The importance of computer-based clinical decision support (CDS) is based on the many problems facing health care, and the need for proactive, point-of-care, patient-specific approaches to ensure that best practices are adopted [1Greenes RA, Ed. Clinical Decision Support: The Road Ahead. New York: Elsevier 2007.]. Among the well-recognized problems are: spiraling costs [2Poisal JA, Truffer C, Smith S, et al. Health spending projections through 2016: modest changes obscure part D’s impact Health Aff 2007; 26(2): W242-53., 3National Coalition on Health Care. Health care Facts: Costs [Retrieved: 11 November 2010]. Available from: http://nchc.org/facts-resources/ fact-sheet-cost 2009.], disparities of access and large numbers of uninsured [4California Health Care Foundation. Health Care Costs 101 [Retrieved: 11 November, 2010]. Available from: http://www.chcf.org/ 2005., 3National Coalition on Health Care. Health care Facts: Costs [Retrieved: 11 November 2010]. Available from: http://nchc.org/facts-resources/ fact-sheet-cost 2009.], errors and unevenness of quality [5Kohn L, Corrigan J, Donaldson M, Eds. Committee on Quality of Health Care in America, Institute of Medicine. To Err Is Human: Building a Safer Health System. Washington, D.C: National Academies Press 1999.-8Mangione-Smith R, DeCristofaro AH, Setodji CM, et al. The quality of ambulatory care delivered to children in the United States N Engl J Med 2007; 357(15): 1515-23.], slow dissemination of advances [9Balas EA, Boren SA. Managing clinical knowledge for health care improvement In: Van Bemmel J, McCray AT, Eds. Yearbook of Medical Informatics. Stuttgart, Germany: Schattauer Verlagsgesellschaft mbH 2000; pp. 65-70.,10Khoury MJ, Gwinn M, Yoon PW, Dowling N, Moore CA, Bradley L. The continuum of translation research in genomic medicine: how can we accelerate the appropriate integration of human genome discoveries into health care and disease prevention? Genet Med 2007; 9(10): 665-74.], inefficiencies and waste [11McKinsey Global Institute. Accounting for the cost of U.S. health care: A new look at why Americans spend more [Retrieved: 11 November 2010 from http://www.mckinsey.com/mgi/publications/ US_healthcare/ 2008.,3National Coalition on Health Care. Health care Facts: Costs [Retrieved: 11 November 2010]. Available from: http://nchc.org/facts-resources/ fact-sheet-cost 2009.], fragmentation and poor communication [12Halamka J, Overhage JM, Ricciardi L, Rishel W, Shirky C, Diamond C. Exchanging health information: local distribution, national coordination Health Aff (Millwood) 2005; 24(5): 1170-9.,13Marchibroda JM. The impact of health information technology on collaborative chronic care management J Manag Care Pharm 2008; 14(2 Suppl): S3-11.], and a lack of patient-centered care [6Institute of Medicine. Crossing the quality chasm: a new health system for the 21st century In: Institute of Medicine. Washington D.C: National Academies Press 2001.].

CDS has been shown to be useful in fostering patient safety, health care quality, and cost-effectiveness of care [1Greenes RA, Ed. Clinical Decision Support: The Road Ahead. New York: Elsevier 2007.]. Further, CDS fulfills the practical need of health care delivery organizations to respond effectively to programs such as pay for performance and prior authorization for medication or procedure orders. Recent emphasis on quality reporting and “meaningful use” [14American Health Information Management Association. Meaningful Use—Provider Requirements AHIMA Meaningful Use White Paper Series. Paper no. 2 2010; (Jan 20); 4 pgs, [Retrieved: November 11, 2010]. Available from: http://library.ahima.org/xpedio/groups/public/documents/ahima/bok1_046481.hcsp?dDocName=bok1_046481 ] under the American Recovery and Reinvestment Act of 2009 (ARRA) gives further impetus to CDS, since CDS can be used to proactively improve performance on the very tasks being measured as a basis for stimulus fund incentives or subsequent reductions in reimbursements.

Nonetheless, successful approaches to CDS have not been broadly disseminated and adopted for a variety of reasons [1Greenes RA, Ed. Clinical Decision Support: The Road Ahead. New York: Elsevier 2007.], including competitiveness, incompatibility of platforms, lack of standards, and under-utilization of electronic health records (EHRs) and computer-based provider order entry (CPOE) – major platforms in which to integrate CDS. But even for those with access to CDS, enormous, obstacles exist, including the lack of (1) sources of high quality medical knowledge in executable form, and (2) infrastructure and processes for managing and updating such knowledge and integrating it into applications. Addressing these needs is very expensive, if not prohibitive, for individual organizations, even large ones. In addition, the need for knowledge management capabilities has not yet been widely recognized except within a few of the larger medical centers that have been in the vanguard of CDS adoption. Further, standards for CDS representation and integration into applications are as yet incomplete.

In 2006 the U.S. Office of the National Coordinator for Health Information Technology (ONC) and the Agency for Healthcare Research and Quality (AHRQ) commissioned a project by the American Medical Informatics Association (AMIA) and produced The Roadmap for National Action on Clinical Decision Support, which proposed a coordinated approach to address the above impediments [15Osheroff J, Teich J, Middleton B, Steen E, Wright A, Detmer D. A Roadmap for National Action on Clinical Decision Support Report. Bethesda, MD: American Medical Informatics Association 2006.]. The report recommended a series of activities to improve CDS development, implementation and use throughout the United States to help enable improvements in health, and the quality, safety and efficiency of healthcare delivery.

There had been little evidence to date that health care organizations, public and private, were willing to share knowledge that they believe gives them a marketplace advantage. Nevertheless this had been done in other fields, even fiercely competitive ones like genomics research, e.g., in terms of contributions to GenBank [16Benson DA, Karsch-Mizrachi I, Lipman DJ, Ostell J, Sayers EW. GenBank Nucleic Acids Res 2010; 38(Database issue): D46-51.], the Genome Knowledgebase [17Joshi-Tope G, Vastrik I, Gopinath GR, et al. The Genome Knowledgebase: a resource for biologists and bioinformaticists Cold Spring Harb Symp Quant Biol 2003; 68: 237-43.], the DAVID Knowledgebase [18Sherman BT, Huang da W, Tan Q, et al. DAVID Knowledgebase: a gene-centered database integrating heterogeneous gene annotation resources to facilitate high-throughput gene functional analysis BMC Bioinformatics 2007; 8: 426.], Reactome [19Vastrik I, D'Eustachio P, Schmidt E, et al. Reactome: a knowledge base of biologic pathways and processes Genome Biol 2007; 8(3): R39.], and other molecular and genetic data and knowledge bases. To the extent that it has occurred at all, the primary sharing of clinical knowledge in executable form has generally been limited to the activities of knowledge content providers, EHR vendors, or user groups of a particular vendor. Other ventures have been less than successful. As an example, the Arden Syntax was developed in the early 1990s, and was adopted as an ASTM standard in 1991 [20Hripcsak G, Wigertz OB, Kahn MG, Clayton PD, Pryor TA. ASTM E31.15 on health knowledge representation: the Arden Syntax Stud Health Technol Inform 1993; 6: 105-2., 21Pryor TA, Hripcsak G. The Arden syntax for medical logic modules Int J Clin Monit Comput 1993; 10(4): 215-4.] and subsequently moved to Health Level Seven (HL7) in 1998; but a website for sharing of Arden Syntax Medical Logic Modules (MLMs) developed at Columbia University in the mid-1990s did not get regularly updated, and is no longer available. Arden Syntax has considerable use, but with many proprietary implementations, effectively limiting the extent to which sharing occurs. A consortium called the Institute for Medical Knowledge Implementation was formed in the early 2000s, and included health information system vendors, academic medical centers, and professional societies as members, with the goal of creating shared medical knowledge modules. The group foundered when no entity was willing to contribute content. (See [22PRNewswire reference to the Institute for Medical Knowledge Implementation [Retrieved: 11 November, 2010]. Available from: http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104&STOR Y=/www/story/02-11-2003/0001889281&EDATE= 2003.] for brief mention, although IMKI website is no longer available.)

Further, biomedical knowledge and technology development continue to advance relentlessly, making it impossible for practitioners to keep up with current knowledge, or for organizations to keep pace by integrating the new knowledge into their own practices. Promising sources of generally available knowledge such as AHRQ-supported National Guideline Clearinghouse [23Agency for Healthcare Research and Quality. National Guidelines Clearinghouse Available from: http://www.guidelines.gov 2010 [Accessed: 11 November, 2010];] and the Evidence-based Practice Center reports [24Agency for Healthcare Research and Quality. Evidence-based Practice Centers Available from: http://www.ahrq.gov/clinic/epc/ 2010 [Accessed: 11 November, 2010];], and the Cochrane Collaboration Library [25The Cochrane Collaboration. The Cochrane Library Available from: http://www.thecochranelibrary.com/view/0/index.html 2010 [Accessed: 11 November, 2010];] seek to establish and update authoritative collections of reports identifying best health care practices.

A current AHRQ program initiated as a part of the American Recovery and Reinvestment Act (ARRA) is the promulgation of Comparative Effectiveness Research (CER) [26Agency for Healthcare Research and Quality Fact Sheets on Recovery Act Investments in Comparative Effectiveness Research. Available from: http://www.ahrq.gov/fund/cerfactsheets/ 2010 [Accessed: 11 November, 2010];], which promises to further enhance such efforts. There are clearly many policy issues involved, such as how to assemble and organize disparate studies, whether a coordinating center should exist, what its role in guiding clinical decisions or influencing health care spending should be, and how it should be governed to ensure appropriate representation of stakeholders [27Wilensky GR. The policies and politics of creating a comparative clinical effectiveness research center Health Aff (Millwood) 2009; 28(4): 719-29.]. Assuming that a politically acceptable approach for these difficult issues is found, these knowledge resources would still not be made available in unambiguously executable (or even near-executable) form.

The technical issues and impediments involved and possible approaches to bridging the gap between such reports and recommendations and the development of ready-to-implement CDS are elaborated in the AMIA Roadmap call-to-action [15Osheroff J, Teich J, Middleton B, Steen E, Wright A, Detmer D. A Roadmap for National Action on Clinical Decision Support Report. Bethesda, MD: American Medical Informatics Association 2006.] and in the book by Greenes [1Greenes RA, Ed. Clinical Decision Support: The Road Ahead. New York: Elsevier 2007.]. One attractive possibility that was highlighted in the call to action is the idea of developing a Web-accessible repository of high quality medical knowledge, in an unambiguous form that can be used as a basis for implementing CDS, and that would be available to all institutions and users. This would avoid the need for each organization to duplicate the effort of creating and maintaining such a repository. Responsibility for managing this communal repository would rest in an authoritative body that would determine knowledge to be included, formalize its representation, index it for retrieval, and keep it updated. Projects initiated over the past two years have begun to tackle this idea.

The Morningside Initiative began when a small group of prominent health care organizations came together for a working meeting at the Morningside Inn, Frederick, MD, August 28-30, 2007, under the sponsorship of the Telemedicine and Advanced Technologies Research Center (TATRC) of the US Army Medical Research and Materiel Command. The purpose was to develop a long-term plan intended to address the challenges by means of a deliberate, gradual process, initially involving selected participants for a limited set of tasks, and then expanding in scope and scale as workable approaches are developed. Participants included representatives of the Department of Defense (DoD) Tri-Care Management/Health Affairs (TMA/HA), the Veterans’ Health Administration (VHA), Partners Healthcare System (Partners), Kaiser Permanente (Kaiser), Henry Ford Health System (HFHS), Arizona State University (ASU), the AMIA, and TATRC.

The Morningside Initiative is the result of that August 2007 initial working meeting, and was the unanimous recommendation of participants. It called for a collaboration to create a repository of shared knowledge for CDS to meet the needs of and be made available to diverse healthcare providers and organizations. Subsequent to the initial meeting, Intermountain Healthcare (Intermountain) joined the other participants to form the Morningside Initiative.

To achieve this vision, the collaboration is intended as a prototype of a larger future knowledge-sharing organization, aiming at developing and refining (1) an organizational framework, (2) a technical approach, and (3) content acquisition and management processes that are functional and can be scaled up to include a broader range of participants as well as an expansion of capabilities and content scope. The organizers of the Morningside Initiative concluded that the challenges of knowledge sharing (access to high quality content, reusability, tools for knowledge management, and an organizational framework to facilitate knowledge exchange) would best be addressed by bringing together its small number of highly visible and respected health care organizations in the public and private sector that could lead by example. They would develop and test a framework for knowledge sharing that could be extended to a national-level effort if successful. The rationale for this approach was based on three primary considerations:

1. Working out the organizational and logistical issues of knowledge sharing could be best done first with a small number of participants, so that the approaches can be refined through that experience, before undertaking a large-scale initiative involving many more participants.
2. Reluctance to share clinical knowledge by institutions could be overcome if high-profile participants have already committed to doing so and seeded the effort with a corpus of useful knowledge content.
3. Tools and methods for knowledge management, for creating, representing, and updating the content, and for facilitating adaptation and reuse in other sites should be developed and piloted with this small group.

The general framework for the Morningside Initiative is shown in Fig. (1). Current efforts are to capture best available knowledge already implemented as CDS in the systems of participant organizations; represent it in a more sharable, less setting-dependent format; drive toward standards-based models for representation; and develop and test approaches for secondary reuse. The key test of reuse is the adoption of this knowledge in settings different from those from which the CDS had originally been implemented.

Fig. (1) General framework for the Morningside Initiative.

In the future, it is anticipated that external knowledge will continually arise from authoritative studies, and that a process will be established to regularly incorporate this knowledge into the shared knowledge base. This is discussed further in the section on Future Directions.

The rationale for the effort has included the following potential benefits:

• In concert with the nationwide efforts to achieve widespread adoption of interoperable health IT, a collaborative approach based on standards and interoperability will help attain greater proliferation of CDS than any one entity could achieve alone.
• The approach will broaden the potential use and application of CDS at the point of care in order to impact health care decisions throughout the health care system.
• It will advance and foster more widespread accessibility of CDS content and interventions.
• By leveraging lessons, tactics and approaches used by others, buy-in across local, regional and national entities will be increased.
• A collective effort will address CDS topics of prime concern and importance to the national health care system in terms of cost, quality, and access to care.
• A national-level resource will facilitate adoption and more effective use of CDS within organizations that might otherwise not have the resources to pursue CDS.
• Availability of the resource will remove or lessen barriers and challenges (in terms of technical, cost, and expertise deficiencies) and thus enhance the likelihood of adoption and proliferation of CDS.
• Professional benefits will accrue to individual investigators for participation (e.g., papers, grant writing, research, and leadership opportunities).
• The collaborative mechanism will provide participating entities with a level of external validation of their CDS work regarding content quality and development approach.

A recognized requirement for a future collaborative knowledge sharing organization is a business model based on sustainable long-term mechanisms of support, yet to be defined, which might be expected to include a combination of organizational subsidies or membership dues, governmental and payer funding, and possibly other mechanisms. For the short term, limited support has been provided through TATRC, and additional sources of support are being sought. Most of the activities described in this report have been carried out through the in-kind contributions of time and effort by the participants and their organizations.

The initiative described here does not duplicate the work of Evidence-based Practice Centers, the Cochrane Collaboration, Comparative Effectiveness Research, or other efforts to determine best practices and to develop guidelines based on meta-analysis and evidence-based medicine. An operational knowledge repository would, however, regularly draw upon the work of these organizations to operationalize their findings. For the near term, the Morningside Initiative takes as its starting point the “operationalized” rules-based knowledge that health care provider organizations have already determined to be useful and have implemented in various applications in their systems, usually in the form of decision rules for alerts, reminders, or medication prescribing recommendations, or as order sets. Typically these CDS modules have been drawn from guidelines, authoritative reviews or other evidence-based medicine sources, but they have been made unambiguous and computable – a process which sounds straightforward but is definitely not. Further, typically these modules are not represented in a language that can be interpreted by other systems or applications, even within the same organization. So a major focus is to develop a common shared representation for such knowledge.

A subsequent phase of the Morningside Initiative is anticipated, which includes a national-scale effort to maintain and update a shared repository of executable knowledge. This would include a regular process to review and incorporate new knowledge coming from such authoritative sources as above. However, that is not the initial focus of the effort.

In 2008, at about the same time as the Morningside Initiative was being formed, AHRQ issued contracts for two projects, the Guidelines into Decision Support (GLIDES) project and the Clinical Decision Support Consortium (CDSC). Like the Morningside Initiative, both GLIDES and CDSC are concerned with the lifecycle of transforming knowledge from narrative form to implementation. GLIDES is particularly focused on narrative guidelines and using the Guideline Element Model (GEM) and GEM Cutter tools [28Shiffman R. GLIDES Project: Guidelines into Decision Support Slide presentation from the AHRQ 2009 Annual Conference. Rockville, MD. 2009; [Retrieved: 11 November, 2010]. Available from: http://www.ahrq.gov/about/annualconf09/shiffman.htm ] to implement narrative guidelines in testbed settings. The CDSC aims to develop cooperative approaches to share and manage CDS knowledge [29Middleton B. The clinical decision support consortium Stud Health Technol Inform 2009; 150: 26-30.] The CDSC, which includes health care organizations and vendors, is using a top-down approach to develop knowledge management models, representations, and lifecycle processes, based on the internal knowledge management portal development at Partners Healthcare, and a proprietary content management platform (Documentum, EMC Corp., Hopkinton, MA).

The Morningside Initiative, by contrast is committed to adopting an open, collaborative process for knowledge management in order to create a shared repository of content for CDS that would be as close to implementable form as possible. Further, a goal of this group is to follow best available models for knowledge representation and to help drive the standards development process where appropriate standards are lacking.

The Morningside Initiative is working in collaboration with the Congressionally-funded Knowledge Management Repository (KMR) project, directed by CAPT Emory Fry, MD, at the Naval Health Research Center, San Diego, CA (see: [30Fry EA. Infrastructure: DDSS and KMR Socratic Grid Project. 2010; [Retrieved: 11 November, 2010]. Available from: http://www.socraticgrid. org/index.php/documentation.html ]). The KMR project is aimed at developing an open-source platform for CDS for the Federal NHIN-Connect implementation. Thanks to this collaboration and a cooperative research and development agreement (CRADA), this NHIN-Connect hardware/software stack is installed at the ASU Clinical Application, Research, and Education Interoperability Testbed (CARE-IT) Laboratory [31Fridsma D. CARE-IT Lab. Arizona State University [Retrieved: 11 November, 2010]. Available from: http://care-it.asu.edu 2010.], and is available for Morningside Initiative work. CARE-IT is a collaboration between academic, government and industrial partners established to promote research into open-source, interoperable and standards-based solutions for health care information technology (HIT) and to provide an innovative infrastructure for the education of health care and technology professionals. We use this resource as the technical framework for hosting the CDS knowledge repository, delivering and ensuring interoperability of the components of this research – both the content repository, tools, and resources needed for their support – and providing a collaboration environment for participants.

Another collaboration that has evolved is with the Structuring Care Recommendations for Clinical Decision Support (SCRCDS) project. This project, originally called the Hardened Rules Project, is a contract funded by AHRQ to Thompson Reuters, Inc., with subcontracts to Morningside Initiative participants at ASU and Intermountain Health, among others (see: [32Agency for Healthcare Research and Quality. Structuring Care Recommendations for Clinical Decision Support AHRQ Publication No. 09-M048. 2009; [Retrieved: 11 November, 2010]. Available from: http://pbrn.ahrq.gov/portal/server.pt?open=18&objID=910228&qid=42777884&rank=2&parentname=CommunityPage&parentid=0&mode=2&in_hi_userid=8762&cached=true ]). A primary goal of that project is to develop a formulation for the United States Preventive Services Task Force (USPSTF) A and B-level prevention and screening recommendations (45 in all), that is unambiguous and sufficiently detailed to facilitate broad uptake and use by EHR implementers and vendors, as well as a small subset of rules based on Meaningful Use quality measures. This involves producing the rules in terms of well-defined data elements and logical constructs, but not incorporating setting-specific factors such as triggering conditions, workflow, and action details that limit their portability. Morningside Initiative participants are seeking to develop formal approaches to extending the output of the SCRCDS project to incorporate a systematic process for adding in the site-specific features, based on requirements of specific sites.

More recently, in 2010, ONC issued grants to institutions for four Strategic Health Advanced Research Projects (SHARP), One, known as SHARP-C, focused on Decision Making and Cognition, was obtained by the University of Texas Health Science Center at Houston to create a National Center for Cognitive Informatics and Decision Making in Healthcare [33University of Texas Health Science Center at Houston. National Center for Cognitive Informatics and Decision Making in Healthcare. [Retrieved: 11 November, 2010]. Available from: http://www. uthouston.edu/nccd/projects.htm 2010.]. A subset of the Morningside Initiative team involving ASU, Intermountain Healthcare, the VA Palo Alto Healthcare System, and the Naval Health Research Center is responsible for one of the projects, that of developing an “implementers’ workbench”. The goal of that effort is to create a resource for enabling CDS rules that capture evidence-based best practices but are not site-specific to be encoded with the triggering, data source, interaction, notification, and other workflow and setting-specific features necessary to be incorporated effectively into a specific site.

The Morningside Initiative’s lifecycle model for knowledge acquisition, refinement, and deployment (see Section B) involves 4 stages of evolution of knowledge content that is the basis for the implementers’ workbench. Fig. (2) shows the relationship of the Morningside Initiative to the foci of the KMR, SCRCDS, and SHARP-C projects. As can be seen, this evolution has enabled the Morningside Initaitive to become a kind of coordinating entity through which participants in other related projects are able to align their efforts.

Fig. (2) Life Cycle of Rules Knowledge Refinement. Note that stage 3 is typically skipped by implementers, such as those deploying Arden MLMs. The Morningside Initiative seeks to make stage 3 explicit, as a means of capturing the experience of successful implementations and also as a means to formalize the range of setting-specific options needed to successfully embed CDS into operational settings. The Morningside Initiative is seeking to do this using open source tools and methods and a collaborative process wherever possible. Relationships to three concurrent projects involving Morningside Initiative participants, the DoD-based Knowledge Management Repository (KMR) project, the AHRQ-funded Structuring Clinical Recommendations for Clinical Decision Support (SCRCDS) project, and the ONC-funded SHARP-C subproject 2B are shown. Two AHRQ-funded initiatives, the Guidelines into Decision Support (GLIDES) and the Clinical Decision Support Consortium (CDSC), have similar goals of supporting the lifecycle shown in this figure, but are using different approaches. GLIDES is using the Guideline Element Markup (GEM) tools to refine guidelines into decision support implementation. CDSC is developing knowledge authoring, sharing, and execution tools and processes based on a formal model of intermediate knowledge artifacts.

The ultimate purpose of the Morningside Initiative is to establish a continually updated and widely accessible national-level repository of authoritative knowledge in executable form. The details of the knowledge are to be fully transparent. That participation will be available to all is a guiding principle.

Since standards are still immature, the goal of being able to provide knowledge in standard form is not fully attainable. Nonetheless, the Morningside Initiative intends to represent knowledge in unambiguous form by using (or developing as needed) conventions for representation that are considered by the team to be the best available, and which are fully documented. The team also works with standards development organizations to help accelerate the adoption of standards that are most urgently needed. It is expected that as this effort gains traction it may provide an important use case for such standards development, to help drive the direction of the process.

A principle of the approach is that knowledge bases are managed using open source tools and non-proprietary data formats wherever possible. This avoids licensing fees or vendor dependence, or the delays in waiting for vendors to support particular needed features. Open platforms, e.g., the J2EE framework, have demonstrated scalability and performance, and there are growing numbers of open-source software products available for various needs.

The functional requirements for delivery define needs for the CDS, the process for selecting and rating the CDS for inclusion in the KR, the tools and methods for representing it, and approaches for contextualizing it for use. Ideally, we further impose that the content and tools be standards-based and open source, wherever possible, and that the processes involved in implementing the life cycle be that of an active, engaged, viral open source community of knowledgeable developers and users. A draft set of functional requirements (some of which are included in Table 1) forms a basis for our approach and is being further refined.

Table 1

Condensed Functional Requirements for a Knowledge Repository. Requirement Categories and Examples of Functional Requirements are Included. The Requirements Seek to Capture Elements of the Knowledge Model, the Knowledge Authoring Environment, the Knowledge Sharing Environment, and the Repository itself. An As-Yet-Unrealized Goal is to Specify Functional Requirements for a Knowledge Execution Component of this Environment

Table 2

Examples of Metadata for Indexing Content in a Knowledge Repository. The Collection is Derived from Metadata Described in a Number of Categorization Schemes for Computable Medical Knowledge

As noted above, the initial focus in the Morningside Initiative has been on CDS content which is already implemented in existing participant sites and which has been shown to be effective. The aim is to learn from examining that content what the unique features of each implementation are, to compare knowledge aimed at similar purposes, and to forge a common representation. Knowledge content foci are determined by the organization/steering committee, and are initially primarily aimed at high priority clinical conditions, notably involving chronic disease. We have further restricted our activities thus far to examination of diabetes rules-based knowledge, e.g., HbA1c testing, Metformin prescribing, eye and foot examination reminders, and operational definitions of chronic diseases.

Once having created a sharable representation, abstracting context-specific and site-specific aspects, the goal is to re-contextualize the content to enable it to be incorporated into another site with distinct EHR platform, organization, and workflow requirements. This analysis of content has to date been very instructive, and has identified the need to separate the core medical knowledge (the stage 2 representation, as per Fig. 2) from the business or workflow logic and to characterize a variety of annotation meta-tags (see Table 2) that relate to things like how the CDS is triggered, how and where it interacts with applications and users during evaluation, and how the results are communicated or inserted into the workflow (the stage 3 representation).

We have also shown that many apparently different rules are for the same medical purpose, once one does the above analysis. This is exemplified in Fig. (3), which compares knowledge content gathered from Morningside Initiative participant implementations dealing with various diabetes-related rules. For example, a variety of different rules are seen all dealing with how to deliver a reminder that HgbA1c testing should occur every 6 months. It is clear that the differences in appearance are all due to the triggering, presentation, and other business logic considerations. Thus a major task of this initiative is to formalize the content acquisition process in order to understand how to represent these layers.

Fig. (3) Comparison of example rules from Partners Healthcare, Intermountain Healthcare, Veterans Administration, and Kaiser Permanente dealing with HgbA1c assessment. Columns indicate individual rules (or disjunctive clauses of rules) by institution. The conditions potentially evaluated are enumerated in the blue section, with those included in a particular rule indicated by the value needed for satisfaction (N or Y). The actions to be performed (sending of a message or ordering of a test) on satisfaction of the conditions for a rule are indicated in the red and yellow sections of that column by the presence of a Y.

The key innovation in the technical approach is to put together a variety of emerging open source capabilities and to engage the community in an active collaborative mode in its development, testing, and refinement.

The organization of the Morningside Initiative is subject to a set of procedures set forth in a Bylaws document. The Morningside Initiative is a public-private collaboration organized under a memorandum of understanding (MOU) signed by all participants. The Morningside Initiative and AMIA have agreed in principle for the Morningside Initiative to be affiliated under AMIA as an intermediate means to formalize the entity and carry out the business on behalf of its constituency. This was done as a way to maximize flexibility for long-term operation and sustainability as the entity expands to be a fully inclusive national-scale initiative. Other organizational and/or governance structures or approaches may be considered in the future.

Draft bylaws and the proposed AMIA affiliation agreement have been developed. Bylaws include language that addresses indemnification and intellectual property protection of contributors of knowledge content, as well as the Morningside Initiative itself. Participants sharing content will need to be indemnified by clinical care organizations accessing and implementing this content in their respective clinical decision support systems. Further, those sharing content will be responsible for respecting their own third-party content intellectual property agreements as well as protected from having their shared artifacts sold by other third parties.

The original intent in forming the Morningside Initiative was to limit the number of participants initially, until the organization, content, and technical approaches were sufficiently defined, before expanding. We have learned much in the past year about the nature of the collaboration process, however, and that it may be possible to actually move more quickly to a broader scale, by fully embracing an open source, open collaboration model. A key reason for this is that we have found that the interest in this activity is much broader than we had originally believed, given the general reluctance to sharing of CDS over the past decade. A large part of this is due to the focus of ONC, AMIA, and various leading health care organizations on CDS, and the new vibrancy of open source communities in production environments, notably in health care.

The formal adoption of the bylaws has been slowed by the recognition that government participants such as the VA and DOD could not be signers. The current status is that the team is operating without a formal agreement in place, because of the flurry of activities in relation to the KMR, SCRCDS, and SHARP-C projects described in Section A.4. Through these projects, we have, in the short term, been able to further the goals of the Morningside Initiative through alignment of our participation in those projects, carried out under their auspices, with Morningside Initiative coordination, so that operation without the Morningside Initiative’s own bylaws in those contexts has not impeded progress. We expect that formal adoption of bylaws will be necessary once it becomes desirable to expand the scope of participants in a broader consortium.

Regarding knowledge content sharing, efforts has been focused to date on a subset of rules relating to chronic disease management, particularly diabetes, and all participants have contributed their content without restriction. Thus, given current focus on the modeling and tool development around a subset of content, the broader issues of knowledge content sharing have been sidestepped for now, even though the draft bylaws address them. Again, we expect that we will need to turn back to this once the scope of participation in the Morningside Initiative expands.

A long-term goal of the process described above is to create a resource that will become increasingly used and valued as more and more of the management and documentation of health care in the U.S. is carried out electronically. We are convinced that development of and adherence to standards are central to effective implementation of EHRs and CDS. We are committed to using existing standards in this knowledge repository and, where standards are insufficient or do not exist, we anticipate applying the lessons learned in the analysis and design of the repository to promote extension of standards or to create new standards as needed.

Two kinds of standards are particularly applicable for those who wish to share CDS knowledge. The first is the language in which this knowledge is expressed. In the realm of medicine, two overlappingHL7 standards exist. These are the Arden Syntax for medical logic modules [21Pryor TA, Hripcsak G. The Arden syntax for medical logic modules Int J Clin Monit Comput 1993; 10(4): 215-4.] and the GELLO expression language [34Sordo M, Boxwala AA, Ogunyemi O, Greenes RA. Description and status update on GELLO: a proposed standardized object-oriented expression language for clinical decision support In: Proc MEDINFO 2004. 164-8.]. However, neither of these has seen widespread adoption in the medical computing community, except in the case of Arden Syntax, within vendor-specific implementations. An interesting alternative to these standards is represented by the growing use of business rules engines and other, general-purpose decision tools in some clinical environments. It appears that a repository that supports a single, computable representation of CDS logic will not be adequate.

In this context, the Morningside Initiative has begun to explore the possibility of an XML-based Interlingua as a tool for translating among different logic representations. Examples of such tools include the W3C RuleML [35RuleML.org. The Rule Markup Initiative [Retrieved: 11 November, 2010]. Available from: http://ruleml.org/ 2010.] and a version of the Arden Syntax expressed fully in XML [36Kim S, Haug PJ, Rocha RA, Choi I. Modeling the Arden Syntax for medical decisions in XML Int J Med Inform 2008; 77(10): 650-.]. One of the goals of this effort will be to evaluate a subset of the existing (medical and nonmedical) standards to determine if the development of a medical rules interchange format is possible. The key goal of such an interchange format would be to support translations of CDS logic between languages.

In this activity we anticipate leveraging the work of others. The Morningside Initiative and KMR collaborators have regular interactions with a variety of standards organizations and expect to promote the adoption or development of a medical rules interchange format if further analysis confirms the feasibility of this approach.

The second kind of standard that is essential to the sharing of CDS logic is a mechanism to readily adapt decision logic built using data of one institution for use with data queried from the clinical database of a second institution. The central challenge when importing medical logic into a new clinical setting is to bind this logic to local data models and terminologies. This problem is known generically as the "curly braces" problem, arguably a key reason for the limited dissemination of the Arden Syntax as a mechanism for exchanging medical knowledge.

The Morningside Initiative plans to approach this challenge through analysis of two evolving standards. One of these, the virtual medical record (vMR) offers the option of mapping a nonstandard data representation onto a simplified version of the HL7, version 3, data messaging standard. The other, a product of the Healthcare Services Specification Project (HSSP), standardizes CDS as a service. The mapping of local data to the symbolic forms used in the CDS logic occurs during the construction of the message used to invoke the service.

We believe that a key product of our future work will be examples, use cases, and functional requirements designed to support the implementation of decision logic, imported from the repository, in settings where data models and terminologies differ from those present in the originating site.

We are exploring this concept in two ways. One is adaptation to local processes/workflows, and the other is adaptation to local data representations. In these activities, we seek to use standards wherever adequate. We intend to demonstrate the ability to technically interface the CDS into different platforms. One target is the DOD AHLTA system being explored through the KMR project. Another, to be done between KMR and the ASU CARE-IT lab, is the Resource and Patient Management System (RPMS) of the Indian Health Service. In both cases, the approaches to be pursued involve use of the vMR and the HSSP efforts to model decision support as a service.

Since our collaborative process is intended to be an open one, a key dissemination strategy will be through use of a web portal. We intend to promote its use through panels and presentations at national meetings and online announcements. Through instrumentation of the web portal and inclusion of Web 2.0 assessment tools which are part of the BioCore portal design, we will track, monitor and assess content growth/update, and assess usability of tools and processes. We will also monitor cost of participation of each site and of coordination/central Morningside Initiative activities; usage logs and internal adoption by sites from repository; and governance issues/conflicts/resolution, and will publish and disseminate our results.

Ultimately, a national-level initiative will have participants in both public (federal/non-federal) and private (commercial/non-commercial) sectors, and with potentially many kinds of stakeholders (providers, payers, professional societies, government agencies, standards organizations, knowledge providers, and health systems providers, to name a few). One model for sustainability would thus be long-term commitment for support by a Federal agency, much as AHRQ supports the EPCs. An alternative is to establish a means for such an initiative to obtain support from its constituency. One model is that used by membership organizations, such as AMIA and HL7. These have both institutional and individual membership categories, with benefits associated with each, but do not limit access of members to their products and services based on type of user. Rather, the fee model essentially offers “quantity discounts” based on numbers of members. Such organizations have boards of directors that govern them, and which are open to broad participation through democratic processes.

The Morningside Initiative is a first step in bringing together on a voluntary basis organizations that have taken a leadership role in developing, deploying, and demonstrating the value of CDS for health care, to pool their expertise for the purpose of jump-starting a national knowledge sharing activity. The road ahead is long, but we have already learned that much can be gained by traveling together.