileges. ROI for the sponsoring organization would be visibility of being a global sponsor.
  • This low level of revenue would be sufficient to run the WGMs, and maintain a minimal secretariat, with most of the membership relations being managed at the realm level. The development and maintenance of the HL7.org website, tooling infrastructure, collaboration infrastructure (webex, etc) would be an opportunity for sponsorship and/or central funding.
  • The role of HL7.org would be limited to facilitating the process of the creation of inter-realm/global models, and creating the necessary global infrastructure (in terms of tools etc.) to facilitate the creation of those standards by the community. The phrase "inter-realm/global models" should NOT be interpreted as just encompassing high level models such as the RIM. Affiliate work should to be done "within the framework" of international domain-specific specifications as much as possible. I.e. You want to have a global framework for how lab should operate and then constrain that to fit the unique operating environments of each country, rather than independently creating multiple lab solutions in each country and then attempting to harmonize.
  • Most (initial stages of) activities will take place at the realm level. The organization of global or realm meetings is left to the hosting realm organization, assisted by the HL7.org secretariat.

• CEO remit and value
  • The major value identified for the CEO is in the US to deliver focus and visibility to activities there. The global spokesperson for HL7.org could be a part-time role played by the same person who plays the role of CEO of one of its affiliates (e.g. HL7 US).

• Marketing, Press Releases and Memoranda of Understandings
  • PressReleases (PR) and many Memoranda of Understanding (MoU) are a realm issues and should be dealt with at a realm level.
  • Strategic marketing (product lines, branding) should be done by HL7.org, and localized by realms in PR activities. Similarly, MOUs with a global (or: inter-realm) scope should be signed by HL7.org.

These things simply have to be adopted to realm specific circumstances. To try and manage this at a global level would require knowledge of the particulars of each and every country. A global org can create “template items” for re-use by realms.

• Governance
  • Global specifications as well as the board of the global organization would be balloted by all members of HL7.org on a one member one vote basis. If HL7 is to truly be an international organization, there can be no variation in the value of a vote from someone in one country to someone in another country. Persons who are members of multiple affiliates (and/or of the current HL7 Inc. organization) have one single vote.
  • A one member one vote policy should be introduced for all issues related to the approval of standards as well as the election of persons in governance roles within the global organization. If HL7 is to truly be an international organization, there can be no variation in the value of a vote from someone in one country to someone in another country. Persons (through organizational membership or not) that are members of one or multiple affiliates (and/or of the current HL7 Inc. organization) shall have one single vote.
    • A. This policy shall be introduced before the election of the chair-elect that will serve after Ed Hammond.
    • B. All necessary preceding steps, including, but not limited to, the creation of a global registry of HL7 members, the resolution of privacy issues related to the creation of such a global registry, and an administrative ballot to approve of this new policy, shall be taken care of before the time listed under A.
  • Work items would get onto agendas in the same way that they do now – with co-chairs managing the committee’s agendas and the WGM agenda being set by those that attend the conference calls and mailing list discussions.
  • The Board of HL7.org (and its Advisory Board) shall be re-organized to reflect a better balance of constituencies, proportionate to the worldwide distribution of the membership. This could be achieved by a "weighted" vote by affiliate to ensure that smaller affiliates have a reasonable voice in the governance of the organization, while still allocating increased influence to those countries with larger memberships.

• Active Content producers set the technical agenda
  • In the current situation, it is the active members who set the technical agenda. This is a concern for “light users” and those who may feel that the HL7 brand is being confused by too many specifications being produced and approved without consistency checking against other specifications. This will be addressed to some extent by the recommendation that HL7 employ a technical project coordinator to identify and report areas of overlapping work.
  • It is, however, reasonable that those producing content should set the agenda for the technical committees, since any organisation with the resource to develop meaningful content, is likely to have some underlying requirement and the prospect of implementing it.
• Global working meetings to be held globally
  • The global meeting should be moved around the world, much as the interoperability conference does. Rotation between North America, Europe and Asia/Pacific should be proportionate with the membership allocation. For instance, perhaps membership in Asia-Pacific isn’t large enough to justify annual meetings, but one every other year would be warranted.
  • While this may result in the loss of some active participants, it is likely that those with global interests will be able to attend rotating meetings. For those that cannot justify the travel, there will be more vibrant affiliate meetings to meet national needs, and the opportunity to travel to a more local HL7.org meeting once a year.

Alternatives (to be avoided)

Do Nothing

For all the reasons given in the Strategic review documents, do nothing is not a viable option. The working group meetings are not effective, and there is a good deal of inconsistency between HL7 products, and greatly reduced engagement in the balloting process.

Consortium Model

The fees for membership would be dramatically increased, and a much small number of organisations would be members and determine the agendas and the specifications that should be worked on. Others would be free to comment, but only the members would be able to vote.

This would be a radical change to the brand, ethos and business model of the organisation. This is likely to result in a significant change to the way the individuals and organisations engage with HL7, and there seems little prospect of success. This is not something that ORC or Strategic Taskforce recommended.

Fully Directed

This option would see more of the activity done within HL7 being managed centrally. This would be achieved by a combination of carrots (professionalizing the organization in areas such as admin, project coordination, technical and publishing support) and ‘Gateway’ processes that controls the right to include material in the ballot.

The process for defining and running the Gateways (by HL7 staff) would be resource intensive and fallible, and so unlikely to add more value than simply improving the visibility of work items, and encouraging realm developments and implementation prior to global harmonisation.

On the positive side, this would result in those with approved projects (projects that can get through the gateways) being comfortable. However, those which were not approved may be likely to go elsewhere to develop their specifications. This would undermine the value of the HL7 brand as a one-stop-shop for Medical Informatics Standards, and result in either a flowering of inconsistent profiles and standards from different organisations, resulting in reduced interoperability and/or expensive harmonisation efforts.

Unresolved issues

ANSI Accreditation

It is assumed that HL7.org would be ANSI accredited, and that both HL7US and HL7.org balloted material would follow the ANSI rules and be ANSI certified.

Q: Could HL7.org be ISO accredited? If not, it would be sensible for global HL7 standards to be developed with input from ISO national bodies being welcomed, as part of going through an accelerated ISO adoption process.

A: (by Melvyn Reynolds): IEEE (also an ANSI accredited SDO) has an agreement (known as the "Pilot process") with ISO whereby the products of IEEE P11073 (Medical Device Communication) are effectively fast tracked into ISO. Ed Hammond has been instrumental in getting HL7 into a similar (but not identical) fast-tracking-to-ISO position - but recent ballot experience there indicates that early consensus formation (and realm sensitivity) is still a necessity.

Literature

The plea for self-organization is based in part on analysis and review of peer production activities such as Open Source Software projects.

• (online report) FLOSS Report, [1].
• (online papers) Free/Libre Open Source Software Research, [2].
• (book) Wikinomics - How Mass Collaboration Changes Everything, Don Tapscott and Anthony A. Williams, 2006, [3].
• (blog) The different aspects of Peer governance, Michel Bauwens, [4].
• (online book) Commons Based Peer Production, Yochai Benkler, 2006, [5].
• (book) The Starfish and the Spider: The Unstoppable Power of Leaderless Organizations, Ori Brafman and Rod Beckstrom, 2006, [6].