The Cure for Dysfunctional CDS

By Sean Benson, VP & General Manager for POC Advisor,
Wolters Kluwer Clinical Software Solutions
Twitter: @POC_Advisor

Effectively delivering clinical decision support (CDS) interventions to the point of care has proven to be a significant challenge for many hospitals and health systems. These projects often fail to drive meaningful improvements and, in some cases, make matters worse. One study found that clinicians ignore EHR safety notifications between 49 percent and 96 percent of the time, primarily due to alert fatigue. Another found a strong association between low perceived value of alerts, clinician dissatisfaction and, subsequently, clinician burnout and turnover.

To succeed in utilizing CDS to improve outcomes, a healthcare organization must apply the CDS Five Rights model as outlined by the Agency for Healthcare Research and Quality, which calls for delivering:

  1. The right information,
  2. To the right person,
  3. In the right intervention format,
  4. Through the right channel,
  5. At the right time in the workflow.

As with many things in healthcare, though, meeting these standards is easier said than done.

Many healthcare organizations attempt to build CDS solutions internally, but this approach can present significant challenges. To be successful in delivering CDS, organizations must have resources and expertise in a number of disciplines, including mapping of clinical information to standardized terminologies, use of Natural Language Processing (NLP) to codify free-text information and the development of highly accurate CDS alerts and interventions.

Most health IT platforms use different terminologies and even have multiple ways to refer to the same medical concept (ex. hemoglobin A1c could also be called HbA1c, A1C, Hb1c, Hb1c or HGBA1C). Additionally, much of the important clinical information is contained in physician notes that are free-text documents. This lack of standardization requires a significant effort to identify, codify and map clinical concepts appropriately to make them usable for delivering CDS to clinicians. To address this successfully, organizations need a team of expert developers, informaticists and clinicians with deep knowledge of mapping, interoperability and NLP, along with the appropriate toolsets to support their efforts.

Even when interoperability and mapping issues are resolved, there is no guarantee that home-grown CDS will achieve long-term success. To sustain improved outcomes, CDS solutions must be able to deliver highly accurate advice into the clinician workflow. This is a significant challenge, as most CDS systems struggle with issues around alert fatigue. For example, in the case of CDS alerting for sepsis, a review of all published studies before 2016 determined that no solutions improved mortality or length of stay. This is primarily due to poor alert accuracy. Essentially, the systems caused alert fatigue and were ignored by clinicians. The effort required to build highly accurate alerting is very significant and beyond the reach of most healthcare organizations.

Fortunately, some health IT vendors have the ability and expertise needed to develop CDS platforms that overcome the deficiencies of home-grown CDS. When exploring vendor-provided options for CDS solutions, hospitals and health systems must consider whether or not the platform will fit into clinical workflows without causing excessive false alerts, and whether it offers the configurable features required for it to evolve over time. They should also seek out a vendor that understands the provider organization’s clinical processes and protocols and focuses on adjusting the technology rather than expecting clinicians to adjust their workflows.

By working closely in partnership during the implementation process, hospitals and vendors can improve care quality and outcomes while preventing disruption to the clinical workflow. Doing so will ultimately improve clinician satisfaction with CDS, encouraging broader adoption of these important solutions and leading to better outcomes.