The concept of clinical decision support is simple. It delivers context-specific education material in real time to a physician trying to make a patient-care decision. Tools can be divided into several basic types. Some help clinicians select the appropriate imaging study, and others help physicians make the appropriate clinical decision. But implementing decision support in clinical practice? That’s a bit more complicated. But like many things that truly matter, where there’s a will, there’s a way.

The need for clinical decision support is clear. “It’s very hard for physicians ordering imaging tests to keep [all of the required] information in their head,” points out Katherine Gray, president of Sage Health Management Systems, Inc. For example, the Ottawa Ankle Rule, which determines ankle sprain patients who require an x-ray, is too specific and detailed for non-specialists to remember. While clinicians deal with thousands of complex rules and guidelines, radiologists contend with other challenges. The number of cases and images continues to skyrocket, and studies are growing in complexity as well. What’s more, the base of clinical knowledge is not a constant. Guidelines change as researchers uncover new findings and publish new studies and physician associations release new guidelines, but it can take five years or longer to incorporate new clinical guidelines into practice.

Medical errors, waste and the increasing adoption of imaging utilization management programs by payors also are driving the market. Nevertheless, current decision support adoption rates hover in the single-digit range, says long-time advocate Ramin Khorasani, MD, MPH, who is director, Center for Evidence-Based Imaging at Brigham and Women’s Hospital in Boston.

The challenge, says Khorasani, is three-fold. First, the tools that enable delivery of decision support — computerized physician order entry (CPOE) and structured reporting — must be more widely adopted. The next challenge is knowledge delivery; developers must engineer multiple programs to deliver content through CPOE or the structured report. Finally, the system must incorporate knowledge management, which comprises mechanisms to track and update content.

Despite the hurdles, decision support is being implemented at facilities across the country. Some sites opt for solutions developed by vendors, while others use CPOE or the electronic medical record (EMR) as a foundation and develop their own solution. This month, Health Imaging & IT visits a handful of these sites to learn about decision-support options and the benefits of the technology.

Decision support & imaging utilization

This graph shows variability in primary care physician use of outpatient radiology at Brigham & Women’s Hospital . The X axis depicts each physician with an identifier and bars of same color denote physicians practicing in the same office location. The Y axis represents the number of imaging tests requested per patient per year.

Imaging utilization is heterogeneous and represents both over- and under-utilization of imaging tests. For example, studies indicate an increasing reliance on appendiceal ultrasound to diagnose acute appendicitis despite its lack of effectiveness. At the same time, mammography is underutilized with only a fraction of the target population receiving appropriate screening exams. (Based on recent government health surveys, the percentage of women 40 and older who reported getting a mammogram in the last one to two years dropped to 74.6 percent in 2005 from 76.4 percent in 2000, according to the Centers for Disease Control.)

Decision support transforms the picture by alerting the physician to patient needs and the most appropriate imaging study. During the electronic order process, a pop-up can inform the ordering physician if the ordered study is inappropriate or contraindicated or if the patient is due for a regular screening exam.

Brigham and Women’s Hospital has taken a leading role in decision support. More than 90 percent of physicians have adopted CPOE, enabling delivery of critical information about imaging studies.

One key to effective decision support is embedding tools in daily clinical workflow. That is, decision support cannot require additional steps; it must integrate both with existing technology and workflow, says Khorasani. In addition, decision support must be actionable and current. It must reinforce or contradict behavior. For example, it might refer to recent studies that indicate MR better detects stroke than CT. If the mechanism provides ambiguous information, its utility is negligible. For example, informing a physician that research data about a certain intervention is varied does not provide clear, actionable guidance. 

“Any decision-support intervention must translate into actionable steps for the computer. The computer understands yes or no scenarios, not shades of gray,” adds Louis Penrod, MD, director of clinical decision support at University of Pittsburgh Medical Center in Pennsylvania. And the decision-support tool must be updated with current data as guidelines change.

The benefits of implementing decision support are both clinical and financial. At the clinical level, the patient is referred for the appropriate study. On the financial side, decision support can help eliminate unnecessary or duplicate studies. In addition, more complete data capture via the structured report could increase revenue, and when mined, further support current clinical practice or indicate a need for a modification. For example, Brigham and Women’s Hospital calculates a net revenue capture opportunity of $4.63 per outpatient CPT billed due to a lack of information like billable diagnosis or referring physician identity.

Decision support for imaging diagnosis

Here’s how it works…

Top: A radiology exam, in this case a chest CT for lung cancer screening, is selected from pre-determined menus. Bottom: Once an exam is selected from pre-determined menus with CPT mapping, the user selects the indication for the test from pre-determined menus, using structured or coded indications mapped to ICD9 codes to ensure successful billing. A user must select at least one entry from the ‘sign and symptoms’ field and at least one entry from the ‘relevant history’ or ‘differential diagnosis’ field to be able to proceed with the order. Coded indications represent most common reasons, appropriate or inappropriate, for the test. Inappropriate indications launch decision support to help the user in test selection. Free text entries account for the balance of reasons a specific test is requested. Screen shots courtesy of Ramin Khorasani, MD, MPH, Brigham and Women’s Hospital, Boston, Mass.

Decision support is a broad topic that also encompasses tools to facilitate accurate diagnosis. For example, Jonathan Gusdorff, MD, a neuroradiologist practicing in suburban Philadelphia relies on Amirsys Inc.’s Statdx. The online decision-support tool replaces textbooks and reference books in challenging cases, says Gusdorff. 

The decision-support model replaces several less effective options. In difficult cases, the radiologist might:

• Complete a lengthy descriptive report that covers everything he or she sees
• Transfer the case to another radiologist
• Order another imaging test
• Refer to textbooks or online search engines

Statdx provides a more accurate and efficient model. Gusdorff uses the program three to four times weekly, accessing Statdx via a desktop shortcut. Take for example a rare case like a suspected aberrant internal carotid artery. The radiologist hits temple bone and then middle ear. The program provides a summary of the condition, relevant images and clinical findings. After about 18 months of use, Gusdorff has whittled the time it takes to use Statdx to support a diagnosis to less than a minute. “It’s replaced walking to the bookshelf and combing through the index searching for the right information. It’s like having 20 experts looking over my shoulder at any given time,” notes Gusdorff.

By helping radiologists provide accurate diagnoses, the system also improves service to surgeons and referring clinicians. Although the initial investment in Statdx was a bit expensive, says Gusdorff, it provides a current, online replacement for reference textbooks. “I haven’t purchased any textbooks since buying Statdx.”

Decision support & chronic disease management

Geisinger Medical Center in Danville, Penn., comprises three hospitals and dozens of outpatient sites and practices throughout Pennsylvania. The health system has relied on Epic Systems Epicare electronic health record (EHR) in its ambulatory facilities since 1997 and plans to complete its inpatient implementation this year.

“The EHR serves as the delivery mechanism for various decision-support efforts,” explains Chief Medical Officer Joseph Bisordi, MD. When the health system deployed Epicare, physicians identified nearly two dozen types of patient visits and reengineered the workflow to accommodate decision support. The target areas include acute chronic diseases such as diabetes and at-risk patient populations like the elderly. “There are 20 to 40 items that may make a difference in patient outcomes. We don’t want to miss any of them,” states Bisordi.

The robust decision support undertaking at Geisinger includes notifications for: drug-drug interactions, drug allergies, healthcare reminders and alerts and automatic orders. “The goal is to make it easier for the physician to do the right thing than the wrong,” sums Bisordi.

A rules-engine runs in the system background and delivers pop-ups related to the patient based on demographics. Take for example, the case of diabetic patients who should be checked for multiple measures including hemoglobin A1c, blood pressure, cholesterol and more during each outpatient appointment. “Compliance with all measures is relatively low at most sites,” notes Bisordi. During a clinical decision support-enabled visit, the physician is reminded to order all tests and can complete the orders with one click. As a result, the medical center has seen “significant movement” toward the numbers of diabetic patient meeting all guidelines.

Similarly, Salt Lake City, Utah-based Intermountain Healthcare aims to develop advances to its GE Healthcare Centricity EMR platform to better manage chronic diseases in adults. The healthcare system has developed a new data-driven model for diabetic patient care. The system tracks each physician’s diabetic patients and analyzes critical parameters such as hemoglobin A1c, blood pressure and lipid testing rates and results. The Intermountain Healthcare primary-care team establishes goals for the population such as 90 percent compliance with hemoglobin A1c testing. After physicians meet goals in one parameter, the team addresses another. Intermountain Healthcare also provides physicians with report cards, so that they can compare their patient results to other physicians in the system. In addition, the healthcare system relies on the parameters for a modest pay-for-performance system that rewards providers with 1 percent of their compensation when their group of patients meets goals. 

The decision support-enabled physician visit

Decision support also transforms the medical visit into a data-driven operation. “Practices with a paper chart can access only a small fraction of the pertinent patient information,” explains Steven Towner, MD, clinical programs, adult primary care implementation leader at Intermountain Healthcare. Intermountain Healthcare developed a Chart Notes tool to code patient data and facilitate consistent care standards. The tool also proactively searches for data a physician may need during an office visit. Target parameters include lab values and results from other providers. Currently, the system is configured to handle diabetic office visits, but Towner hopes to expand it to other conditions.

At Geisinger Medical Center, the Epicare system facilitates the templated visit, which facilitates physicians’ ability to a complete all required tasks during a patient visit. A kidney transplant patient visit, for example, requires close to 50 questions and tasks. “It’s hard to do consistently,” says Bisordi. The template reminds the physician and provides a structure for consistent documentation.

At Geisinger, Epicare also issues reminders about payors’ formulary requirements, alerting physicians when an insurer will not cover a specific drug. The formulary feature alone translates into $1,000 in annual savings for each of the system’s 650 physicians for a total annual savings of $650,000.

Integration, interoperability & improvement

At Geisinger, Epicare is integrated with a host of legacy systems including PACS, RIS and LIS. Another enabling technology is the Enterprise Master Patient Index (EMPI). University of Pittsburgh Medical Center (UPMC) plans to deploy Initiate Systems Inc. Initiate Identity Hub software to help facilitate interoperability across disparate systems and to provide accurate interactive delivery of patient information. The software will help connect patients to their medical information across the entire health system, which, in turn, may help facilitate the medical center’s robust decision support undertaking.

UPMC has developed multiple decision-support tools. “Hospitals with CPOE have implemented interactive decision support with order entry. Those without CPOE use alternative methods of messaging,” says Penrod. CPOE provides different tools, he says, but isn’t necessary for all flavors of decision support.

Take, for example, patients receiving heparin therapy. UPMC aims to more accurately identify heparin-induced thrombocytopenia (HIT). The internally-engineered UPMC system checks absolute platelet values and measures counts over time. If the count falls below a cutoff value or drops more than 50 percent in one week, the system sends the attending and primary-care physicians an email warning that the patient may have HIT and suggests a treatment approach. This type of mechanism can be implemented with CPOE or without CPOE, says Penrod. UMPC sites without CPOE rely on a communications call center that houses a database of providers and handles messaging according to each provider’s preference: email or fax. On the other hand, decision-support mechanisms such as medication allergies and contraindications require CPOE as the physician needs to receive the message at the point of care.

The foundation of the in-house system is a structured database that tracks the various decision-support rules. “Decision support requires a substantial commitment to infrastructure, quality assurance and maintenance,” explains Penrod. UPMC uses both regular review cycle and content experts to ensure content reflects current standards of care. The system also tracks each rule with regard to volume and analyzes whether or not each rule is functioning as expected. 

The workflow issue

To be truly effective, decision support must integrate smoothly into physician workflow. That’s the really tough part. At UPMC, Penrod and his team invest a significant amount of time discussing workflow with physicians. “We need to understand how to fit decision support into workflow. If we develop nuisance alerts that annoy physicians, they will begin to ignore them,” says Penrod. “There are right and wrong times [and ways] to deliver decision support,” adds Bisordi. Some reminders should be passive; others like drug interactions should be active pop-ups. In addition, developers need to respect the limits of the 15-minute office visit. “There’s only so much that can be crammed into the visit,” says Bisordi.

UPMC tests each decision-support mechanism in a pre-production environment that helps the team better understand the impact of the rule. The pre-production environment doesn’t fully capture all of the complexities of the production environment, but it does help the team refine and revise rules.

In addition to selecting the right time, place and format for decision support, Bisordi recommends sites spend time preparing physicians and other users. The best way to garner acceptance for decision support is to approach it as a quality measure, says Gray. “Decision support helps physicians do the right thing for their patients in the most efficient way.”

Future directions

Decision support is a journey, not a destination. After early adopters tackle the low-hanging fruit, many plan to move into other areas. Geisinger Medical Center, for example, has other chronic conditions in its sights. Next targets include hypertension, depression screening and autism evaluations. The challenge as the number of rules grow, says Penrod, is managing the increasing detail. Content experts need to stay on top of changes in standards, and multiple systems may need to be tweaked with each update. For example, a change in a normal lab range may require an adjustment in the LIS and decision-support system.

Intermountain Healthcare envisions the “smart visit,” where the computer automatically orders necessary tests such as lipid panels. The current process requires multiple manual processes including physically documenting, sending and coding the order. These processes are prone to error and impede efficiency. The smart visit can reduce paperwork, minimize keyboard entry and standardize processes, says Towner.

The enterprise view

Various types of clinical decision-support solutions are available. Enterprises can develop or deploy systems to standardize and improve patient care, reduce waste and trim costs. Target areas include radiology where decision support can steer referring physicians to the appropriate test or help radiologists better diagnose complex cases. But decision support can be an enterprise tool; numerous healthcare systems are tapping into decision support to better manage patients with chronic diseases. Regardless of the target or application, however, effective solutions integrate with existing technology, require significant oversight and regular updates and work with, rather than against, physician workflow.

On the Decision Support Drawing Board

As early adopters implement decision-support tools, vendors are stepping to the plate and investing in new solutions. Recently released and next-generation decision-support tools could re-invent the decision-support arena.   Sage Health Management Solutions Inc. offers Radwise, a computerized physician order entry (CPOE) system that incorporates evidence-based decision support. The system recommends appropriate studies according to current guidelines. The goal is to reduce the number of inappropriate imaging studies. McKesson Corporation wraps decision support into its Horizon Cardiology and Horizon Cardiology Hemodynamics platforms to provide better, safer, more efficient cardiac patient care. The systems gather data at the point of care to help physicians determine appropriate interventions for specific patients. Built-in dashboards help managers evaluate performance and outcomes and identify red and yellow flags. The next step is integration with Horizon Performance Manager, which would open the systems to general medical data and feed that information into the order set. The ultimate goal is to increase adherence to American College of Cardiology (ACC) and American Heart Association (AHA) guidelines. Siemens Medical Solutions’ answer is its Remind know-ledge platform. The system currently provides retrospective decision support and quality measurement, but Siemens wants to focus on point-of-care decision support. The plan, says Bharat Rao, senior director for knowledge development, is to integrate Remind into the Soarian and syngo families of solutions to extract data at the point of care and facilitate clinical decision-making. During a normal day, physicians must apply thousands of patient care guidelines, says Rao. Remind can simplify and streamline the process by helping physicians order appropriate follow up procedures. These developments will arrive “in the near future.” Siemens’ three-to-five-year plan hinges on imaging, proteomics and genomics and deploying decision support to personalize guidelines and adjust treatment based on specific patients. For example, Remind might analyze biomarkers and genetic information to determine which patients might benefit from diagnostic ultrasound to determine cardiac risk factors. Decision support is evolving. Market penetration will increase as more sites adopt the required underlying infrastructure and enhancing quality and containing costs move to center stage. At the same time, vendors are prepping for the future and developing tools to expand decision support throughout the enterprise and into new areas such as molecular imaging.