Skip to main content

Emphasis on Support in Decision Support

By Greg Freeman for HealthLeaders Media  
   May 14, 2012

This article appears in the May 2012 issue of HealthLeaders magazine.

Computer-based clinical decision-support systems offer great opportunities to improve care and reduce costs, but healthcare leaders have to remember who's ultimately in charge: the human operating the computer. Implementing even the best technology for decision support can become a costly, frustrating failure that ultimately degrades patient care if you don't factor in the human element.

That was one of the lessons learned when Penn Medicine in Philadelphia adopted a computerized physician order entry system. Penn Medicine used the Eclipsys Sunrise Clinical Manager to achieve 100% CPOE in the inpatient setting. In addition, 1,800 physicians actively use the Epic electronic medical record system in the ambulatory setting.

Physicians make about 15 million hits per year in Penn's internally developed physician portal to view patient information and results. All physicians have access to an internally developed data warehouse that maintains 2.4 billion rows of data to help ensure patient safety and quality care, as well as support clinical trials and research.

That success would not have come without ensuring clinician participation by involving them from the start, says Michael Restuccia, vice president and chief information officer of Penn Medicine, a $3.8 billion-a-year academic health system with nearly 2,100 physicians providing services to the Hospital of the University of Pennsylvania, Penn Presbyterian Medical Center, Pennsylvania Hospital, and more than 250 ambulatory practices.

"The biggest concern for clinicians was how much time it would take for them to use this system, and how it would impact their work flow," Restuccia says. "We were successful because we worked side by side with our physician community to make sure that what we developed actually was developed by them."

Physicians and nurses are included in the planning, implementation, and optimization of all technology initiatives, Restuccia says. By incorporating their feedback, Penn Medicine created customized solutions that addressed the needs of the users and generated enthusiasm and support for the projects.

A concern for the physicians was that they would still be able to use their judgment when appropriate rather than having a computer dictate patient care to them, Restuccia explains. Also, Penn Medicine assigned dedicated project managers to each of its physician practices during IT implementations.

"Designing decision-support systems requires the coming together of two very different skill sets," Restuccia says. "Clinicians know how to deliver care, and they know the work flow of their employees and coworkers. They generally know what will work and what won't. Then you need to partner that with the skills of the information services team, who tend to be structured, programmatic, detail oriented, and they're management oriented."

Restuccia recalls how it took three months of discussion with all the interested parties to determine what type of device would be used for entering documentation into the new system—a device in the room, a cart that would be wheeled from one room to another, or a device at the nursing station. The consensus was to use a device in each room. The partnership with clinicians, he says, is the reason for success.

"If I had told them it was going to be a cart because I unilaterally decided that was the best option, the installation would have been a disaster," he says. "In fact, I'm sure that if I told them it was going to be a device in each room—the same decision they came to—it still would have been a disaster because we didn't take the time to get the buy-in."

The results were impressive:

  • A 38% decrease in preventable deaths over several years, ranking the Hospital of the University of Pennsylvania as a national leader in this category, due in part to electronic alerts in patients' EMRs. For example, the system requires clinicians to consider the risk of venous thromboembolism for all patients they admit and offers support ordering prophylactic medications.
  • Reduction in many hospital-acquired infections by introducing evidence-based initiatives whose outcomes are built in to the inpatient CPOE system.
  • A 50% improvement in the appropriate use of blood thinners and a 10% reduction in complications through the introduction of evidence-based order sets that were designed in part by physicians.
     
  • Increased internal hospice referrals within the health system by more than 24% in a 12-month period supported by changes to inpatient discharge documentation and electronic referrals.
     
  • Increased ambulatory patient volume by 6.8% through the opening of the new Perelman Center for Advanced Medicine facility, which uses an EMR system throughout the facility.
  • Improvements in patient flow based on interventions that have resulted in significant decreases in length of stay, transport time, and bed-turnaround time.


At Adventist Health System, based in Altamonte Springs, Fla., CMIO Phil Smith, MD, says the system's clinical decision support is provided in two ways. First, the system provides order sets, reminders, and suggestions to clinicians regarding patient care, and then, after orders are placed, the system checks for allergies, drug interactions, and other possible problems.

Adventist Health System implemented CPOE using clinical decision support from Zynx Health and measured impressive results in only 27 months at 26 of its hospitals, with about 9,000 providers. The results included:

  • An 11% decrease in length of stay and 16% decrease in costs per case for heart failure patients.
     
  • A 95% reduction in callbacks from pharmacists to physicians to clarify orders, since medication orders are now submitted electronically through the CPOE system.
     
  • 100% CPOE and electronic documentation adoption in the emergency departments of all hospital sites. (Overall CPOE adoption rates are measured at 87%—well beyond the 30% noted in meaningful use requirements.)

Getting there meant designing a system that clinicians would actually use because it fit their needs and was practical in everyday use, Smith says. Adventist Health System formed 10 different committees consisting of physicians in different specialties, nurses, clinical pharmacists, and others who would study the literature to determine best practices for various conditions and situations, plus reminders for what not to do. Those reminders are useful for alerting doctors that a particular treatment has become outdated or is no longer appropriate.

"We look for opportunities to guide behavior in the direction we want to go. An example would be building aspirin into the order set for emergency room chest pain and adding an alert that says, ‘It looks like you're treating a heart attack and haven't given an aspirin. Do you want to order aspirin or is there a reason not to?'" Smith explains. "It's not only reminding the doctor, but it's also giving the doctor an opportunity to document why it might not be appropriate with this patient."

Out of every 100 orders for medication, about 10 have an alert that changes the doctor's behavior by canceling the drug, changing the dosage, or adding a monitoring test, Smith says. Last year, Adventist Health System clinicians ordered 6.6 million medications by CPOE, and there were about 650,000 alerts that changed behavior.

Those are encouraging numbers, Smith says, but Adventist Health System also tries hard to avoid "alert fatigue" in which so many alerts fire during CPOE that clinicians ignore them. The provider monitors the override rate for each alert. In March 2011, the rate climbed from the typical two overrides per 100 alerts to almost nine overrides per 100. That was a sign of alert fatigue, Smith says, so the committees went back and revised some alerts to make them more effective.

Striking a balance between too many nuisance alerts and not enough alerts to keep the patient safe can be a challenge, Smith says. In many scenarios, the best strategy is to incorporate alerts that give the physician the ability to override the decision support and use his or her own judgment. That is not prudent in some extreme cases in which conflicting orders or drug interactions are always wrong, but relying on the physician's expertise is key to making decision support work, Smith says.

"We always try to remember that we're supporting the physician and providing information so he or she can make the best decision," Smith says. "We found out early on in the process that doctors don't want a computer barking orders at them and telling them how to take care of their patients. They want information and polite reminders, but they don't want a computer taking over their job."

Greg Freeman is a contributing writer for HealthLeaders Media.


This article appears in the May 2012 issue of HealthLeaders magazine.


Reprint HLR0512-6

 

Pages

Tagged Under:


Get the latest on healthcare leadership in your inbox.