Designing Experiences To Counter Alert Fatigue

BACKGROUND

It’s no secret that doctors despise the technology they use. It’s as if medical software was designed to be as infuriating as possible. A big part of providers’ frustration is how their systems handle alerts. For example, electronic health records (EHR) can easily produce hundreds of notifications a day, many of them duplicated or outdated. Alerts are only effective when their human counterparts are engaged and willing to react. Because of this, alerts are only helpful in small doses and when they’re properly designed. If they’re not, it will lead to alert fatigue, producing apathy [1] that harms both the patient and the provider.

Almost all alerts in medical software inherently promote alert fatigue. Alerts are intrusive with flashing visuals and loud noises that unnecessarily alarm patients and interrupt the healthcare provider. Oftentimes they’re overly simplistic, giving equal weight to both life-threatening and non-life-threatening situations. This quickly desensitizes the doctors to the alarms. It is similar to your response to phone calls and your likelihood of engaging, which decreases as more telemarketing calls come in. This can have devastating consequences if a doctor ignores a life-threatening alert.

WAYS TO IMPROVE ALERT DESIGN

By focusing first on only the drawbacks of the current design, we can already establish a few ground rules moving forward:

1. Talk to real, practicing doctors and caretakers before trying to implement new alerts.

2. Only allow audio alerts and pop-up visuals in extreme circumstances. Alerts must be critical,otherwise, doctors and other healthcare providers will become desensitized.

3. Notifications must have legible font and color combinations to make them easier to skim.

4. Allow providers to customize alerts so doctors may determine what is important to them and eliminate low-value notifications.

5. For lower-value notifications, have a dedicated section where providers can check these non-emergency alerts at-will.

6. Finally, don't alert if an alert is not needed. Put the information in the log instead.

Secondly, we can look at design guidelines for establishing life-threatening alerts:

1. “An alert signal should be accompanied by a clear, consistent, concise, and candid warning message"

2. Flashing, rather than static lights, preferably in one standard color…can be used to gain attention to visual warning messages.

3. To ensure all affected building occupants receive information, use multiple channels to disseminate the warning message – including visual means, audible means, and tactile means. Ensure there are no contradictions in the messages.

4. For limited message length, message writers could draft the message in a bulleted form; each of the five topics in the warning should be separated as individual bullet points.” [3]

SOLUTION FOR HIGH PRIORITY ALERTS

Many devices have alerts that can create a sense of chaos when dealing with a patient. Therefore, it would be more effective to centralize all alerts. Alerts from monitoring devices could be funneled into one system, which would lessen the cognitive load of keeping track of specific devices.

For the most urgent of emergencies, design should target three of the human senses—vision, tactile, and auditory—by using a standard flashing red light, some sort of vibration, and loud noise. This system could be a pager, or a stationary tablet at select positions throughout the medical facility.

There should also be a screen to communicate the exact nature of the emergency. The screen should be large enough to accommodate sentences of text. Ideally, a visual interface would read aloud and display a transcription of what the issue is, including:

1. Source: Which room and which patient does this indicate?

2. Hazard: Which measure of health is at critical levels?

3. Time: When did this start, and how long has it been going on?

For this message to be most effective, there must be a means for doctors and nurses to customize patient names and room numbers or room names.

OTHER ALERTS

We can determine how to account for less urgent alerts by doing the opposite of what a true urgency looks like and using rules for limited communication instead. For example, if a true emergency looks like flashing lights, plenty of noise, and tactile sensations, then an example of a lesser priority would only display an orange light that is stable without further sensory alarms.

For on-screen messages for provider interpretation, we can turn to communication rules to understand that the language should include “using simple, clear sentences - avoiding double negatives…[and] limit jargon.”[3] Although the medical field is a space filled with jargon and specialized words, caretakers must spend little time interpreting messages. Short messages make alert systems an occasional interaction and a way of logging the patient’s experience.

There should also be a way for providers to listen to alerts instead of always needing to read them. Alternate means opens an opportunity for multitasking, where a nurse could listen to a patient’s less urgent alerts while visually looking up information. It also accommodates the visually impaired.

POSSIBLE APPLICATION

Let's dive into what this solution would look like cohesively. We will take a hypothetical doctor and operate with the assumption that the system is software that can be downloaded onto a tablet. Additionally, we’ll assume this software can be paired with medical monitoring devices.

Persona

User Flow

BASIC SUCCESSFUL USE CASE

Users: Doctor/Nurse

Goal: Receive Emergency Alerts

Actions:

1. Before use, Dr. Doe and their team would customize the software parameters, including general patient information, connection to monitoring devices on the patient through wireless means like Bluetooth, and room identification. After a quick initial setup, a dashboard displaying overall patient health and options to view specific patients’ sub-dashboards appears.

2. When an emergency appears, the outer rims of the dashboard flash with a red light, and a transcript of the emergency alert displays as it is read out loud and the device vibrates. The alert is concise and clear, referring only to the main points of urgency. An emergency alert would override even the use of the dashboard, forcing attention to it.

3. To stop the alarm, Dr. Doe would need to do two actions:

   1. Tap the screen to indicate attention provided to the alert.

   2. Enter a short, individualized pin for each healthcare provider to indicate who is addressing the alert. This action also logs who cleared the alert and is now available in alert history for medical staff to review later.

ALTERNATIVE USE CASE

Users: Doctor/Nurse

Goal: View non-urgent alerts

Actions: Urgent alerts go into the history log of the main dashboard. Meanwhile, the specific patient sub-dashboards have another log of their less urgent alerts.

1. Less urgent alerts would have a brief orange outer rim light appear but not stay.

2. The orange alert is translated into an orange dot on the specific patient’s profile as an unread message with a number to indicate how many new alerts a patient has.

High Fidelity Example Mockup of Central Dashboard

CONCLUSION

Solving core issues like reducing alerts and creating a priority system that is easy to read and interact with may help reduce the severity and scale of medical alert fatigue. With this solution, a doctor like Dr. Doe can reduce the amount of interruption in his day to the bare minimum with a central place to view all alerts. Additionally, the system is simple, ensuring a minimal learning curve.

Although this solution may not be perfect in accounting for edge cases and more complex medical situations, it has the potential to save lives by addressing the widespread problem of alert fatigue. More consideration for the caregivers will benefit patients as well. It is imperative to continue iterating and testing these solutions so we may eventually optimize medical workflow.

References

[1] https://psnet.ahrq.gov/primer/alert-fatigue

[2] https://ehrintelligence.com/news/best-practices-to-mitigate-ehr-clinical-decision-alert-fatigue

[3] https://nrc-publications.canada.ca/eng/view/accepted/?id=6907d282-823c-41e3-908a-95c29adc7f05

[4] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5387195/

[5] https://www.premierinc.com/newsroom/blog/reducing-alert-fatigue-in-healthcare