How Community Health Network is using analytics to streamline its clinical decision support

Clinical decision support can be a complicated issue. It is one Community Health Network has been managing for several years. The health system has more than 200 sites of care around Indianapolis and Central Indiana, including nine specialty and acute care hospitals where care is delivered by hundreds of physicians and other clinicians.


Like all care providers, these physicians and clinicians are extremely busy, and there has been considerable evidence that responding to an overabundance of CDS alerts generated by electronic health records has contributed to their burnout over the last few years.

Therefore, one of the issues healthcare organizations must be cognizant of when developing or implementing CDS is the impact on the end user, said Dr. Patrick McGill, executive vice president and chief analytics officer at Community Health Network.

“We also must determine what action we want them to take when they receive the alert,” he said. “In short, the main problem that we are always trying to solve is ensuring we are presenting the right alert, at the right time, in the right place in the workflow, for the right person, in the right situation.

“Understanding the role of the end user, whether it is a physician, nurse, pharmacist or other clinician, also is important,” he continued. “We ask ourselves: How is the person in that role interacting with that alert? Is it relevant? How often are they acting on it or are they usually closing it out?”

When the health system began exploration of CDS several years ago, it calculated all of its clinicians received a combined total of more than 30 million alerts per year. Through its own analysis, it turned off six alerts determined to be a nuisance to most clinical staff.

Just turning off those six alerts eliminated more than 12 million alerts per year. The remaining 18 million are still far too many, McGill noted, which is why the health system has continued to study this and why it is expanding its relationship with CDS vendor FDB by implementing the vendor’s CDS Analytics module.


Community Health Network’s informatics team, which meets monthly and includes physicians, nurses, pharmacists and informaticists, has been analyzing CDS as one of its target areas for quality improvement.

While CDS is only one aspect of the team’s charter, its scope regarding decision support is quite broad: Some informaticists are focused on order sets, others are focused on medication alerts, some are focused on inpatient alerts, and some are focused on ambulatory alerts. 

“Healthcare organizations considering implementing analytics technology for CDS evaluation should have a data governance structure in place first.”

Dr. Patrick McGill, Community Health Network

Advancing CDS quality improvement in this way has required having a data governance structure so policies and a process are in place for adjusting or turning off alerts.

“About two years ago, before the onset of the COVID-19 pandemic, this team began exploring ways to expand our relationship with FDB,” McGill recalled. “FDB told us about the new CDS Analytics module, which sounded ideal for us at this point in our evolution as the clinicians that led our data governance team are not report writers.

“Manually generating CDS reporting can be quite time-consuming and may not always deliver the insight required or present data visually in a way that can be easily interpreted,” he added.

The team needed a tool that would help them immediately visualize how CDS is being used across the health system, across departments and by individual clinicians. Ideally, this tool would drill down into how providers are interacting with or dismissing alerts.

“The other tools we had at the time did not give us the complete picture like FDB’s CDS Analytics module does,” McGill explained. “This capability will drastically improve how we analyze our CDS performance. Not only will the tool let us know which alerts are working and not working, but it can also give us projections about how planned CDS changes might impact care or safety.

“This insight, combined with our data governance policies and procedures, will enable us to continue our CDS optimization and eliminate more unnecessary alerts,” he added.


Community Health Network is in the final stages of integrating FDB CDS Analytics with its EHR and other systems. It has several goals it wants to achieve with the analytics.

First and foremost, it wants to improve its provider experience with the EHR and optimize provider workflows. The goal is to ensure providers receive relevant decision support alerts at the right point in their workflows, but that they are not being over-alerted.

“The second goal is to drive the best outcomes for our patients through CDS,” McGill said. “For example, we want the alerts to encourage appropriate care at the optimal time, close care gaps, and help clinicians follow the best practice guidelines.

“The easiest way to sum this up is we want to make it easy for the clinicians to do the right thing and make it harder for them to do the wrong thing,” he continued. “That means, for example, we want one click to order all the health maintenance items. Conversely, if they are going to order something inappropriate, the CDS system should make it harder for them to do that, which is the universal tenet of clinical decision support.”

The third goal is to ensure appropriateness of care while reducing wasteful tests and services and driving down care costs. Community Health Network is involved in many value-based care payment programs with both Medicare and commercial payers, so it is always examining ways to deliver the best care more efficiently and cost-effectively.

“Accomplishing these three goals requires data illustrating how clinical decision support is performing and how providers are using it,” McGill said. “FDB CDS Analytics will fill in much of this information in a way that is faster and easier than what we can achieve on our own.

“Accomplishing our goals for optimizing CDS will require provider adoption, and in some cases, behavior change,” he added. “The most relevant, actionable clinical decision support delivered at the right time to the right provider will not move the needle on quality and costs if providers do not act on these alerts.”

The analytics tool will help the health system significantly with education and action, he said. The health system will be able to approach providers with data supporting alerts they may be overriding and reasons why the team is recommending the providers take the action, he said.

“Often, when we talk to clinicians, we discover they simply do not realize that a particular guideline has changed; they are following protocols they have always followed,” McGill noted. “When we show them the evidence for a new CDS guideline, we can see a learning effect in their reaction and the behavior change that follows. But again, creating this learning effect requires supporting data, which is where CDS analytics comes in.”

In a few clicks, the health system is going to be able to pull information to present to providers so that education and learning can occur. Being able to present this data is satisfying and aligns with the health system’s values, he said.

“We believe adoption and behavior change occur best when education is delivered from a place of learning, a place of continuous improvement and collaboration, and mutual understanding,” he added.


“Healthcare organizations considering implementing analytics technology for CDS evaluation should have a data governance structure in place first,” McGill advised. “While this structure does not have to be as evolved as the Community Health Network’s structure that we have been developing for several years, healthcare organizations should have at least a preliminary foundation in place.

“This structure should help all stakeholders understand and agree to the goals of the CDS evaluation process,” he continued. “It also should establish guidelines about how data will be used and analyzed, and how decisions regarding results will be made.”

Another important element to consider is stakeholder involvement, he suggested.

“Healthcare organizations do not want CDS analysis to merely be a theoretical exercise or something that only involves senior leadership,” he said. “It requires engagement from frontline clinicians who receive and interact with the bulk of these CDS alerts. I would bet if you asked any one of those clinicians to name a CDS alert that annoys them, they could easily cite several and that these alerts might all be different depending on that clinician’s role.”

Some of the most bothersome alerts may not even fire that often, but they still are burdensome because they are delivered at the wrong time or are not clinically relevant for specific patients, McGill explained.

“For example, when we were doing our own analysis years ago, providers would always tell us about a deep vein thrombosis (DVT) prevention alert that was very irritating to them and interruptive to their workflow,” he recalled. “The alert did not fire that often, but it was causing plenty of misery for some of our providers, so they were asking us for help.”

Fortunately, the health system was able to adjust it so it would be more relevant.

“That is an example of why you want to get the appropriate stakeholders involved so you can learn about how the CDS is helping or interfering with their workflows and their care practices,” he concluded. “Understanding these dynamics is key to optimizing CDS alerts. Therefore, listening to providers’ input across roles is vital for any CDS quality improvement initiative.”

Twitter: @SiwickiHealthIT
Email the writer:
Healthcare IT News is a HIMSS Media publication.

Click here for More Health News

Share this:

Leave a Reply

Your email address will not be published.