Patient safety organizations have repeatedly pointed to concerns surrounding alarm fatigue and its implications for patient harm.
In May, the American Association of Critical Care Nurses (AACN) issued a Practice Alert for both bedside care providers and nurse leaders to address this “significant threat to patient safety,” noting that recent studies estimate that “approximately 90 percent of alarms in various critical care settings are either false or clinically irrelevant.”
The source for alarm fatigue is an excessively high volume of these nonactionable alarms, which desensitize nurses and causes them to miss or not respond to alarms that could put patients at risk. For example:
• At just one hospital, more than 2.5 million alarms were triggered on bedside monitors in a single month.
• A Joint Commission infographic estimates that 85 -99% of alarms do not require clinical intervention.
We’ve been addressing alarm fatigue at the Johns Hopkins Health System since 2006. Over the years, our alarm committee has developed broad guidelines that have address the need for:
• Appropriate alarm settings
• Effective notification channels such as smartphones
• Clear and actionable alarm information
• Clear and timely response protocols
• Effective staff training
Technology has come a long way since we used floppy discs to share data and couldn’t imagine the capabilities of today’s middleware and smartphones. Here are five technology tools and data that we use (or are considering using) to improve our alarm management practices.
Suspend features
One of the best ways to lower the alarm burden for caregivers is to be extremely careful about what gets automatically forwarded to them. The suspend feature in our Connexall software platform can determine if an alarm is sent via the nurse call system or physiological monitor. Using customized algorithms, we can delay sending those that do require immediate attention.
The software also has a suspend and count feature that can filter and bundle high-volume alarms commonly seen in certain populations. In our cardiology unit, for example, our software counts tachycardia alarms less than 10 seconds in length; when six of these events occur in less than five minutes, the 6th alarm triggers a notification to the nurses’ phone.
The settings for parameters on suspend functions will vary depending on the care unit, the type of alarm (crisis, warning or technical) and events. At Johns Hopkins, crisis alarms are sent immediately to the nurse, while certain warning and many technical alarms (e.g. lead failures or bed disconnects) can be delayed or sometimes auto-corrected or silenced.
The impact of the suspend feature on alarm fatigue cannot be overestimated. For example, during just one week in April, we found that although we had 61,227 warning levels alarms on our physiological monitors, only 3,103 of these were messaged out to mobile devices.
At Johns Hopkins, patient calls are sent to the unit clerks before going to the nurses. We also use the scheduled suspend feature, which allows us, for example, to send alarms directly to the nurse at night in pediatric units since these do not have unit clerks at that time.
Algorithms for alarm escalation and thresholds
Alarm escalation pathways need to be highly customized to determine who receives the alarm, when escalation should occur, who receives the second notification and when this should stop. We’ve developed these escalation algorithms in partnership with Connexall based on:
• The type, age and condition of patients in the unit
• Clinical workflows
• The unit’s architectural layout, which affects proximity and visibility of room monitors
• Staffing schedules
• Number of patients per nurse
• Time of day
We continually adjust the algorithms for the patterns and timing of escalation as well as our alarm thresholds. More importantly, we encourage our nurses to make these changes in real time. In many cases, our nurses will not require physicians’ orders to adjust alarm parameters on the patient monitors to make them more actionable. For example, if a nurse already knows a patient has atrial fibrillation, the alarm doesn’t have to go off each time a non-threatening episode occurs.
“The beauty of middleware is being able to write all these rules within the software,” said Maria Cvach, director of policy management and integration for Johns Hopkins Health System. “For example, our escalation system is set up so that if somebody fails to put in the primary or the secondary person, the alert will still go automatically to the charge nurse.”
Alarm data and reports
Weekly reports enable us to identify trends and adjust our practices accordingly. By using the reports’ detailed information, we can identify outliers and dig deeper to find out what happened and why did it happen. Report data includes:
• Total weekly alarms per unit
• The units’ average number of alarms (advisory, warning, crisis and system) per bed/day
• Average number of alarms/day by alarm type
• Average duration of alarms by type
• Average duration to respond to nurse calls
• Graphics on frequency on alarm types for high volume units (e.g., SpO2LO in the PACU)
These reports give managers a tool to compare their numbers with peers in other units to understand why their alarm volume is out of range and how they might improve their practices. The data also is used by the units and alarm management committee to adjust alarm settings and timing for escalations. Other uses include:
• Daily reports as needed to alert the staff of events that warrant immediate attention
• Custom reports to address specific issues; for example, examining whether longer response times to patient calls contributed to lower patient satisfaction scores
Alarm filtering to smartphones/devices
When reports show lags in response times likely caused by alarm fatigue, custom filters can be created to ensure that notifications are timely and meaningful without overburdening caregivers. These can be based on the alarm level (advisory, warning, crisis), type and other conditions that determine when alarms will go directly to nurses’ smartphones. It’s also helpful to have a smartphone platform that integrates with nurse call systems so that lower priority alarms can be sent to unit coordinators.
One of the key goals of alarm filtering is to ensure that only actionable alarms appropriate to the unit’s patient population are sent directly to nurses. We have had some success with the “less is better” approach and continue to look for ways of filtering out useless alarms.
Real-time location services
This proprietary tool from Connexall can suppress an alert if a caregiver is already at the patient’s bedside, thus avoiding an unnecessary secondary alarm. It also allows you to set up parameters for escalations to alert the closest available staff member if the primary caregiver is in another room or on a break.
We’ve included this feature on our roadmap for future implementation because it would be especially helpful in our larger units, some of which are as long as two football fields. We’re also interested in employing a tool in development that would offer real-time monitoring of alarm volume for each bed. This would allow bedside nurses and managers to see alarm-related workloads relative to other beds in the same area and intervene when needed.
What we’ve learned
An alarm management program is always changing – it’s not something you can set and forget. Everyone needs to be involved, starting with an interdisciplinary alarm committee with representatives from every care unit that meets regularly to understand how alarm fatigue impacts nurses, physicians and patients. Informaticists and educators need to work with clinicians to understand how workflows intersect with and affect the use of technology.
Flexibility also is essential. Most alarm practices need to be customized to accommodate variations in staffing, physical settings and patient populations. (However, it is important to standardize the colors and sounds of auditory and visual alarms systemwide.) It’s also critical to allow nurses to customize alarm settings to make them more clinically relevant and actionable.
One of the most effective ways to minimize alarm fatigue is to have unit champions in each unit. These are the problem-solvers who ask questions, dig deeper and keep attention focused on how their units can improve alarm management practices.
Lastly, we all can learn how to take better advantage of technology. As my colleague Maria Cvach noted in a recent interview about alarm fatigue, “Don't forget to engage your manufacturers for assistance… I always say that we use our monitors to about 20 % of their capacity and they have a lot of features we don't even know about. So, get to know your monitors and get your vendor to help you with that.”
Robert “Jeff” Frank is a Clinical Engineer at Johns Hopkins Hospital, where Jeff coordinates and develops medical systems integrations including, nurse call, physiological monitoring, and Connexall middleware systems.