Abstract
Background: Telemetry monitoring is intended to improve patient safety and reduce harm. However, excessive monitor alarms may have the undesired effect of staff ignoring, silencing, or delaying a response due to alarm fatigue. Outlier patients, or those patients who are responsible for generating the most monitor alarms, contribute to excessive monitor alarms.
Methods: Daily alarm data reports at a large academic medical center indicated that one or two patient outliers generated the most alarms daily. A technological intervention aimed at reminding registered nurses (RNs) to adjust alarm thresholds for patients who triggered excessive alarms was implemented. The notification was sent to the assigned RN's mobile phone when a patient exceeded the unit's seven-day average of alarms per day by greater than 400%.
Results: A reduction in average alarm duration was observed across the four acute care telemetry units (P < 0.001), with an overall decrease of 8.07 seconds in the postintervention versus preintervention period. However, alarm frequency increased significantly (χ23 = 34.83, P < 0.001).
Conclusion: Implementing a technological intervention to notify RNs to adjust alarm parameters may reduce alarm duration. Reducing alarm duration may improve RN telemetry management, alarm fatigue, and awareness. More research is needed to support this conclusion, as well as to determine the cause of the observed increase in alarm frequency.
Telemetry monitoring is intended to improve patient safety and reduce harm. However, excessive monitor alarms may have the undesired effect of staff ignoring, silencing, or delaying a response because the alarm is falsely believed to be a “nuisance alarm.” Nuisance alarms are defined as monitoring device alerts that may be either false or true but are nonactionable.1–5 Research has shown that the majority of monitor alarms are nuisance alarms and can occur hundreds of times per day.5–9
Using alarms safely has been a priority of The Joint Commission (TJC) since it published a sentinel event alert in 2013 outlining the severity of alarm fatigue in the United States. Of every 98 alarm-related sentinel events, 80 resulted in patient deaths.10 TJC subsequently implemented National Patient Safety Goal 6 on clinical alarm systems and began auditing hospital compliance in January 2016.11
Research has found that between 72% and 99% of monitor alarms do not require intervention and are nuisance alarms that contribute to alarm fatigue.1,2,5,7,12 Bonafide et al.6 found that nursing staff had slower response times to true alarm events among patients who had a larger number of nuisance alarms preceding a true alarm. Research supports excessive alarms as being an important contributing factor to registered nurses (RNs)failing to respond to critical alarms in a timely fashion.2,7,10,12
A major contributing factor to excessive monitor alarms is patient outliers. Patient outliers are those patients whose monitoring devices generate the most alarms. In two studies, only a few patients generated greater than 50% of total telemetry monitor alarms.12,13 Yeh et al.12 noted that two patients caused 54% of all alarms in 24 hours, with 91% of those alarms being due to low heart rate. A team of experts assembled by the National Coalition for Alarm Management Safety found a consistent problem across health settings: Many of the alarms on a given unit were caused by only a few patients.13 When data were analyzed on a large scale across multiple health systems within this group, it was found that telemetry units had the largest quantity of alarms compared with both intermediate and intensive care unit settings,13 with high and low heart rate alarms being the most common. In one of the reported practice settings, three patients were responsible for generating 83% of the 719 monitor alarms on a unit. In another, 78% of alarms were attributed to four of 45 patients.13 Alarm customization, personalized for patient-specific parameters (e.g., heart rate), is suggested as a best practice to reduce the quantity of preventable nuisance alarms.2,12–15 By customizing and adjusting alarm parameters appropriately, the total quantity of nuisance alarms can be reduced.7
Alarm technologies (e.g., middleware) are effective at improving alarm notification and response times. Middleware technology allows for customization and filtering of alarms between the primary alarming device and the second receiving device.1 Using alarm escalation rules within middleware can reduce false alarms1 and disable alarm escalation after clinical staff acknowledge the alert on the receiving device.16 In one study in which middleware technology was used to filter alarm tones, the positive predictive value improved, demonstrating a reduction in staff response time to critical alarms.17
In many health systems, acute care telemetry units may have the largest quantity of generated monitor alarms13; however, limited research exists regarding alarm management in these settings. Most alarm research has occurred in critical care settings. RNs on acute care telemetry units are prone to experiencing the effects of alarm fatigue because they have an increased number of patients, and these patients are more susceptible to frequent false alarms because of increased mobility and care activities. Alarm fatigue was found to be the contributing factor that led to a delayed response to a patient event at the project facility. When RNs were surveyed before project implementation, most responded they “usually/always” become indifferent to alarms when they sound repeatedly.18 Alarm data from the project facility indicated that patient outliers may be a contributing factor leading to alarm fatigue.
This quality improvement (QI) project took place between August 1, 2021, and December 31, 2021, in the Department of Medicine (DOM) at a large academic medical center after review by the hospital's institutional review board.
Objectives
The purpose of this evidence-based QI project was to improve patient safety by implementing a technological intervention aimed at optimizing alarm systems on acute care telemetry units. The specific aim of this project was to facilitate a statistically significant reduction from the baseline for both alarm duration and frequency.
Methods
Description of the Unit
The DOM is a 185-bed division within the hospital. Patients in the DOM typically present with medical problems such as cardiac, renal, pulmonary, or infectious diseases. This QI project took place on four acute care telemetry units within the DOM. Alarm data from the project facility's acute care telemetry units demonstrated that one or two patients were responsible for generating the greatest number of monitor alarms.
At the project facility, all monitor alarms are conveyed to a central monitor station in each patient care unit, as well as to monitors in each patient's room. In addition, select monitor alarms are sent to the primary RN's mobile phone (iPhone; Apple, Cupertino, CA) using a middleware software solution (Connexall, Toronto, OH) that filters alarms using hospital-defined algorithms before sending the alerts to the RN's mobile phone. These algorithms include alarm escalation rules to be applied through the middleware, allowing for a reduction of total alarm quantity by filtering and/or delaying an alarm that may be due to patient movement or care activities, thereby allowing for system autocorrection.19 Mobile phone alarm priority and tones are set to a high priority for lethal rhythms; medium priority for heart rate, peripheral oxygen saturation (SpO2), and leads failure; and low priority for other select notifications.
At the project facility, RNs on acute care telemetry units are responsible for responding to all patient alarms without the use of a central monitoring staff. Many best practices for alarm management already have been included in the hospital's alarm management policy and implemented by the staff. These practices include proper skin prep and electrode placement, standardized telemetry alarm presets, alarm customization, and a telemetry discontinuation protocol.
As written in the current policy, RNs are permitted to change alarm parameter settings within 10% of the alarm preset without a provider order (heart rate 50/140 bpm, SpO2 88%). RNs are required to obtain a provider order for adjustments that exceed 10% from the alarm preset and document every shift that alarms are on, audible, reviewed, and customized on the patient's assessment flowsheet. To maintain consistency, the project was implemented on four DOM acute care telemetry units with a combination of monitored and nonmonitored beds. Each of the four units had a total patient capacity of 24 patients (mix of monitored and nonmonitored) but only had the capability for 12 telemetry-monitored patients. The units have a staffing pattern of one RN for every four patients during the day shift (7:00 am to 7:00 pm) and one RN for every five patients at night (7:00 pm to 7:00 am). All four units care for similar patients with diverse medical conditions.
Intervention
The technological intervention was conceptualized by the QI project lead and mentor and implemented by clinical engineering using the hospital's middleware solution (Connexall). The QI project team leveraged an existing monitor alarm “flood report,” which is emailed to select alarm “champions” daily (by noon). The flood report identifies telemetry-monitored patients who have exceeded the unit's rolling seven-day alarm per bed per day average by 200%.
The end goal is to heighten staff awareness of patient outliers and to prompt the RN to customize alarms to reduce the number of nuisance alarms and, hence, potential subsequent fatigue. Often, the assigned RN does not receive the emailed flood report to be notified of the need to adjust the alarms. The intervention capitalized on technology, as each assigned RN carries a mobile phone. Alarm flood data were sent directly to the RN's mobile phone when a patient's alarms exceeded the set threshold.
Although the hospital's traditional alarm flood report is set at 200% of the unit's seven-day rolling average, for this project, the report threshold was created to be more sensitive at 400%. This was done to avoid further contributing to alarm fatigue as a result of excessive alerts being sent to RNs' mobile phones. After a monitored bed reached 400% of the unit's seven-day rolling average alarms per bed per day, an alert was sent to the assigned RN's phone indicating that the patient had reached the unit's alarm threshold, as well as instructing the RN to review and adjust the patient's alarm settings. The alert was sent one time per shift, as soon as the patient violated the alarm threshold. This could occur whenever the alarm threshold was exceeded.
The intervention alert went through two modifications. The original (Figure 1, left, midintervention) alert included the following message: “ACTION REQUIRED: Bed # has triggered X monitor alarms today. Please review and adjust alarm thresholds as needed.” The alert then was modified (Figure 1, right, postintervention), in the following two ways, to give more precise information. (1) RNs were notified regarding which type of alarm was causing the patient to exceed the unit alarm threshold. (2) The information was adjusted to include a more precise calculation, updating every 12 hours, with the alert reading as follows: “UNIT BED# Alarm ACTION REQUIRED: patient has triggered x monitor alarms in the last 12 hours. Please review and adjust alarm thresholds as needed. Highest Offending Alarm is X Alarm.” The decision to modify the alert midway through the project was based on providing assigned RNs with more specific information to guide their alarm-adjustment decision making.
Figure 1.
Alarm alerts for the midintervention (left) and postintervention (right) periods. Left: The midintervention alert, which was deployed from Oct. 15, 2021, to Nov. 11, 2021. Right: During the end of the midintervention period, a way to make the alert more specific was discovered and implemented during the postintervention period (Oct. 22, 2021, to Dec. 31, 2021). This improvement had two benefits. (1) It allowed the type of alarm that was causing the patient to exceed the unit alarm threshold to be shown. (2) It allowed a change in alarm frequency from once every 24 hours to once per shift, which empowered registered nurses to impact alarms on all shifts, as clinically indicated. Abbreviation used: SpO2, peripheral oxygen saturation.
The QI project director led a team consisting of nurse educators, nurse leaders, unit RNs, and clinical engineering staff. Before project implementation (by Sept. 1, 2021), the project director met with RNs in the four acute care telemetry units and conducted education on the hospital-specific alarm management policy, alarm adjustments, and technological intervention.
An alarm threshold adjustment decision tree tool (Figure 2) was developed by the QI project team and emailed to all RNs, posted in staff breakrooms, and placed at all central monitor stations. This education tool was designed to assist RNs with alarm-adjustment decision making. For program success, staff in clinical nursing leadership on each of the four DOM acute care telemetry units were leveraged as key stakeholders.20,21 These individuals helped the QI project director educate on the intervention. Education was done individually on each unit through rounding and via an annual education session.
Figure 2.
Clinical alarm decision tree tool. The decision process shown reflected the organizational policy for alarm parameter adjustment and was posted at all central telemetry monitoring stations in participating study units. The tool guided actions of registered nurses based on what type of alarm was triggering a technological alert. Abbreviations used: HR, heart rate; SBP, systolic blood pressure; VT, ventricular tachycardia.
Alarm Data Collection
For data collection, the objective was to analyze 100% of available alarm data during the predetermined project intervals. Data were collected in three distinct phases. Phase 1 (preintervention) took place from Sept. 1, 2021, through October 14, 2021; phase 2 (midintervention) from Oct. 15, 2021, through Nov. 21, 2021; and phase 3 (postintervention) from Nov. 22, 2021, through Dec. 31, 2021. Alarm data were organized in Microsoft Excel spreadsheets (version 16; Microsoft, Redmond, WA), and significance was determined using run through the statistics program R (version 3.6.3; R Foundation for Statistical Computing, Vienna, Austria) to determine the significance of the intervention.
Only alarms lasting longer than 10 seconds and through a duration of 1,800 seconds were included in the analysis. These preestablished criteria sought to eliminate monitor alarms for which RNs would not have had time to respond, those that autocorrected without intervention, and those that were due to patients being detached from monitors to go off the unit for tests, procedures, or hygiene activities.
Results
Alarm Data: Duration and Frequency
Two-tailed Welch's two-sample t tests were run to compare alarm durations for the preintervention (77.92 ± 1.27 s [mean ± SE]), pre- to midintervention (74.35 ± 1.27 s; t39682 = 2.04, P = 0.041), and pre- to postintervention (69.84 ± 2.00 s; t11869 = 3.45, P < 0.001) periods (Figure 3). To further evaluate the relationship between intervention and reduced alarm duration, a linear regression (Table 1) was conducted to predict alarm duration based on intervention time point (pre-, mid-, and postintervention). The regression analysis showed a reduction of 3.57 seconds from pre- to midintervention (P = 0.040) and a reduction of 8.07 seconds from pre- to postintervention (P < 0.001).
Figure 3.
Mean alarm duration in each intervention phase with standard error. The bar graph demonstrates the mean alarm durations during each phase of the intervention. The preintervention period ranged from Sept 1, 2021, to Oct. 14, 2021; midintervention from Oct. 15, 2021, to Nov. 11, 2021; and postintervention from Oct. 22, 2021, to Dec. 31, 2021. Preintervention alarm data had a mean duration of 77.92 ± 1.27 seconds, midintervention alarm data 74.35 ± 1.27 seconds, and postintervention alarm data 69.84 ± 2.00 seconds.
Table 1.
Alarm duration regression. The results from the regression analysis, predicting alarm duration using the preintervention alarm condition as the intercept for the midintervention (Oct. 15, 2021, to Nov. 21, 2021) and postintervention (Oct. 22, 2021, to Dec. 31, 2021) periods, are presented. No statistically significant change was predicted from pre- to midintervention (P = 0.040), but a significant change was predicted from pre- to postintervention (P < 0.001) (F2,46906 = 6, P = 0.002).
To assess the impact of the intervention on the average frequency of alarms per day across the four acute care telemetry units, chi-square analyses (Figure 4) were performed, comparing the pre- to midintervention and pre- to postintervention time points. Comparing pre- to midintervention (χ23 = 3.07, P = 0.380), no meaningful change in alarm frequency was seen. Comparing pre- to postintervention (χ23 = 34.83, P < 0.001) a significant increase in frequency was observed.
Figure 4.
Average alarm frequency per unit per day, based on chi-square analysis. The preintervention period ranged from Sept 1, 2021, to Oct. 14, 2021; midintervention from Oct. 15, 2021, to Nov. 11, 2021; and postintervention from Oct. 22, 2021, to Dec. 31, 2021. Comparing pre- to midintervention (χ23 = 3.07, P = 0.380), no meaningful change in alarm frequency occurred. Comparing pre-to postintervention (χ23 = 34.83, P < 0.001), a significant increase in frequency was observed. This increase may be attributed to a surge in COVID-19 patients requiring continuous peripheral oxygen saturation monitoring during the postintervention period.
Alarm Perception Survey
A total of 93 RNs responded to the preintervention survey, and only 21 RNs responded to the postintervention survey. Due to the low postsurvey response rate, no statistical conclusions were drawn from the data.
Discussion
Aim 1: Alarm Duration and Frequency
Alarm data analysis demonstrated a statistically significant reduction in overall alarm duration across the four DOM acute care telemetry units. These data suggest that the technological intervention was a beneficial alarm management strategy. Targeting patient outliers by notifying RNs through their mobile phone when patients exceeded the unit's average alarms per bed per day empowered them to customize alarm threshold adjustments for patients. Literature supports that nuisance alarms often are caused primarily by only one or two patients and that nuisance alarms are an important cause of alarm fatigue.2,7,10,12 The four DOM acute care telemetry units' RNs recommended continued use of this intervention as a method to reduce alarm duration, and plans are in motion to expand to other units within the organization.
An unexpected finding, based on chi-square analysis comparing the pre- to postintervention periods, was that an increase in alarm frequency occurred rather than a decrease (Figure 4). A decrease in alarm frequency was predicted based on the significant reduction in alarm duration. A possible explanation for the increase in alarm frequency was the Omicron variant surge in COVID-19 patients admitted to the four DOM acute care telemetry units during the postintervention data collection period (December 2021). The patients may have required increased monitoring. However, when SpO2 data were excluded from the data set, a statistically significant increase in alarm frequency persisted. Considering the significant reduction in overall alarm duration, paired with an increase in alarm frequency, any alarm duration reduction was viewed as a meaningful finding. Alarm frequency data analysis should be reassessed at a time point not affected by a COVID-19 surge.
Occasionally, RNs communicated to the QI project director that they received mobile phone alerts sooner than anticipated. Units could have as many as 12 telemetry-monitored patients at any given time. It was determined that when fewer patients were on telemetry, the unused telemetry monitors were recorded in the middleware data as zero. The zero recording was being calculated into the unit's seven-day rolling average, thereby falsely lowering the unit's average alarms per bed per day. This caused alerts to be sent to the RNs' mobile phones sooner than if more patients were being monitored with telemetry during the unit's seven-day rolling average time frame.
Limitations
Because this research was conducted on four acute care units with a specific type of telemetry-monitoring device and policies, it's uncertain whether the results would be generalizable to other similar care units. Data from this project were contingent on the hospital's available alarm technology, which may not be available at other facilities. In addition, the project facility has a policy that allows RNs to individually adjust alarm thresholds for patients. This is a recommended best practice, whereas institutions with policies that do not allow for RN-managed alarm adjustments could have limited success with this intervention.
A significant limitation of this QI project was the implementation timing. The postintervention time frame had a large fluctuation in COVID-19 patients and staffing challenges. During the project rollout, the number of RNs floating among the DOM acute care telemetry units increased. It is possible that RNs floating among units were not educated on the technological intervention because it was only implemented on four of the eight DOM acute care telemetry units. Although zero sentinel events were reported on the acute care telemetry units during project implementation, additional research is recommended to determine the true impact and clinical significance for patient safety.
Conclusion
In the QI project described here, a technological intervention was implemented with the goal of heightening RN awareness of patient outliers—or those patients responsible for most monitor alarms—and thereby empower RNs to adjust alarms in a meaningful way. The QI project demonstrated a reduction in alarm duration but an increase in alarm frequency. This may have been due to the timing of the project, which occurred during a surge in COVID-19 cases. More research is needed to determine if the technological intervention described here can reduce alarm duration and frequency.
Acknowledgments
This QI project would not have been possible without assistance from Matthew Lehet, PhD (Michigan State University), who helped with programming the statistical software used for data analysis and interpretation. The authors also thank Yih-Jang Chang (Johns Hopkins Hospital), who allowed access to the raw alarm data for multiple months on the participating inpatient units. Without access to these data, the project would not have been possible.
References
- 1. Winters BD , Cvach MM , Bonafide CP , et al. Technological distractions (part 2): a summary of approaches to manage clinical alarms with intent to reduce alarm fatigue . Crit Care Med. 2018. ; 46 ( 1 ): 130 – 7 . [DOI] [PubMed] [Google Scholar]
- 2. Graham KC , Cvach M . Monitor alarm fatigue: standardizing the use of physiological monitors and decreasing nuisance alarms . Am J Crit Care. 2010. ; 19 ( 1 ): 28 – 34 . [DOI] [PubMed] [Google Scholar]
- 3. Cvach MM , Frank RJ , Doyle P , Stevens ZK . Use of pagers with an alarm escalation system to reduce cardiac monitor alarm signals . J Nurs Care Qual. 2014. ; 29 ( 1 ): 9 – 18 . [DOI] [PubMed] [Google Scholar]
- 4. Cvach M , Rothwell KJ , Cullen AM , et al. Effect of altering alarm settings: a randomized controlled study . Biomed Instrum Technol. 2015. ; 49 ( 3 ): 214 – 22 . [DOI] [PubMed] [Google Scholar]
- 5. Sendelbach S , Funk M . Alarm fatigue: a patient safety concern . AACN Adv Crit Care. 2013. ; 24 ( 4 ): 378 – 86 . [DOI] [PubMed] [Google Scholar]
- 6. Bonafide CP , Lin R , Zander M , et al. Association between exposure to nonactionable physiologic monitor alarms and response time in a children's hospital . J Hosp Med. 2015. ; 10 ( 6 ): 345 – 51 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Gazarian PK . Nurses' response to frequency and types of electrocardiography alarms in a non-critical care setting: a descriptive study . Int J Nurs Stud. 2014. ; 51 ( 2 ): 190 – 7 . [DOI] [PubMed] [Google Scholar]
- 8. GE Healthcare . Bringing order out of chaos . www.gehealthcare.com/insights/white-paper/bringing-order-out-of-chaos . Accessed March 28 , 2023. .
- 9. Turmell JW , Coke L , Catinella R , et al. Alarm fatigue: use of an evidence-based alarm management strategy . J Nurs Care Qual. 2017. ; 32 ( 1 ): 47 – 54 . [DOI] [PubMed] [Google Scholar]
- 10. The Joint Commission. Medical device alarm safety in hospitals . www.jointcommission.org/resources/sentinel-event/sentinel-event-alert-news-letters/sentinel-event-alert-issue-50-medical-device-alarm-safety-in-hospitals . Accessed March 28 , 2023. . [PubMed]
- 11. The Joint Commission . National Patient Safety Goals: effective January 2021 for the hospital program . www.jointcommission.org/-/media/tjc/documents/standards/national-patient-safety-goals/2021/npsg_chapter_hap_jan2021.pdf . Accessed July 31 , 2022. .
- 12. Yeh J , Wilson R , Young L , et al. Team-based intervention to reduce the impact of nonactionable alarms in an adult intensive care unit . J Nurs Care Qual. 2020. ; 35 ( 2 ): 115 – 22 . [DOI] [PubMed] [Google Scholar]
- 13. Cvach M , Kitchens M , Smith K , et al. Customizing alarm limits based on specific needs of patients . Biomed Instrum Technol. 2017. ; 51 ( 3 ): 227 – 34 . [DOI] [PubMed] [Google Scholar]
- 14. Bi J , Yin X , Li H , Li H , et al. Effects of monitor alarm management training on nurses' alarm fatigue: a randomised controlled trial . J Clin Nurs. 2020. ; 29 ( 21-22 ): 4203 – 16 . [DOI] [PubMed] [Google Scholar]
- 15. Voepel-Lewis T , Parker ML , Burke CN , et al. Pulse oximetry desaturation alarms on a general postoperative adult unit: a prospective observational study of nurse response time . Int J Nurs Stud. 2013. ; 50 ( 10 ): 1351 – 8 . [DOI] [PubMed] [Google Scholar]
- 16. Allen SH , Doyle PA , Sapirstein A , Cvach M . Implementation of alarm reduction interventions in a cardiovascular surgical ICU . Jt Comm J Qual Patient Saf. 2017. ; 43 ( 2 ): 62 – 70 . [DOI] [PubMed] [Google Scholar]
- 17. Hansen CJ , Rayo MF , Patterson ES , et al. Perceptually discriminating the highest priority alarms reduces response time: a retrospective pre-post study at four hospitals . Hum Factors. 2021. ; 187208211032870 . [DOI] [PubMed] [Google Scholar]
- 18. Torabizadeh C , Yousefinya A , Zand F . A nurses' alarm fatigue questionnaire: development and psychometric properties . J Clin Monit Comput. 2017. ; 31 ( 6 ): 1305 – 12 . [DOI] [PubMed] [Google Scholar]
- 19. Hoglund D , Elms J . The use of mobile devices to improve alarm systems . Biomed Instrum Technol. 2011. ; 45 ( 5 ): 381 – 4 . [DOI] [PubMed] [Google Scholar]
- 20.Bryson JM, Patton MQ. Analyzing and engaging stakeholders. In: Newcomer KE, Hatry HP, Wholey JS, Eds. Handbook of Practical Program Evaluation. 4th ed. Hoboken, NJ: Wiley; 2015:36–61. [Google Scholar]
- 21.Dempsey C, Reilly BA. Nurse engagement: what are the contributing factors for success? Online J Issues Nurs. 2016;21(1):2. [DOI] [PubMed] [Google Scholar]