The starting point for the panel was that the ability to share medical knowledge is a major pre-requisite for the successful implementation and large-scale deployment of enactable clinical guideline applications. The challenges involved in formalizing, sharing and maintaining machine-interpretable clinical knowledge are considerable, but, equally, so are the potential benefits.

Much progress has been made in formalisms for encoding knowledge and building useful clinical applications over the past two decades. The panel took the view that enough is now known about these formalisms to permit the primary focus of research and development to be moved towards addressing the practical challenges involved in developing repositories of content which are sufficiently convincing to attract the support of publicly-minded contributors and also the attention of organizations which influence take-up of innovations in clinical practice. The latter include medical IT and publishing companies as well as healthcare service providers.

As we know from the Open Source software experience, the notion of publicly shared intellectual property is, to a considerable degree, opposed to the objectives of commercial organizations whose assumption is that competitive advantage depends on exclusive ownership of intellectual property. The vision of sharable medical knowledge is therefore only likely to be achieved if two main conditions can be met:

1. An open knowledge community can be established which can operate sustainably as an independent sector (e.g., by servicing a market which does not depend exclusively on commercial players);
2. The development and acceptance of business models in which publicly-shared knowledge can be licensed for commercial use (such as the selling of applications which add value to open content).

The panel discussed many of the technical tasks which need to be accomplished to enable knowledge to be shared in a form which computers can interpret and used in practical clinical tasks such as decision support and care planning. The development of standardized platforms for deploying scalable knowledge-based services of these kinds would include a set of requirements to ensure, for example, that:

• Services are as far as possible mutually compatible and interoperable and free of institution-specific details;
• Content and service components are reusable;
• Quality and safety can be assured by automated cross-verification.

But the panel felt that the vision of sharing executable clinical knowledge will not be achieved without strong co-operation between groups of people, communities of practice, able and willing to share, maintain, update, and improve content, in a manner analogous to the publicly-minded communities of programmers who develop and maintain open source software.

Presentations were made by the four panel participants to cover some of the main challenges involved in sharing and maintaining repositories of executable clinical knowledge, and covered the following topics:

1. Practical methods to weave medical knowledge services into an application and to map clinical abstractions into electronic health records - Mor Peleg.
2. Lifecycles for creating, publishing and maintaining open source, sharable knowledge-based services for supporting patient care - John Fox.
3. Methodologies for distilling sharable knowledge from encoded knowledge that is tied to particular institutions and systems - Bob Greenes.
4. Incentives and business models for creating a knowledge-sharing community - Sheizaf Rafaeli.

A general discussion followed the panel presentations. A number of the prominent points raised in the discussion are summarized below, along with an assessment of the general feeling from the hall towards each, where this could be readily identified:

1. Can a national or international effort be put together to create a repository of implementable knowledge? There was general agreement in the hall that this should be possible.
2. Can the goal of guideline sharing be achieved within 10 years? There was general agreement that this should be possible.
3. Guideline-sharing at the implementation level requires separation into component steps that can be individually implemented, because differences in process/workflow can prevent a guideline from being adopted in its entirety. The hall was mainly in disagreement with this statement.
4. True sharing of executable medical knowledge can never be fully achieved because knowledge cannot be separated from local institutional requirements and constraints. There was general agreement with this statement.
5. Guideline formalization activities do not typically address implementation settings and requirements. The hall was split on this proposistion.
6. A major challenge now is to establish principles of safe deployment and use in clinical service design. There was general agreement with this.
7. Open technical standards (covering medical concepts, clinical vocabulary, task models, for example) will be key. The hall was split around 50:50 in its response.
8. A vital challenge will be to persuade the commercial world of medical IT publishing etc. to develop business models that accept and build on open standards. There was general agreement with this.
9. Incentives, whether material, social or ego-oriented, are important to generating results. The hall was split around 50:50 in its response.
10. Social and ego-oriented incentives are more available, more likely to generate results, and have has more leverage for potential participating scientists as compared to material incentives.