One of the main values of an EMR is to alert and guide the clinical decision making process -hence came the concept of clinical decision support (CDS) modules-. Clinicians have to deal on a daily basis with a flood of data and information that they need to make decisions upon, and if everything is in greyscale, mistakes are more likely to happen. CDS simply paints this grey image with bright colors for clinicians to easily identify areas that need attention. Examples like drug-drug interactions, drug-allergy interactions, doses outside recommended range, or labs that need to be checked before prescribing medication. Possibilities are endless, and here lies the problem.
Due to the fact that there is so much information to present to the clinicians, and each piece of information is important -subjectively-, in the old days of technology -and EMRs-, it was common to have all sorts of flashing, blinking, red alerts everywhere on the screen. And similar to having everything painted in shades of grey, painting the screen with too much color dilutes the message as well, and while everything is important, nothing is important anymore! Here is where the alert fatigue starts to take place.
How many is too many?
Think about traffic signs. You’re driving down the road and still you need quick indications on danger, directions, while keeping an eye out for pedestrians, and other sources of danger. In addition to street signs, and signals, modern cars’ dashboard present the driver with all sort of indications and alerts for objects in blind spots, lane change, speed, navigation, and probably a lot more.
Notice in the previous example, as a driver; you are showered with hints, cues and alerts all the time; however, attention fatigue for drivers is still much less of a problem, and usually is related to other factors -like lack of sleep-, rather than the street and car design. While in healthcare systems, alert fatigue has been observed for clinicians presented with as little as 12 alerts per day in a busy clinic seeing almost 40 patients a day. That means that 1 alert every 3 patients, was enough to kill the concept of CDS, cause alert fatigue, and drive the override rate all the way >90% for all types of alerts. Why?
Relevance is the key!
The more I look into alert fatigue cases, the more I realize; it’s mainly around alerts relevance rather than numbers. There are many factors that can lead clinicians to decide not to pay attention any more to alerts, but relevance is the most important one. Once you present clinicians with multiple irrelevant alerts that they have to override without a positive impact on their current plan of care, their perception of the CDS will shift down. We (as humans) are programmed to filter unimportant information. If at your workplace you hear a loud explosion you might get alarmed once or twice. However, if you work near a construction site, before you know it; your brain will start filtering out these explosion noises as unimportant and not worth the attention (Unless something changes significantly in the nature of the sound).
How different types of alerts look, is also important. During alert fatigue, once clinicians decide (even unconsciously) that these alerts are not relevant they simply look for the “OK”, or the “Override” button regardless of the content (Whether it’s a drug-drug interaction, drug-allergy, drug-lab check, etc.). I still remember the first time a physician told me “I just press OK, and move on!”. As long as the alert screen looks similar, people won’t invest the time to read the fine print.
Not all alerts are created the same
To fight this, you must categorize your alerts according to number of factors and carefully assign them to a different tool with a different look, feel and interaction. For example; before you put an alert in the system, ask yourself:
- Does the clinician need to take an action NOW?
- Should the alert stop the clinician’s current workflow and require them to perform something else that’s critical?
- Are there any valid reasons to override the alert? What’s the acceptable override rate?
- Would notifying the clinicians unobtrusively achieve the same goal?
- Can the alert be placed in the context of a current screen in the EMR (Like the medications list for example) instead of a pop-up type of screen?
Going through such process for every type of alert you have will help you achieve your objective with as little disruption as possible to the clinicians, which is a main factor in compliance with your CDS recommendations. Alerts should vary in their methodology according to answers to questions like the above. Some examples would be:
- A peculiar icon in a screen like the Medication Administration Record, or the medications list that notifies the physicians of a moderate interaction
- A message in their inbox that reminds the clinician of an action that should be taken, but not urgent
- A non-modal pop-up message (i.e. it’s shown on a corner of the screen and doesn’t interrupt the current workflow. However, clinicians see immediately that they have a message with a recommendation, but don’t need to attend to them on the spot. They can gracefully finish what they’re doing and then attend to the alert
- Finally, the infamous modal pop-up message with the red title on top, that interrupts current workflow, and forces the clinician to take immediate action. These should be your last resort
Improve based on evidence
It’s critical to always revisit the design of the alerts, especially if the logic is something that you can control in your design. Most CDS systems provide statics and reports on the number of fired alerts, clinicians’ response to the (Override rate, override reasons, etc.), triggers, trend, etc. It’s critical to regularly visit these data and look closely for trends like:
- Rise, or fall of alerts fired (Especially the top 10)
- Rise, or fall of override rate (%)
- Significant changes in the type of fired alerts
- Clinical outcomes related to these alerts
These numbers should drive you to continuously change your design and review your alerts system, and provide feedback to clinicians on their behavior. If (for example) some clinicians override alerts and don’t provide proper reasons, or if near misses happen despite the system’s log showing an alert properly presented; you should interview them and inquire about the effectiveness of CDS. Such actions could be early signs for alert fatigue.
Clinical Decision Support systems, should support your decisions in their design, as much as they support clinicians in providing better and safer care for the patients.