Clinical decision support may help automate the ED, but as it reduces certain risks, is it creating new ones?
With the High Tech Act offering dollars for automation, the interest in emergency department information systems (EDIS) is growing exponentially. Hospitals are pushing the agenda of EDIS vendors, for the causes of cost containment and patient safety, the driving forces for these federally funded incentives.
One component of many emergency department information systems is clinical decision support. Hypothetically, such systems trigger actions and initiatives to reduce risk to patients presenting with high-risk clinical scenarios. In example, if a patient presents with chest pain between above the age of 40 years, clinical decision support tools may prompt the physician to consider many things, including ordering an ECG. Theoretically, such prompts would only occur for items with a recognized, irrefutable best practice or standard. Thus, the right things should happen for the right patients, mitigating individual practice styles.
The goals for such systems are improving quality of care and reducing risk. The initial genesis of clinical decision support came from the latter, risk management. Even before the explosion of electronic medical records and comprehensive ED information systems, standardized charting systems, specifically templates, were used for this purpose. Many of these tools included check boxes for items the clinician may not have considered or didn’t intuitively include in their evaluation. For instance, in abdominal pain patients, the physical examination section may include items such as rebound tenderness, guarding and McBurney’s point tenderness. Knowing that appendicitis is a frequent misdiagnosis, including such items may prompt the physician to perform these physical examination items and improve the quality of care delivered, their documentation and thus, reduced their risk.
Throw the power of an electronic solution behind the same concept and these prompts become active, as opposed to the passive, first-generation templated systems. In the paper template, you could just ignore those items in the chart. However, in the electronic age, for better or worse, the provider will be forced to participate. Furthermore, decision support has evolved into a tool applied to most patient encounters and is no longer limited to select high-risk encounters.
The concept of clinical decision support is an excellent one. However, just as with other components of EMRs or ED information systems, the law of unintended consequences comes into play. At what expense do we gain the benefits of decision support. Interestingly enough, in reducing risk from one direction, it has been theorized that risk is actually created in another. Most systems of this kind are designed by non-end users, and thus, have not encountered the impact of such tools on provider workflow. Many physicians have reported frustration with these systems, despite the value they are reported to provide. The end-users, emergency physicians, often find that what is gained in safety and reduced risk is lost in operational inefficiency, resulting in delays, extended length of stays and an increase in patients left without being treated; all elements we know to increase risk and negatively impact patient safety.
Through 390 hours of observation, Campbell identified 324 adverse and unintended consequences. These included the addition of extra work (19.8%) and workflow issues (17.6%). (Campbell, E.M., et al, J Am Med Informat Assoc 13(5):547, September-October 2006). In 2009, Campbell reported that workflow is disrupted by these systems, substantially increasing the provider’s workload by addressing unhelpful alerts and various other prompts. (Campbell, E.M., J Gen Intern Med 24(1):21, January 2009). Such prompts are the primary issue associated with these systems. Some physicians have reported having to transition through 10 or more prompts to order a single aspirin in systems with computerized physician order entry, accompanied by clinical decision support. Keeping the physician behind the computer and away from the bedside is unlikely to be a viable strategy for reducing risk and improving safety. Although the delivery of that one aspirin may or may not be safer, the operational cost of several mouse clicks for multiple medications for multiple patients in the course of a shift interferes with workflow to such an extent that the newly introduced risk may outweigh the hypothetical benefits gained. Fairly recently, in June 2008, the ACEP EM Informatics section reviewed the feasibility of translating the ACEP clinical policies into a real-time, computerized decision support tool. One of their primary conclusions was consistent with the concerns that have been raised. They reported that the subject matter is so complex in Emergency Medicine that with today’s technology, it is not possible to develop such a tool that will not significantly and negatively impact workflow.
They also raise the important point that in the most critical patients, the patient will require stabilization long before the physician can or will access the computer system, rendering clinical decision support useless in such situations. (Melnick ER, et al., Annals of emergency medicine 5 April 2010.)
Another element of clinical decision support systems are forced functions. Forced functions require action before the provider can do anything else. The value in forced functions is that they force immediate attention to an issue and compliance with the requirement. This could be making certain that a pregnancy test is ordered in all childbearing-aged females when the provider orders an X-ray or making certain a critical lab value is reviewed. Forced functions are valuable tools. However, they should be reserved for the most critical of actions, as they not only disrupt workflow, they stop it completely until the alert is cleared. Depending on how these are selected and how they are integrated into the system, they can be invaluable or simply a nuisance, resulting in the provider routinely clicking the mouse just to bypass them.
The ideal system is one designed to work in parallel with our work processes. Much like the Microsoft word notepad assistant that pops up on the page when assistance is needed, clinical decision support should be an ongoing guide throughout the ordering and documenting process of the encounter. It should interact with the physician, not force them to interact with it. It should provide ready access to brief summaries of the current evidence on the topic for that decision point. If such a system were not intrusive or obstructive, the opportunities to provide support would be limitless, as workflow would not be interrupted and provider resistance would be minimal. Physicians and patients will benefit from a system that provides guidance and helpful reminders at critical decision-making points. However, systems that routinely force interaction, in fact taking the primary role away from the physician, derail the delivery of care and can be counterproductive at best and have the potential to create substantial harm.
Ultimately, computerized clinical decision support will prove to be a useful tool. However, this assumption holds true only if the physician drives the care with support provided by the system and not the other way around.
Kevin Klauer, DO, is the Editor in Chief of Emergency Physicians Monthly