FREE Health Data Management e-newsletter 
Decision Support Not an Exact Science
Health Data Management Magazine, November 1, 2007
For example, some, like Mercy Health Partners, a Toledo, Ohio-based integrated delivery system, want to ensure the decision support they offer reflects their best practices. So they develop an extensive evaluation process for each clinical intelligence tool. Others are concerned that too many alerts will disrupt physician workflow. Because of such challenges, some provider organizations spend a vast amount of time assessing each clinical intelligence tool before offering it in a CPOE system. "Some larger provider organizations do considerable decision support evaluation both before and after rollout," says Jonathan Teich, M.D., assistant professor of medicine at Harvard University, Cambridge, Mass., and an attending physician in emergency medicine at Brigham & Women's Hospital, Boston. "The results can be surprising for organizations to learn which types of interventions work well for their clinical staff." Offering decision support from a CPOE system has been an imperative, yet complex process for Mercy Health Partners. Nine months before CPOE rollout began, executives at the delivery system invited physicians to participate in a paid committee to determine the type and frequency of intelligence tools to be used in the new system. The physician design committee selected various drug interaction alerts and created about two dozen rules for the system that direct physicians to some of Mercy's clinical best practices. But Mercy had considerably more difficulty creating automated order sets-another type of decision support tool. Order sets are groups of drugs and tests that physicians commonly order simultaneously or that reflect other clinical best practices. To develop automated order sets, the committee first had to collect paper-based order sets from physicians and nurses across Mercy's seven hospitals. Committee members spent months going from unit to unit looking for such order sets, even searching through desk drawers for some unofficial ones that were written on scraps of paper, says Kenneth Bertka, M.D., chief medical information officer. Then they had to consolidate the order sets and present them to the pharmacy and therapeutic committee at each hospital for approval to be used in the system. Incorporating more than 700 order sets into the CPOE system was a time-consuming project, Bertka says. "We started by implementing standard order sets, but that was more of a challenge than we thought," he says. "In some cases we went live with different order sets at different hospitals because we couldn't make the changes overnight. There was a need to find a faster way to develop order sets." Shortly after the CPOE roll out began, Mercy changed its clinical governance structure to enable the physician design committee to implement order sets without waiting for other committees' approval. Physician design committee members now talk to various clinicians about prospective changes, then invite other committees to comment after they have been implemented. Though Mercy has shortened the process for integrating decision support into its CPOE, the committee still hasn't unveiled all of the alerts, rules and order sets it plans for the system, Bertka says. "In an ideal world, we would have had the clinical guideline development process streamlined across our hospitals before we implemented order entry," he says. "We didn't have that. But we also made a conscious effort to go forward slowly so our clinicians wouldn't be overwhelmed by alerts." Critical Element While CPOE has been touted as the key to most hospitals' patient safety and efficiency initiatives, combining it with decision support is essential to meeting such goals, many industry executives say. Most CPOE systems include or can be integrated with third-party optional applications to give physicians alerts for drug interactions and duplicate tests at the point of order. They also can include dosing calculators, applications for creation of automated order sets or best practice rules, and links to interactive clinical reference information. Additionally, many CPOE systems offer reporting capabilities to enable hospitals to track how physicians use decision support. But it's the integration of all these tools that makes CPOE such a powerful patient safety system, says Sharon Young, senior manager at Healthia Consulting. The Minneapolis-based firm in September was purchased by Ingenix. "CPOE just gets you to the door," Young says. "It's the decision support and reporting that allows hospitals to measure how effective they've been." Mercy Health Partners' physician design committee still is evaluating decision support for the delivery system's three hospitals that aren't yet live with its CPOE, from Siemens Medical Solutions, Malvern, Pa. But Mercy's executives want to take the next step with the intelligence tools at the four hospitals that are, says Bertka, the CMIO. "The challenge going forward is how we will review order sets," he says. "We hope to do this on a regional approach." Mercy plans to integrate decision support software from Zynx Health, Los Angeles, with its CPOE system to aid in such analysis. Executives will use the vendor's technology as a template to create order sets so they can centrally track and archive them. They also will use it to receive updates on evidence-based clinical guidelines they can use when evaluating order sets, Bertka says. Having begun a CPOE system rollout only five months ago, Decatur (Ill.) Memorial Hospital isn't quite ready to begin an extensive technical analysis of its decision support. But because the 356-bed hospital automated all order sets before rollout, its clinical intelligence tools already have received enough scrutiny from physicians to warrant numerous changes, says Michael Zia, M.D., vice president of medical affairs and quality. The hospital also has implemented other decision support tools, such as interaction alerts and links to medical articles and journals, accessible through the CPOE, from San Francisco-based McKesson Corp. Big Bang After Decatur Memorial executives began talking to clinicians about plans for CPOE three years ago, they decided to take a "big bang" strategy with decision support and unveil multiple clinical intelligence tools at go live rather than waiting until physicians were more accustomed to the system. They believed the approach would help physicians better adhere to hospital best practices, enter orders faster and be a motivator for system use, Zia says. Clinician leaders and I.T. staff evaluated various alerts and order sets, then integrated them into the system. They planned to roll out the technology by department over six months. But as soon as the system was unveiled in the first department, some physicians started using it enterprisewide. And many of the early adopters began submitting suggestions on how to improve decision support. "We couldn't say 'don't order things in this area' because it was good that they were using the system," Zia says. "But we were getting requests for more order sets or updates to care plans. It was a boon to patient care, but it gave us a lot more things to do." Decatur Memorial executives decided to build on the early CPOE momentum by promising physicians they would try to implement their suggestions in the system within 48 hours. But the undertaking stretched resources so much that the hospital had to delay its implementation schedule. The hospital also delayed planned mandates for enterprisewide physician use of CPOE and the termination of all paper orders, which has caused headaches for some clinicians and department secretaries, Zia says. The current dual use environment at Decatur Memorial is "the worst of all times," Zia says, but it's actually creating a motivation for the I.T. department. "We are overwhelmed with trying to keep up with input physicians have given us," he says. "But if they have great suggestions, we'd rather slow down implementation and have physicians be positive on their experience. It's that response that will make them put their time and effort into it. It's a trade off, but we have a commitment to them." Once Decatur Memorial finishes rolling out CPOE, it plans to conduct a formal analysis of all order sets and alerts. Executives also plan to create more sophisticated best practice alerts in the system. "We hope to eventually have a database we can share with other McKesson CPOE users so we can work together to get best care algorithms," Zia says. "We can't mandate physicians order one test over another, but we can give them as much information at the point of care as they need." Analyze this All hospitals should use automated analysis to determine how physicians are using decision support in their CPOE, says Young, at Healthia. For example, they can measure how frequently an order set is used or how many times an interaction alert is overridden to better understand if such applications are helping meet clinical goals. Hospitals also should collect benchmarks before implementation to get a more accurate picture of use data, Young says. Additionally, physicians should be involved in the analysis process so they can help determine which decision support applications are important to their organization's goals and understand that their use of them will be measured, she adds. "Every facility and physician has to decide what they will do with this analysis," Young says. "Will they resist it and keep measurements low and in-house, or will they make the most of it to improve patient care?" In the seven years it's been using CPOE, Alamance Regional Medical Center has made numerous changes to its integrated decision support as the result of analyzing physician use. The Burlington, N.C.-based, 238-bed hospital initially offered only interaction alerts and automated order sets in the system, from Eclipsys Corp., Boca Raton, Fla. But those applications have been evaluated ever since via analysis reports generated by the system, says Ken Fath, chief medical information officer. "We continue to monitor the firings of alerts and how they are responded to," he says. "If alerts over time aren't changing the behavior of ordering, then we will run a batch by our physician review board for evaluation. There's a fine line of how many alerts we want to fire at physicians." Alamance also is consolidating the personal order sets in its CPOE system. The hospital originally had enabled physicians to create order sets that no one else could use to increase adoption. While the strategy spurred about 400 personal order sets, I.T. staff has had difficulty managing decision support they couldn't access, Fath explains. For example, if a drug name changes, I.T. staff has to contact the physician and persuade them to make the modification in their order set. "We tried to simulate the paper world as much as we could when we started CPOE because it wasn't a popular project," he says. "But if we had to do it again, we wouldn't allow personal order sets." Alamance also continues to add new types of decision support to its CPOE system. In 2002, the hospital began using tools within the application to build various advanced alerts based on its clinical best practices that are designed not only to improve patient safety, but also workflow efficiency. For example, if physicians overrule an alert, the system presents another screen directing them to choose from a list of acceptable explanations. This helps the hospital immediately document why a best practice wasn't followed instead of tracking down the physician after the fact, Fath explains. Alamance also plans to integrate established clinical ordering guidelines from Zynx and other organizations into the system, he adds. While physicians have suggested many of the alerts Alamance has integrated in its CPOE, it's been a cultural change for them to accept being monitored on how they use them, Fath says. "No one had really looked at whether they were using best practices before," he says. "Some of the resistance has been that they don't like that I can look at what they are doing." Alamance, however, plans to continue monitoring decision support use, and might even break down the analysis by physician, Fath adds. Although integrating decision support in a CPOE system isn't new for Children's Hospital of Pittsburgh, analyzing how physicians use it is. The 296-bed hospital implemented the system, from Cerner Corp., Kansas City, Mo., in October 2002, but only recently acquired the capability to monitor decision support use, says Jim Levin, M.D., chief medical information officer. While I.T. staff still is learning about the analysis tools, they've spent the past five years developing and customizing alerts and order sets for the hospital's pediatric specialty. Basic decision support applications, such as drug interaction alerts and dosing calculators, have been adjusted for the smaller stature of Children's Hospital patients. The hospital also created alerts in the system to lead physicians to pediatric best practices recommended by its physician advisory committee, which comprises clinicians and pharmacists. For example, one alert requires a patient's weight and allergies be entered before an order can be processed. I.T. staff also enabled the CPOE software to pull data from Children's Hospital's electronic health records system, also from Cerner, to help physicians better analyze how patients are responding to drugs, as well as how they measure on pediatric growth curves. Additionally, the hospital offers access to clinical reference content from Lexi-Comp Inc., Hudson, Ohio, and its own pharmacists to aid physician ordering. "The whole art is to tailor use of the appropriate tool for the situation," Levin says. "While each alert sounds terrific in isolation, when you look at a workday, physicians can be receiving an overwhelming amount of alerts not because they are making errors, but because the alerts are overly conservative." Although Children's Hospital has yet to terminate alerts based on use analysis, its physicians advisory committee has been evaluating the impact of some decision support since before go live. For example, it decided to use just the highest severity level of alerts available in the system when it was implemented. Further, it's evaluating suggestions from physicians about eliminating some alerts that are irrelevant to pediatrics, such as ones that warn physicians not to prescribe a particular medication if the patient is drinking alcohol.
For more information on related topics, visit the following channels:
