In Pursuit of the Holy Grail: Scalable, Interoperable Clinical Decision Support

Nov. 8, 2013
When the physicians in the ambulatory clinics of Wishard Health Services in Indianapolis get an alert or reminder in the computerized physician order entry (CPOE) module of their homegrown electronic health record (EHR), they may not be aware that the alert was generated at Partners HealthCare System in Boston, and sent to Indiana as a Web service.

When the physicians in the ambulatory clinics of Wishard Health Services in Indianapolis get an alert or reminder in the computerized physician order entry (CPOE) module of their homegrown electronic health record (EHR), they may not be aware that the alert was generated at Partners HealthCare System in Boston, and sent to Indiana as a Web service.

The project extracts a limited data set about a patient, including labs and allergies, in the form of a Continuity of Care Document (CCD), which is sent to Partners. Its system adjudicates the rules against that data, and sends back information on which reminders or alerts fire.

Although there was a considerable amount of technical and policy work behind the scenes to make it happen, “the physicians’ experience is that the alerts are not a whole lot different than homegrown ones — and that is the goal,” explains Brian Dixon, a research scientist at the Regenstrief Institute in Indiana. “By demonstrating that clinical decision support (CDS) can work as a Web service, we are not trying to do something magically different. The value is in the economies of scale.”

Brian Dixon

The meaningful use incentives are not enough money for all hospitals, especially community hospitals, to create their own systems for decision support, says Dixon, who is also an assistant professor of health informatics at the Indiana University-Purdue University Indianapolis School of Informatics and Computing. “They cannot form the committees or have knowledge management and informatics people to decide on 300 rules.” Then there are the IT requirements that stem from those decisions, and getting the rules into Epic, Cerner, or whatever EHR you are using, he adds. “That is time-consuming and costly.” 

Having academic health centers provide CDS as a service is one potential solution. “The idea of a service is that it can be robust content at a reasonable cost that multiple health systems on a regional or national level could take advantage of,” Dixon says, “and a small group could maintain the knowledge base. That is the dream.”

It’s easy to find discouraging statistics about CDS. Outside of four or five leading integrated health systems and the Veterans Administration, robust CDS use is still quite rare beyond alerts for drug-drug interactions and drug-allergy contraindications. Stage 2 of meaningful use requires that providers implement five clinical decision support interventions related to four or more clinical quality measures at a relevant point in patient care.

Yet most providers know they are nowhere near where they need to be on CDS, claims Jerry Osheroff, M.D., principal at TMIT Consulting in Cherry Hill, N.J. “Even the ones held up as beacons of light realize they have a long way to go,” Osheroff says.

Regenstrief’s Dixon adds that although vendors offer tools for drug-drug interactions, “there really isn’t anything comprehensive you can buy off the shelf in one place. You still have to do a lot of configuration, maintenance and updating.” Smaller providers will have to rely on EHR vendors to provide out-of-the-box rules or order sets that they can quickly implement, just to get something in place, he says. “Unfortunately, just as in Stage 1,” he says, “we will see some people focusing on checking the box, rather than what will actually improve healthcare.” 

Osheroff agrees there are ways people can find to check the box to do five interventions, but adds that, “to really attack this systematically and follow the spirit of meaningful use and improve information flow is going to take a lot more effort.”

Yet despite the difficulty provider organizations and vendors have had in designing CDS interventions that don’t annoy physicians or lead to “alert fatigue,” there is a lot of exciting work taking place in CDS research, and many health informaticists see themselves in the early stages of an important journey.

Osheroff himself is working diligently to get more CIOs and CMIOs working together on CDS. He heads up the Collaborative on CDS for Performance Improvement, a nationwide volunteer initiative encompassing more than 130 individuals representing dozens of hospital organizations and numerous EHR vendors. The collaborative is an outgrowth of work on the second edition of a book called “Improving Outcomes with Clinical Decision Support: An Implementer’s Guide,” published by HIMSS. In the last year, Osheroff also has worked as a subcontractor to Deloitte on worksheets and implementation guides the Office of the National Coordinator will offer on for CDS-enabled performance improvement to help providers with meaningful use, Stage 2.

Osheroff says wider use of CDS will require both enhancements to technical standards and more process improvement work. “You need both tools and strategy,” he says. “You need a rich marketplace of high-quality interventions and you also need a rich ecosystem of providers working on performance improvement together in a sharing environment to get better faster.”

Clinical Decision Support Consortium 2.0

The project between Regenstrief and Partners is just one piece of a larger effort, the Clinical Decision Support Consortium (CDSC), a five-year project funded by the federal Agency for Healthcare Research and Quality (AHRQ) to find ways to make CDS knowledge more easily shareable. The effort actually started as an effort to share CDS modules more easily within Partners, which is well known as a pioneer in CDS.

“We had been getting value from CDS at Partners for several years,” explains Howard Goldberg, M.D., senior corporate manager for enterprise clinical informatics infrastructure services at Partners. “But we had little CDS modules all over the institution, and they were hard to maintain and to make sure you were on the most updated version. We wanted to get to one source of truth, so several years ago, we decided to create a central service for all the ambulatory EHRs in use.”

Once Partners created a CDS Web service for use internally, AHRQ got interested, Goldberg says. “They said it is wonderful that Partners is doing this, but nobody else benefits. How can we find a way to share the benefits on a regional or national scale?”

Besides the work with Regenstrief, CDSC projects also include work with the Horsham, Pa.-based EHR vendor NextGen and its customer WVP Health Authority in Salem, Ore., as well as the University of Medicine and Dentistry of New Jersey (now part of Rutgers, the State University of New Jersey, in New Brunswick), a GE Centricity customer. “We are all learning how to do this together,” Goldberg says. “There are technology challenges. They have to be able to consume the CDS as SOAP [Simple Object Access Protocol] services in the EHRs. And that is easy or difficult depending on your platform.”

CDSC Version 2.0

Although CDSC’s federal funding ran out in 2013, the project is far from over. Blackford Middleton, M.D., who led the effort while at Partners, has since moved to Vanderbilt University in Nashville, Tenn., and is working on creating CDSC Version 2.0 there. “I got institutional support to rebirth it and continue that work here,” says Middleton, assistant vice chancellor for health affairs and chief informatics officer for Vanderbilt University Health System.

Middleton says he is concerned that the great potential of CDS is not being realized yet. “Any hospital that buys health IT has a dickens of a time with decision support to transform healthcare because the tools are so difficult to use, and informatics expertise is in short-supply,” he says. “It’s like buying Excel with no macros or function keys.”

Blackford Middleton, M.D.

There are vendors such as FDB (First DataBank, South San Francisco, Calif.) working on drug-drug interactions, but Middleton says he is worried that certain aspects of CDS are not being addressed by the marketplace well or at all, and he fears they may never be.  “For instance, creating and maintaining the knowledge base to do pharmacogenomic drug dosing in ambulatory care requires a whole academic medical center,” he says.

He envisions the development of an ecosystem of a few large academic medical centers, including Partners and Vanderbilt, maintaining and providing CDS content to smaller hospitals and provider organizations via Web services.

Of course, organizations that do this work to curate, encode and make the content machine-readable and executable will want to be paid to sustain their efforts. Middleton says that perhaps there could be two levels of service, one that is free to get people started and then tiers or levels for fees.

Referring to the meaningful use requirements, Middleton says five interventions might not sound like too much to ask, but there are multiple detailed requirements that will make it a challenge for some providers. “Of course, there is also a lot more to do beyond that. We are just scratching the surface, and the challenges of meaningful use Stages 2 and 3 will just further highlight the need for academic medical centers to start making CDS knowledge artifacts available at low cost or even free. It’s part of our mission,” he says.

Health eDecisions’ Standard Development Work

The Office of the National Coordinator for Health IT also has turned its focus to the “holy grail” of scalable, interoperable clinical decision support. Its effort, a public-private partnership called Health eDecisions (HeD), “is seeking to define and validate standards that enable clinical decision support at scale,” said Kensaku Kawamoto, M.D., Ph.D., the initiative’s coordinator.

Kensaku Kawamoto, M.D., Ph.D.

Kawamoto, who is also associate chief medical information officer and director of knowledge management and mobilization at the Salt Lake City-based University of Utah, says one need recognized early was a standard data model that is simple and intuitive for a typical CDS knowledge engineer to understand and use. After considering several possibilities, the team chose the HL7 Virtual Medical Record (vMR), which is a simplified and computable representation of the clinical record relevant for CDS.

One focus of HeD has been defining and validating a standard, shareable format for CDS knowledge artifacts, including order sets, documentation templates, and decision rules.  HeD has completed several pilot projects in this area, in which CDS content from four suppliers (Zynx Health, newMentor, CDC, and Wolters Kluwer Health) were generated in a standard HL7 format and consumed by EHR vendors. For instance, newMentor developed an artifact around a National Quality Forum rule on the use of aspirin or other antithrombotic for patients with ischemic vascular disease, which was translated and consumed in the Allscripts EHR.

In addition, HeD has defined a standard approach for EHR systems to access and consume CDS as a Web service. This type of approach has been evaluated by several groups, including an open source CDS organization Kawamoto oversees at the University of Utah. OpenCDS (, which has members from over 150 organizations, enables the provision of CDS as a service using the HL7 vMR and Decision Support Service standards. Kawamoto says an example implementation of OpenCDS is the Immunization Calculation Engine (ICE), which is being put to use by organizations including the Alabama Department of Public Health, the New York City Department of Health and Mental Hygiene, and eClinicalWorks. The ICE Web service evaluates a patient’s immunization history and generates the appropriate immunization recommendations for the patient. The use of the standards defined by HeD could enable such CDS Web services to be widely adopted at scale.

On a practical level, Kawamoto says, organizations may be hesitant to consume decision support as a service if it involves sending patient data out to another institution’s servers, even if it is de-identified. But the creation of solutions built on open source tools such as OpenCDS may allow those organizations to efficiently deploy the services locally and within their own firewalls. Also, Kawamoto notes that “software as a service” is relatively commonplace today in healthcare, and that the consumption of remote CDS services could easily follow a similar pattern of adoption. —David Raths

Sponsored Recommendations

Going Beyond the Smart Room: Empowering Nursing & Clinical Staff with Ambient Technology, Observation, and Documentation

Discover how ambient AI technology is revolutionizing nursing workflows and empowering clinical staff at scale. Learn about how Orlando Health implemented innovative strategies...

Enabling efficiencies in patient care and healthcare operations

Labor shortages. Burnout. Gaps in access to care. The healthcare industry has rising patient, caregiver and stakeholder expectations around customer experiences, increasing the...

Findings on the Healthcare Industry’s Lag to Adopt Technologies to Improve Data Management and Patient Care

Join us for this April 30th webinar to learn about 2024's State of the Market Report: New Challenges in Health Data Management.

Findings on the Healthcare Industry’s Lag to Adopt Technologies to Improve Data Management and Patient Care

2024's State of the Market Report: New Challenges in Health Data Management