Clinical Decision Support Roundtable: Benefits, Battles & Bettering for the Future

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - clinical decision support

Clinical decision support (CDS) can produce impressive results in improving clinician workflow, reducing utilization and allowing easier access to evidence-based care for better patient outcomes. However, as with any endeavor that impacts a widespread enterprise, the effort is rife with challenges. Even years after implementation, facilities face obstacles such as optimizing alerts and managing liability concerns. CMIO hosted a CDS roundtable with physician leaders from facilities well along the CDS spectrum. They discussed implementation, ongoing challenges and their views on the future of CDS. 

Roundtable Participants
Donald L. Levick, MD, Editor-at-large, CMIO Magazine
 - Donald Levick-Portrait
Donald L. Levick, MD, MBA, CMIO Lehigh Valley Health Network, Allentown, Pa. Co-author of Improving Outcomes with Clinical Decision Support: An Implementer’s Guide, Second Edition; live on computerized physician order entry (CPOE) and clinical decision support (CDS) since 2001
Kenneth Ong, MD, MPH, CMIO, New York Hospital Queens, New York City
 - Kenneth Ong, MD, MPH
Kenneth Ong, MD, MPH, CMIO, New York Hospital Queens, New York City, 2007 Spirit of HIMSS Award recipient, editor of the 2007 HIMSS Book of the Year, Medical Informatics: An Executive Primer, Edition 2.0; live on CPOE and CDS since 2006
Natalie Pageler, MD, Medical Director of CDS, Lucile Packard Children’s Hospital, Palo Alto, Calif.
 - Natalie Pageler, MD
Natalie Pageler, MD, Medical Director of CDS,
Lucile Packard Children’s Hospital, Palo Alto, Calif.;
live with CPOE and CDS since 2007

How was your CDS implementation experience?

LEVICK: When we first went live with computerized physician order entry (CPOE), U.S. penetration was pretty low and it was new to most people. One thing we did well was implement CDS in a very slow, deliberate manner. We started with the least intrusive interventions and then over time became more intrusive—a continuum of intrusion. Once the physicians start to acclimate to CPOE, see the alerts and work with order sets, you can get more aggressive with how much you steer them. As a result, we never endured huge push-back on the number of alerts or alert fatigue, which are seen in other hospitals.

ONG: New York Hospital Queens went live with CPOE and CDS long before I came. The systems were put in place with guidance from the Clinical Informatics Committee, which was comprised of chairs of the clinical departments. I am told by physician leadership that they wish they had a CMIO at that time to help shepherd the process because there were dozens, if not hundreds, of requests. Prior to the formation of the CDS Committee, which I chair, there was no real process in place to review and prioritize requests.  

PAGELER: We took the slow route. We had the pharmacy IS live before we went live with CPOE, so we could track the alerts from the pharmacy. We made some very big decisions before we went live, including the possibility of turning off drug-drug interactions for the provider. The pharmacy already was getting overloaded, we could see how many alerts there were and that many of the alerts weren’t clinically significant in our pediatric population. Currently, the drug-drug interaction alerts are still live for the pharmacists, but not for the physicians.  

Turning off the drug-drug interaction alerts caused us to garner really low scores on our Leapfrog evaluations, but we decided that it was a good decision for our institution based on our strategy of guiding alerts to reducing harm and our understanding that drug-drug interactions are a rare source of harm in pediatrics. Based on further evaluation, we decided to recreate some level of drug-drug interactions through custom work. We are in the process of evaluating which alerts are the most clinically relevant.

What teams do you have in place to manage your facility’s CDS?

PAGELER: We have several teams devoted to CDS issues. We created all of our electronic order sets on paper before we went live with CPOE and an electronic order set committee reviews order set utilization and requests for new order sets or changes to our existing order sets. A CDS committee evaluates all new rules and alerts, ensures the best practices are updated and monitors for alert fatigue. All of our committees include multidisciplinary representatives such as physicians, nurses, respiratory therapists, pharmacists, patient safety and quality representatives, as well as informaticists.

LEVICK: We started first with an order set committee that still meets weekly. That’s where all requests for order sets are vetted and then undergo a vigorous review and approval process, including nursing informatics and clinical pharmacy with multiple sign-off points. The CDS Committee started as an IT Committee to prioritize the requests for alerts since those had to be built into the system. Over the years, it has enlarged its scope and become a Medical Staff Committee, so all decisions are endorsed up through the Medical Executive Committee and to the board of the entire institution. That committee now deals with the prioritization of requests, helping set direction for order sets in terms of general framework, as well as assessing other types of CDS interventions, such as  evidence-based resources and how to link those into varied systems. We now hold a monthly meeting that practicing clinicians as well as ancillaries and IT people attend.

ONG: I guide the order sets team within IS and the CDS committee, the latter of which works closely with the Medication Management Committee, our Pharmacy and Therapeutics [P&T] Committee. I vet their recommendations and our team in IS implements the approved ideas. The IS team includes not only an analyst devoted to orders and order sets but also a Medical Logic Module [MLM] developer.

How have you engaged physicians with CDS?

ONG: We rely on communications with existing venues in the clinical departments as well as the CDS Committee, P&T Committee, the Chiefs Committee, which meets Monday mornings, and with the Stroke Committee and other related committees. The Chiefs Committee includes leadership of the voluntary medical staff. Adoption is driven by outcome metrics delivered by quality and performance improvement.

We employ hospitalists and have a residency training program. The residents write more than 90 percent of orders and do more than 90 percent of the clinical documentation. There was some resistance among the affiliated voluntary attending physicians but over the years, they have adapted. They now use CPOE, CDS and other EHR components every day.

PAGELER: We engaged the attending staff through a clinical resource management program focused on improving quality and efficiency and chaired by the clinical leadership. The CPOE system is viewed as a tool to support the goal, but the intervention can be any combination of synchronous and asynchronous alerts, order sets and/or education.  

We also have a large trainee population who are very adaptable and complete most of our order entry and documentation. We have a large informatics education component for our trainees and a person within the IS department specifically dedicated to education, which has been instrumental in improving the effectiveness of our CDS. For example, we recently have been evaluating drug-allergy alerts, because they have been documented in the literature as being particularly problematic. Our recent audit showed a 90 percent override rate of drug-allergy alerts. One of the major problems is that staff doesn’t enter allergies correctly—they enter normal known effects of the medications and allergies. Now, we are changing several pieces of the CDS component to help achieve more effective allergy alerting, as well as initiating an education drive.

LEVICK: We went from voluntary to mandatory use of CPOE in 2005 when probably 60 percent of our orders were still being entered by community-based independent physicians. It was a cultural challenge to get there, but refusal to use the system would lead to suspension from the medical staff. We do have a challenge with utilization of diagnosis-specific order sets. Thus, to increase utilization, we have been revamping our order sets based on user input.

How are you handling the legal implications of alerts?

PAGELER: This is a really important topic for CDS but, as of yet, unexplored. There still is no case law. We have made a conscious decision that our record of alerts and utilization of alerts not be part of the medical record and, therefore, not discoverable. We have taken a minimalist approach and are erring on having fewer alerts to minimize alert fatigue, while simultaneously hoping that it doesn’t become a liability issue. This is an area for which the broader health informatics community needs to create a clearer policy as we move forward.

LEVICK: We have had people concerned about legal issues of documentation in the system, the availability of data and the responsibility to look at it once they know it’s available than we do with CDS.

ONG: Alert fatigue was a long-standing issue in the hospital, and the P&T Committee spent one year cutting back on a lot of the alerts. The vast majority of alerts eliminated were the drug-food interaction alerts. The P&T Committee determined that the vast majority of those alerts were not contributory to improving patient care. When they eliminated those alerts, about 40 percent of alerts disappeared.

It would really be helpful if we had a nationwide database of evidence-based drug-drug interaction alerts that we could adopt without having to examine each drug-drug interaction, which number in the tens of thousands. That was one of the tasks charged to the ONC-funded Advancing CDS project. Hopefully, in the future, it’s something we can all share. An evidence-based database of drug-drug interaction alerts would help decrease the noise-to-signal ratio that physicians are exposed to. Physicians can then focus on those alerts that really impact patient care.

What is your role now that it has been several years since your facility implemented CDS?

PAGELER: Even though we have taken a thoughtful approach to alert fatigue, it’s still a huge issue. The CDS Committee is charged with constantly reviewing the alerts to figure out which are clinically meaningful and if there are ways to optimize our alerting system.

We are looking at customizing alerts for individual users. There is still a lot of work to do to get the optimal number of alerts which elicit appropriate responses from our clinician providers.

LEVICK: My role is to chair the CDS Committee and report up to the Medical Executive Committee. I also participate in the order set team. Our CPOE team handles screen design and requests for new orders, but anything that impacts CDS gets moved to the CDS Committee for discussion and vetting.

ONG: As chair of the CDS Committee and lead person for CDS within IS, nothing that affects orders or clinical documentation is activated without my vetting. CDS definitely goes way beyond just drug-drug interaction alerts and order sets, embracing other technologies such as clinical documentation. With the meaningful use requirements for Clinical Quality Measures, the value of integrating CDS into clinical documentation becomes more apparent. For example, with venous thromboembolism (VTE), we auto-calculate the risk of VTE from the Caprini VTE risk assessment and also capture contraindications to VTE prophylaxis in clinical documentation.

For facilities that have not yet implemented CDS, what is your top advice for success?

LEVICK: Everyone from the board to the end user needs to understand why you are doing this, the value-add that it brings and the potential costs in terms of resources, money and even the unintended costs that it will place on providers who will deal with extra clicks.

Then, you need a team of people who can make the translation between clinical demands and clinical requests—as to what's technically feasible with the systems. That’s where a CMIO or physician champion acts as the interface and can help lead the charge. It’s important to start with a small win that can quickly show the value without being too intrusive. For example, something as simple as order sets facilitate quick entry and allow you to embed some form of evidence-based practice into the care delivery.

ONG: CDS helps us comply with mandates from the Joint Commission, CMS, meaningful use and others. For example, to meet the American Stroke Association’s Get With The Guidelines program, we converted the NIH Stroke Scale from paper to electronic format. The stroke team appreciated this change, because they simply have to input a few numbers and it auto-calculates the total score. Not only do they have their stroke scale documented and auto-calculated, it’s legibly put into the record.  

PAGELER: Start small and build from there. Implementing too many rules and alerts in the beginning overwhelms the end user, contributes to alert fatigue and can lead to frustration with the CDS system.

It also is important to evaluate the effects of CDS interventions, to understand what works well and what needs to be optimized. As an academic institution, we have had some great collaboration between our researchers and our IS department. For example, one of our big wins was a CDS alert about blood transfusion thresholds. We implemented an alert that referred the physician to a journal article about safely lowering the blood transfusion threshold in pediatrics. We found that this intervention decreased our blood transfusion threshold and decreased blood transfusions by more than 400 per year.  

How do you see CDS changing in the future?

ONG: It's going to expand beyond the walls of the hospital to include ambulatory and long-term care. Instead of looking at CDS just for the hospital, we will be talking about CDS across the continuum of care.  We also are extending our audience beyond healthcare providers. Some of the most exciting innovations now and in the future will engage patients and their families, e.g., CDS on patient portals and smart phones.

PAGELER: I am hoping to see ways to share CDS across systems. Our CMIO, Chris Longhurst, [MD,] helped develop BiliTool which is a web-based module. Our EMR actually sends the neonate bilirubin level, weight and risk factors to this tool, which responds with suggestions for the appropriate therapy. Tools like this have the potential to reduce some of the work at the individual hospital level, so everyone is not recreating the same CDS. From the national policy level, these are some of the issues we are going to have to assess: How can we make CDS easier to implement at a wide range of hospitals? How can we ensure that CDS tools are kept up to-date?

LEVICK: To me, the next phase is integrating more tools directly into the core systems, so a provider doesn’t have to go out of the system and lose context. To have more of those tools, even third-party external tools, integrated into the core systems, is where we need to head to really make CDS more useful.