What Is Clinical Decision Support?

CDS encompasses a wide variety of tools, such as alerts and reminders, clinical practice guidelines, customized order sets, data visualization dashboards or interfaces, documentation templates, diagnostic support, and other reference information—all tailored to clinicians' data, information, and knowledge needs (see Table; AHRQ 2022).

CDS provides clinicians, staff, patients or other individuals with knowledge and person-specific information, intelligently filtered or presented at appropriate times, to enhance health and health care. CDS encompasses a variety of tools to enhance decision-making in the clinical workflow. These tools include computerized alerts, prompts, and reminders to care providers and patients; clinical guidelines; condition-specific order sets; focused patient data reports and summaries; documentation templates; diagnostic support, and contextually relevant reference information, among other tools. — Office of the National Coordinator (ONC)

The idea of CDS tools emerged from the recognition that medicine and healthcare have become data and knowledge intensive. Clinicians as humans have limited cognitive capacity to assimilate or process high volume and high dimensional data, and can require tools that filter, prioritize, and present this information. Take for example that there are a number of protocols that health systems have established for suspected pneumonia, often including a combination of labs, medications, and imaging. This evidence-based knowledge can be represented by a CDS tool in the form of an order set, alleviating cognitive burden and saving time that could otherwise be spent with the patient. CDS allows for more streamlined workflows and helps clinicians adhere to standard clinical processes. Without such tools, clinicians may be overwhelmed with keeping track of patient data and acting on it in a timely manner.

Further, our knowledge of biomedicine is growing exponentially (Fontelo and Liu 2018). No clinician can keep up with the medical literature or guidelines. Therefore tools that assist clinicians to search, retrieve, and integrate relevant knowledge and guidelines are important and have been a driving force for CDS.

CDS Delivery, Design, and Challenges

The near ubiquitous adoption of electronic health records (EHRs) creates potential for CDS tools to be integrated into clinical workflows. In fact, the adoption of CDS has become a primary focus of informatics and a driving use case for the adoption of EHR systems. There are many examples where CDS has been integrated into EHR systems to support improvements in clinical care. For example, CDS has been shown to improve adherence to guidelines (Lu et al 2016; Silveira et al 2016) and to improved patient outcomes (Heekin et al 2018). CDS has also been used in patient-facing applications (Goehringer et al 2018) and through mobile health technologies (Raghu et al 2015). The reach of CDS is ever expanding and proving to be an effective mode of delivering real-world interventions.

While CDS can be very effective, there are also many reported cases where CDS does not show benefit, potentially being detrimental (Roshanov et al 2013; Matui et al 2014). Despite the potential of EHR CDS, it has not been adopted and scaled in the way informatics visionaries had intended. There are several reasons for this, including a lack of a central resource for CDS guidance and problems in interoperability. Even when CDS is shown to be effective at one site, it cannot be “scaled up” or rapidly implemented into other organizations and settings due to the lack of standardization of EHR systems (i.e., function, data) and clinical workflows. However, there are areas that can be controlled during its design.

To start, if clinicians do not trust or find value in the CDS tool, then its guidance will most likely not be followed. Many studies have shown that a large proportion of CDS alerts are ignored or dismissed (Carroll et al 2012), often due to a lack of specificity and executable action. In other words, if CDS tools do not offer an executable recommendation that is relevant to the patient, it is likely that it will be ignored and not have the desired effect. Most clinicians can easily recall a time that a CDS alert has not been helpful, or interrupted their workflow. A reminder to schedule a pap smear for a male patient is not an effective use of an alert. For a less extreme example, imagine the frustration of a consulting surgeon attempting to document a note, and being interrupted to order a flu shot. Other side effects of ineffective CDS include alert fatigue, workarounds, and the dissemination of out-of-date content (Ash et al. 2007). At best this frustrates clinicians, and at worse, it creates unsafe conditions. However, not all CDS are alerts, as for the purposes of research, alerts can be built in many different ways and can target a broad range of clinical stakeholders. The majority of current alerts are not “hard stop” requiring a user to interrupt their workflow and interact with the alert.

Despite the many challenges for implementing CDS that is useful and safe, there is tremendous enthusiasm for the potential of CDS to transform and improve care. In fact, the routine use of automated CDS is a fundamental component of the national healthcare reform strategy endorsed by the Centers for Medicare and Medicaid Services (CMS) and the Office of the National Coordinator (ONC). Other requirements for CDS will most certainly increase over time. Meeting the requirements will be a challenge from both an implementation standpoint and from its ability to be useful in real-world workflows. There are many ways to mitigate the challenges and move toward success in CDS design, which has been, and is, a major focus for clinical informatics.

Next, we describe the 5 rights of CDS framework and the GUIDES checklist to provide a foundation for how to think about CDS design.

5 Rights of CDS

Each potential CDS application must be carefully considered and thoughtfully designed to achieve the desired organizational goals without any unintended effects. This process is partially tied to the 5 Rights of CDS framework, a well-known framework used to plan, assess, and deploy CDS interventions. First developed by Osheroff (Osheroff et al 2007), this framework has been adopted by several developers, implementers, researchers, and organizations such as the Agency for Healthcare Research and Quality (AHRQ).

The 5 rights of CDS are as follows:

The right information, to the right person, in the right format, through the right channel, at the right time in the workflow.

The right information refers to what content is presented to the end-user of the CDS tool. This may seem straightforward, but there are several considerations to keep in mind. First, the information presented should be derived from a reputable source, such as from evidence-based practice, government regulations, or clinical practice guidelines from an accredited agency. It should not be based solely on expert opinion, or without consensus on the recommendations of the CDS tool from the targeted audience (Campbell 2016). If the information is not universally accepted, there may be issues in adherence and acceptance from the end-user, and the tool will not have the intended effect. Often to fine tune the CDS tool, override rates and comments are collected (either by the tool or qualitatively) and analyzed to understand the reasoning of why the alert is causing an unneeded disruption in the clinician’s workflow.

The right person refers to the CDS tool reaching the individual who may take action based on the information given. This can mean a number of different individuals of the care team (and in different combinations), including nurses, physicians, respiratory therapists, physical therapists, pharmacists, patients, and patient caregivers. The information may need to be adjusted depending on the audience, especially in team-based CDS. For instance, there may be a protocol around antibiotic dosing, with different alerts targeting the nurse, physician, and pharmacist. The nurse should not receive dosing instructions, as that is not in their scope of practice. Instead, there may be an alert when scanning medication if there is a worry for overdose. Upon ordering, dosing information may be better suited to the pharmacist or physician. Again, if action cannot be taken by the individual, then they should not receive an alert; this only increases alert fatigue.

The right format refers to how the CDS tool is presented, whether it be an alert, order set, infobutton, clinical calculator, clinical practice guideline, or any other format. The developers and implementers of CDS should consider what problem needs to be solved, and how it may be done in the least interruptive format as possible. Again, if all CDS were interruptive alerts, there would be severe alert fatigue, and the use of these tools overall would become worthless. Format may be chosen due to acuity, or it may depend on what makes the most sense. For instance, if there are issues with clinicians remembering the correct orders to initiate blood transfusion, an order set may be developed so there is no delay in the patient receiving the appropriate blood product. This not only ensures the correct ordering process but reduces cognitive burden on the clinician and is done so in a non-interruptive, transparent, customizable format. Second, tying into the right person, the information should be presented in a way that is usable to the end-user. There should be just enough information to be usable, and not cause cognitive overload. Alert fatigue is a serious concern in CDS systems (Agency for Healthcare Research and Quality 2019), and cognitive overload is a contributive risk factor. Information should also be tailored to the audience, and considerations include what information should be given to a clinician versus a patient.

The right channel refers to how the CDS tool is delivered. This may be through the EHR, a patient portal, another clinical system (such as a separate computerized physician order entry system or radiology service), a smartphone app, or by paper. This is becoming an increasingly important factor to consider with the expansion of digital health, as patients and caregivers are now becoming the primary focus of many decision support tools. The EHR should not always be considered the primary modality of delivering care, and consideration should be made in how to expand the access of decision support. In addition, downtime procedures should be put in place for when access to the EHR (or other system) is unavailable, making paper decision support tools not yet completely obsolete.

Finally, the right time in the workflow refers to the fit of the CDS tool into current clinical processes. This is often considered one of the more difficult parts of the 5 rights of CDS, as informatics solutions such as CDS must be fit into workflows that may have been in place for many years without consideration to new technology. This often results in reworks of workflows, which can lead to resistance and strain on the end-user. In contrast, the deployment of CDS may be ineffective when built around workflow if the information is not delivered in a timely manner, or if the information needed for the CDS tool to function is not yet available. For instance, there may be an order set or alert based on past medical history. If the patient is new to the health system or arrives in an emergency and is not identifiable on arrival, this information may not be available for some time. Another example may be that there could be multiple times information could be presented. For example, a warning could be developed for prescribing a sleep aid for someone who is receiving high doses of narcotics, as this could cause severe respiratory depression. Should the alert be presented when the drug is selected during order entry, or only when the order is attempted to be signed? Implementation of such alerts requires and in-depth understanding of the clinical workflow and the needs of the end-user. The alert should not be interruptive to workflow and be presented at the most effective time to reduce alert fatigue and support clinical processes by saving time and frustration.

While the 5 rights are a good set of principles, it does not fully address all implementation issues. Specifically, when developing a CDS tool, you must consider the context of how this new tool would fit with the others already in place. Careful consideration of this is needed to prevent unintended consequences, such as unexpected changes to workflow or conflicting information. Additionally, continued performance monitoring is needed to ensure the intended function of the tool, as well as to monitor patient safety, both in terms of intervention efficacy and whether the clinicians act on the CDS. Multisite CDS may also present unique challenges, whether that be different workflows, data availability, or other facts that may vary how the CDS should be implemented.

These challenges represent areas that are open to research and development, many of which have been called out as priority areas to explore (Osheroff et al 2007). The The Agency for Healthcare Research and Quality (AHRQ) (see Additional Resources) is an active funder and coordinator of collaboration in this space. Despite the need for more detailed guidance, the 5 Rights of CDS provides a solid framework to conceptualize the early design of a CDS intervention.

GUIDES Checklist

The GUIDES checklist is a self-assessment tool developed to help potential CDS implementers (and health system leaders) identify and address factors that affect the success of CDS interventions (Van de Velde et al. 2018). The checklist identifies 4 domains with a total of 16 factors essential to the success of CDS:

The 5 Rights and the GUIDES checklist are well-known frameworks to help conceptualize and evaluate CDS, but others exist. Regardless of the framework you use, the core principals are usually the same: CDS should be based on established evidence, be easy to use, target the right people, and have regular evaluation. While these frameworks are a solid base to work from, there are many other considerations including how different CDS tools interact with each other within an EHR system in real settings. Next, we describe specific use cases for CDS in PCTs and further explore specific considerations for their implementation and evaluation.