Clinical Decision Support: Defining the Right Strategy, Making the Right Decisions

CMIO_0103_01Although the entire healthcare IT universe is waiting on a final definition of meaningful use of electronic health records, the July draft circulated for public consumption gives plenty of clues as to what the rules ultimately will look like for the Medicare and Medicaid health IT subsidy program.

We won’t know the ultimate parameters until spring 2010, but it’s a pretty safe bet at this point that clinical decision support (CDS) will be included. According to the proposed matrix for meaningful use, hospitals will need to implement one CDS rule for a high-priority condition to qualify for federal stimulus funding in their first year of participation. HHS has said that it probably will apply the 2011 standards to a provider’s first year of participation, even if it’s as late as 2014.

And the requirements will get more stringent in 2013 and 2015 as CMS and the Office of the National Coordinator for Health Information Technology (ONC) encourage the adoption of advanced electronic health records (EHRs) to drive quality improvement.



Guidelines to drive decisions

The Healthcare Information and Management Systems Society (HIMSS) broadly defines CDS as a clinical system, application or process that aids practitioners in making proper clinical decisions to enhance patient care. It’s more inclusive than evidence-based medicine, which generally means delivering care based on the most current evidence directly relevant to a patient’s current condition.

“Clinical decision support is providing information that’s tailored to the patient and the workflow and delivered for maximum impact,” says Jonathan Teich, MD, PhD, CMIO in the health sciences division of professional publisher Elsevier. “CDS is about putting the information in front of you to solve the problem at hand.” Evidence-based medicine, according to Teich, is the content or guidelines that drive decision support.

“Decision support is all about adding in and getting patient-specific information rather than just getting flat text,” explains Art Papier, MD, chief scientific officer of Logical Images, the Rochester, N.Y., company that produces the image-based VisualDx diagnostic decision support system.

In years past, many dismissed the concepts of evidence-based medicine and CDS as “cookbook medicine” that forced doctors to follow prescribed actions that weren’t necessarily in the best interests of the patient in question. Teich says this criticism seems to be subsiding, as long as the evidence and support are properly structured. “In the aggregate, it does help you narrow the percentages,” he says, but that only works when there is good evidence on a condition or symptom.

“Clinical decision support is a way of informing practice,” says Gregory H. Dorn, MD, executive vice president and chief operating officer of Zynx Health, a clinical decision support company founded at Cedars-Sinai Medical Center in Los Angeles. Dorn calls CDS a regularly changing summary of current evidence rather than a single, fixed piece of information. “The evidence is perishable. It changes over time,” he says.

The goal of CDS is to standardize processes and evidence. “There are areas where the evidence can inform practice, and there are areas where the evidence does not inform process,” Dorn says.

“When there’s no solid and universal evidence, you do have to rely on experiences,” Teich says. “If there are no national guidelines, I’ll rely on what the chief of pulmonology recommends,” he adds, as an example.

These days, it’s getting rare to find an EHR vendor that doesn’t include evidence-based protocols and clinical decision support in its product. GE Healthcare in October announced that it was going to commercialize the CDS system developed at Intermountain Health Care in Salt Lake City, along with the help of Mayo Clinic in Rochester, Minn. The CDS pilot launched in November at Intermountain. The new CDS will be branded under GE Centricity and officially launched at the HIMSS meeting in March 2010.

It is entirely possible that any product lacking CDS will fall short of national certification standards when the federal stimulus program begins, although many organizations that take a best-of-breed approach rather than purchasing an end-to-end system from a single vendor have successfully interfaced CDS with their other IT systems.

An excellent example is Maimonides Medical Center, in Brooklyn, N.Y., which has a NextGen ambulatory electronic medical record (EMR), an EMR for the emergency department from A4 Health Systems—now a part of Allscripts-Misys Healthcare Solutions and an E&C Intelligent Record for OB by PeriGen in obstetrics. For nearly 15 years, most other inpatient departments have run an E7000 system, originally a product of Technicon Data Systems and now owned by Eclipsys.

E&C had clinical decision support “integrated into the fabric of the product,” Davidson says. “That’s why we bought it. It is very sophisticated.” However, he says it has been tougher for physicians to get used to that system than other departmental EMRs.

Maimonides soon will be standardizing on a newer Eclipsys product, Sunrise Clinical Manager, and will be installing the Zynx Health CDS database. But the hospital won’t just dispense with the rules it already has in place. The obstetrics system alone, installed in 2003, now contains some 2,000 rules, says Steven J. Davidson, MD, who chairs the Department of Emergency Medicine at Maimonides and serves as the in-house medical informatics guru.

 Zynx Health was bought by Cerner in 2002. The connection to Cerner caused customers of other EMR vendors to shy away, so Cerner sold Zynx to publishing giant Hearst Corp. two years later.

It wasn’t a completely futile exercise, however. “With Cerner, we learned how to integrate CDS into technology,” Dorn says.

Dorn breaks CDS down into four domains: order sets; plans of care; rules and alerts; and at its most basic, reference information. The last one is less effective than the others, since it tends to be passive support, while the others are active forms of CDS. “We come at CDS with a broad library in each of these domains, and [customers] can choose which they want in their local practice and select custom order sets or care plans or reference information locally,” Dorn says.

Zynx believes in expert independence. “The field of evidence-based medicine is tied to a rigorous process to extract that evidence,” according to Dorn.

There’s a similar philosophy at Logical Images, as VisualDx has more than 100 physicians on its editorial board to curate the database of some 70,000 images, of which 16,000 are actually in the product. “We have, we would argue, the best collection of medical images in the world,” says Papier, who also is associate professor of dermatology and medical informatics at the University of Rochester College of Medicine in New York.

VisualDx launched in 2001 with a single module. There are now 30 modules in the system, most related to dermatology. Each helps physicians make pattern-recognition diagnoses.

In the past, doctors would just flip through an atlas. “But those books are organized by disease,” Papier says. “What you really need are images organized by symptom,” particularly those symptoms widely seen in primary care or emergency medicine. “You can search by patient factors.”

Also, skin conditions look different on every patient. “Decision support systems need to capture that variation,” Papier says.
 

CPOE or CDS – which comes first?

It’s possible to implement computerized physician order entry (CPOE) without clinical decision support, but not advisable, experts agree. “If you don’t have good CDS, you just automate current processes,” Dorn says. “You deliver mediocre care faster.”

Similarly, CDS is still a powerful tool as an interactive reference, support for e-prescribing and various forms of analytics, even in the absence of an EMR. But linking the two unleashes the power of both.

A serious issue with CDS, however, is overcoming physician reluctance to change, particularly when a computer system fires so many alerts that users start to ignore them. (See “Overcoming Alert Fatigue” on next page.) Dorn recommends making alerts as specific as possible targeting areas with the greatest opportunity for improvement. “You probably want to test these alerts in the background before you make them live,” he says.

Resistance to CDS comes back to workflow and efficiency, according to Teich. “In our experience, people are willing to welcome advice if it helps their decision-making,” he says. “I think it’s a lot about the delivery.”

The content has to be clinically rigorous and, ideally, well-integrated into the EMR, according to Dorn. “Like any drug or device, you need to understand the effectiveness of it,” Dorn says.

Failures more often are a result of change management than what the screen looks like. “When you do CDS, you have to focus on what’s important without hindering the daily work of the provider,” says Teich, who was a key figure in the development of the in-house EMR and related CDS system at Brigham and Women’s Hospital in Boston.

“I think it’s really important to understand the culture that you’re working with,” Teich adds. He also recommends to those seeking to implement CDS that they visit sites similar to theirs to see what others have done. What works? What doesn’t?
 

Gaining physician mindshare

There are a number of ways to get physicians on board with a CDS program. Maimonides offers a nice enticement for following hospital policies on CDS and CPOE. “The full-time medical staff can join with the hospital’s malpractice insurer if they adhere to the requirements of the program,” Davidson says.

And then there are order sets, letting doctors enter one order for a series of actions associated with common conditions and procedures. “Order sets are used for a couple of purposes,” Teich says. They help make care more efficient and easy for the majority of patients without outlying conditions, improve the standardization of care and remind doctors of safety measures they should be taking. “Doctors like them because they are quick.”

Order sets work best for the small number of conditions that the typical physician sees high volumes of. “There’s no order set for familial Mediterranean fever,” Teich notes.

CPOE and CDS are “not paternalistic,” according to Davidson. Properly implemented, they won’t tell doctors how to make decisions, just help them make better ones. And just like with CDS itself, it’s important to get physicians involved in building order sets.

In general, Teich advises, make sure people in each department have their say in developing and refining order sets. “There’s a lot of local flavor,” he says.

Many members of the Maimonides medical staff worked with the MIS team and were involved in the decision-making. “We had doctors who really wanted a sense of ownership on this,” Davidson says. “The medical staff was certainly engaged.”

In migrating from paper to electronic order sets, strategy often depends on the skill of clinicians in learning the editing system, as ?well as the quality of paper-based protocols the electronic system is replacing. Teich believes that things that are fuzzy or not well defined on?paper—dosing that says “as needed” or advice that says “consider”—are difficult to migrate to electronic clinical decision support.

Be as specific as possible, he advises, and start with something that’s relatively easy, such as discharge orders. “Go after the low-hanging fruit to show people that it works,” Teich says. “Once you’ve been through a few, it becomes much easier.”

A good system pushes information to the user rather than require the user to “pull” by going outside the normal work processes, Teich says.

Dorn, of Zynx and Cedars-Sinai, recommends that hospital teams should perform gap analyses and augment current practices with evidence, so there’s an alignment between CDS and organizational culture. “You’re not discarding the institutional memory,” he says. This process also draws in clinicians.

It’s a “delicate” process to get the right amount of precision to alerts and volume of order sets, Dorn says. “There’s an emerging body of evidence over how many order sets are required,” Dorn says, but the correct answer if there is such a thing, is more than 50 but less than 1,000. “Two hundred is a popular number.” He says that should cover about 80 percent of typical diagnoses for a typical acute care hospital.

 

Diagnostic & Management Error: The Missing Link
Clinical decision support really goes far beyond ordering support and rules engines, according to Art Papier, MD, chief scientific officer of Logical Images, which developed the VisualDx diagnostic decision support system. “Order entry, dosing and surgical errors are only the tip of the iceberg,” he says.

Below the water line, as Papier calls it, are diagnostic and management error. Leading authors in the field of patient safety, including Jerome Groopman, MD, Atul Gawande, MD, and Lisa Sanders, MD, have written extensively about the problem of diagnostic errors, and at least one CDS company, Isabel Healthcare, is singularly dedicated to addressing this category as well. Now, VisualDx is taking a close look at diagnostic decision support as well, Papier says. The majority of vendors, however, are still focused on CDS exclusively for order entry.


Overcoming Alert Fatigue
CMIO_0103_01aThe literature on clinical decision support is decidedly mixed.

A study from the Netherlands, reported in the Oct. 18, 2007, issue of the Journal of the American Medical Informatics Association, showed the effects of a poorly planned CDS system. The Sophia Children’s Hospital at Erasmus Medical Center in Rotterdam pulled the plug on clinical decision support after finding that the system helped increase adherence to best practices regarding pediatric patients with fever, but it also led to longer stays in the emergency department and the ordering of more laboratory tests.

More recently, two studies published in the same journal in September reach interesting conclusions about how to improve clinical decision support. One suggests that the phenomenon known as “alert fatigue” or “alert inappropriateness”—the problem created when a computer system overwhelms clinicians with so many alerts that the user tunes out the warnings—is a bigger risk to efficiency than safety. The other highlights a workflow problem: that physicians need to be trained to read and respond to their electronic messages.

Writing in the Sept. 14 issue of Archives of Internal Medicine seems to question this hypothesis. Researchers from the Center for Patient Safety at Boston’s Dana-Farber Cancer Institute and other Massachusetts institutions report finding that drug interaction alerts in ambulatory electronic prescribing systems are effective at preventing adverse drug events (ADEs) and reducing medical costs, even when there are excessive alerts.

The paper, based on the study of close to 280,000 electronic alerts delivered to Massachusetts prescribers in 2006, found that e-prescribing systems generated alerts in 7.3 percent of the 1.8 million e-prescribing attempts examined, but physicians manually overrode a staggering 91.1 percent of the 133,051 alerts that qualified for analysis.

Still, the nearly 12,000 alerts that were accepted likely prevented 402 adverse drug events, three deaths, 14 permanent disabilities and 31 cases of temporary disability. The warnings also may have averted 39 hospital admissions—at an average cost of $9,000 per admission—kept 34 people out of ERs and avoided 267 physician office visits, for an overall savings of $402,619, the researchers report.

“Although overriding alerts may jeopardize the potential impact of these systems, it is possible that even the small number of accepted alerts may reduce patient harm, decrease unnecessary utilization of healthcare services and save money,” the study says.

Clinicians did have to see 331 alerts to prevent one adverse drug event and 2,715 alerts to avoid a death or disability, suggesting that alert fatigue is real, though it may not lead to widespread safety problems. “[W]e believe that the technology’s ability to prevent ADEs makes it worthwhile, and our findings suggest that significant efficiencies could be gained by reducing overalerting,” they conclude.

“Doing so would mitigate alert fatigue, thereby increasing the percentage of clinically significant alerts accepted and the number of ADEs averted. Previous studies have demonstrated that tiering alerts and interrupting prescribers for only the most serious warnings are effective strategies for increasing alert acceptance rates.”

Meanwhile, a team from the DeBakey Veterans Affairs Medical Center and several other Houston-area teaching hospitals report in the Sept. 28 Archives of Internal Medicine that physicians often failed to respond in a timely manner to electronic messages alerting them of abnormal diagnostic imaging test results. Of the 1,196 alerts examined, 18 percent went unopened for two weeks and nearly 8 percent were ignored for at least four weeks, suggesting that more than 200 patients in the study group didn’t get timely follow-up.

“Critical imaging results may not receive timely follow-up actions even when [providers] receive and read results in an advanced, integrated electronic medical record system. A multidisciplinary approach is needed to improve patient safety in this area,” the researchers conclude.

Neil Versel joined TriMed in 2015 as the digital editor of Clinical Innovation + Technology, after 11 years as a freelancer specializing in health IT, healthcare quality, hospital/physician practice management and healthcare finance.

Around the web

As debate simmers over how best to regulate AI, experts continue to offer guidance on where to start, how to proceed and what to emphasize. A new resource models its recommendations on what its authors call the “SETO Loop.”

FDA Commissioner Robert Califf, MD, said the clinical community needs to combat health misinformation at a grassroots level. He warned that patients are immersed in a "sea of misinformation without a compass."

With generative AI coming into its own, AI regulators must avoid relying too much on principles of risk management—and not enough on those of uncertainty management.

Trimed Popup
Trimed Popup