Emergency department monitor alarms rarely change clinical management: An observational study?

William Fleischman, MD, MHS ^a,b,?, Bethany Ciliberto, RN ^c, Nicole Rozanski, RN ^c,

Vivek Parwani, MD ^b, Steven L. Bernstein, MD ^b,d

^a Department of Patient Safety & Quality, Hackensack Meridian Health, Edison, NJ, United States of America

^b Department of Emergency Medicine, Yale School of Medicine, New Haven, CT, United States of America

^c Yale-New Haven Hospital, New Haven, CT, United States of America

^d Department of Health Policy, Yale School of Public Health, New Haven, CT, United States of America

a r t i c l e i n f o

Article history:

Received 8 June 2019

Received in revised form 22 July 2019

Accepted 25 July 2019

Keywords:

Alarm fatigue Monitoring

Emergency department alarms Telemetry

Monitor alarms

a b s t r a c t

Study objective: Monitor alarms are prevalent in the ED. Continuous electronic monitoring of patients’ vital signs may alert staff to physiologic decompensation. However, repeated False alarms may lead to desensitization of staff to alarms. Mitigating this could involve prioritizing the most clinically-important alarms. There are, how- ever, little data on which ED monitor alarms are clinical meaningful. We evaluated whether and which ED mon- itor alarms led to observable changes in patients’ ED care.

Methods: This prospective, observational study was conducted in an urban, academic ED. An ED physician com- pleted 53 h of observation, recording patient characteristics, alarm type, staff response, whether the alarm was likely real or false, and whether it changed clinical management. The primary outcome was whether the alarm led to an observable change in patient management. Secondary outcomes included the type of alarms and staff responses to alarms.

Results: There were 1049 alarms associated with 146 patients, for a median of 18 alarms per hour of observation. The median number of alarms per patient was 4 (interquartile range 2-8). Alarms changed clinical management in 8 out of 1049 observed alarms (0.8%, 95% CI, 0.3%, 1.3%) in 5 out of the 146 patients (3%, 95% CI, 0.2%, 5.8%). Staff did not observably respond to most alarms (63%).

Conclusion: Most ED monitor alarms did not observably affect patient care. Efforts at improving the clinical signif- icance of alarms could focus on widening alarm thresholds, customizing alarms parameters for patients’ clinical status, and on utilizing monitoring more selectively.

(C) 2019

Introduction

Background and importance

Many ED patients are placed on continuous cardiopulmonary moni- toring with telemetric monitoring of heart rhythm, pulse oximetry, and respiratory rate. However, subsequent alarms may not reflect true clin- ical circumstances [1], or be Clinically meaningful [2]. The increasing use of continuous electronic monitoring in the acute care setting has spawned the phenomenon of “alarm fatigue,” or desensitization of Healthcare staff to electronic alarms. Alarm fatigue has been tied to pa- tient harm including death. In 2013, The Joint Commission issued a

? Prior presentation: Robert Wood Johnson Foundation Clinical Scholars Program Annual Meeting, November, 2015, Seattle, WA.

* Corresponding author at: Hackensack Meridian Health, 343 Thornall St, 8th fl, Edison, NJ 08827, United States of America.

E-mail address: [email protected] (W. Fleischman).

“sentinel event alert” for 98 reported alarm-related events, 80 of which resulted in death. While multiple contributing factors were iden- tified for each event, alarm fatigue was the most common contributing factor [3].

Human responses and adaptations to alarms are complex, and they have been shown to vary based on a variety of factors including the sound pattern such as alarm learnability and urgency [4], concurrent “masking” alarms [5], as well as by how reliably “true” a particular alarm tends to be [6]. Response may also be related to the importance ascribed to specific alarms [7]. Improving alarm response requires a multimodal approach, with a key focus on prioritizing alarms that are most clinically important [8-11]. There are, however, little data on the what ED monitor alarms are most clinical meaningful.

A prior study of alarms in ED chest pain patients found that over 99% of alarms represented “false” alarms, or alarms that did not represent a true clinical picture, and fewer than half of true alarms changed clinical management [2]. The clinical significance of monitor alarms in a general ED population remains unclear.

https://doi.org/10.1016/j.ajem.2019.158370

0735-6757/(C) 2019

Goals of this investigation

We sought to evaluate the clinical value of monitor alarms in a gen- eral ED population by observing patients being monitored and the resulting alarms, noting the types of alarms, the staff responses, and whether and how the alarms affected clinical management.

Methods

Study design and setting

We conducted a prospective observational study at an urban, aca- demic ED with an annual census of approximately 100,000. The ED is di- vided into three care areas by acuity. Observation was conducted in the 23-bed highest-acuity care area. Nurses are generally assigned no N5 pa- tients, with extra nurses assigned to the 4 resuscitation bays. Rooms are separated by hard walls and sliding glass doors, and each room is equipped with a Phillips MP30 or MP5 (Phillips USA, Andover MA) monitor. The monitor settings were default manufacturer settings com- monly found in hospitals throughout the United States (Table 1). The monitors also transmit rhythm, heart rate, respiratory rate, pulse oxim- etry, and, if measured, end-tidal CO₂ to central stations located at desks usually occupied by nursing staff and department clerks. Clinicians, in- cluding attending physicians, residents, and physician assistants, gener- ally sit in a glass-enclosed space that has a telemetry monitor installed at ceiling level, but which does not have the ability to silence alarms. The study was approved by the Yale University Human Investigations Committee. Informed consent was not required of staff or patients.

Data collection and processing

The study protocol was developed by two ED physicians (WF and SLB) and two emergency registered nurses (NR and BC). A single ED physician (WF) completed 53 h of observation from June 9, 2015 to Sep- tember 21, 2015. Observation times were during weekdays between 10 AM and 8 PM, during periods of peak ED activity. The observer was stationed at either of two ED staff desks that provided direct line of sight coverage of 7 to 8 patient rooms, as well as a central station mon- itor mirroring data for the rooms being observed. The observer moni- tored the patients in the selected set of rooms and the patients’ alarms during their time in the ED. To reduce Hawthorne effect [12], staff was told a study was being done to “determine what vital signs were impor- tant for ED patients.” For each alarm the observer recorded patient age (up to age 90), sex, chief complaint, type of alarm, observable response by staff including technicians, nurses, or clinicians, whether the alarm was “real” or “false,” and whether the alarm led to changes in clinical

Table 1

Monitor thresholds for alarms in the study ED.

Setting Parameters for alarming

Heart rate, bpm b50, N120

Pulse oximeter, % b90

Systolic blood pressure, mm Hg b90, N160

end-tidal CO2, ppm b25, N60

management. Logistical restraints did not permit the observer to record details about patients who did not generate alarms. We excluded alarms that sounded as the patient was being connected to the monitor or alarms during initial Patient evaluation and stabilization with clinicians at the bedside. Data was collected using an iPad with a customized elec- tronic data collection form (Filemaker Pro, Santa Clara, CA).

Categorization of alarms as “real” or “false” was based on the observer’s physical observation of the patient and on their previous and/or subsequent vitals and electrocardiographic rhythm strips, using methods similar to a previously published study of ED alarms [1]. For example, if a monitor sounded an alarm for a premature ventricular contraction (PVC), the observer reviewed the recorded rhythm strip and determined whether a PVC had likely occurred. Similarly, if an alarm sounded for a low pulse oximetry reading or an abnormal respi- ratory rate, the observer evaluated the relevant waveforms and physi- cally observed the patient to determine the likely “true” status of the patient.

Staff actions were categorized in one of three ways: overt response with an attempt to address the alarm source through a change in mon- itor settings or adjustment of leads/sensors; overt response without change (staff walked to the bedside to view the alarm but was not ob- served changing leads or monitor settings or viewed and silenced the alarm at the central station), or no overt response (staff either did not notice the alarm or noticed the alarm but took no action in response). A change in clinical management was defined as one resulting as a di- rect consequence of the alarm. This was determined by the observer by following patient care visually, tracking the orders and notes in the electronic health record , and by querying the care team. For alarms that were not noticed by staff and were judged by the observer to have the potential to affect patient care, the observer informed the care team of the alarm and tracked for any changes in patient care. Ad- ditional details on observation methods are included in the Supplement.

Outcome measures and primary data analysis

The primary outcome measures were the number and type of alarms, the proportion of false alarms within each alarm type, the re- sponses of staff to alarms, and the number and proportion of alarms that led to changes in patient care or clinical management. Results are presented with descriptive statistics. We report medians and interquar- tile ranges (IQR) for non-normally distributed data and means and stan- dard deviation for normally distributed data. Data management and analysis was performed using Stata (version 14).

Results

During 53 h of observation, there were 1049 alarms associated with 146 monitored patients for a median of 18 alarms per hour of observa- tion (IQR 10 to 26), and a median of 4 alarms per patient studied (IQR 2 to 8). The distribution of the number of alarms by patient and by hour of observation is illustrated in Fig. 1. Patients were 50% female, with a me- dian age of 66 years (interquartile range, 51-79). The most common chief complaints were chest pain, shortness of breath, and trauma (Table 2).

Premature ventricular contractions/min Premature ventricular contractions/min

>=2

N10

The most common alarms were for, in order of decreasing frequency, premature ventricular contractions, disconnected pulse oximetry leads, and disconnected ECG leads (Fig. 2 and eTable 1 in the Supplement).

Respiratory rate, breaths/min b8, N30

The vital signs associated with 76% of alarms (95% confidence interval

Cardiac rhythms with audible alarms as part of arrhythmia monitoring

Asystole, ventricular fibrillation, ventricular tachycardia, ventricular rhythm, sinus pause, atrial fibrillation, irregular heart rate, supraventricular tachycardia (sustained and nonsustained), premature ventricular contractions, ventricular bigeminy, ventricular trigeminy, missed beat, QTc prolongation.

[CI] 73% to 79%) appeared to represent the true status of the patient and were thus categorized as “true” alarms. Only 8 out of 1049 observed alarms led to changes in clinical management (0.8%, 95% CI 0.3% to 1.3%) in only 5 out of the 146 patients studied (3.4%, 95% CI 0.2% to 5.8%, Table 3). Staff did not respond overtly to 64% of alarms (95% CI, 60 to 66), and silenced 31% of alarms (95% CI 28 to 34) without changing alarm settings or leads (Fig. 3).

a

0

10

20

30

40

50

N alarms

Premature vent. contraction

25

SpO2 sensor off ECG leads off

SPO2 low Atrial fibrillation Bradycardia

20

Blood pressure high Respiratory leads off Respiratory rate low

Tachycardia Irregular heart rate Respiratory rate high blood pressure cuff off

N patients

10 15

Unspecified lead off (SpO2/ECG)

Pacer not pacing Non-sustained V-tach Blood pressure low

BP measurement failed Ventricular bigeminy End-tidal CO2 high

5

Asystole

Ventricular tachycardia Pacer not capturing End irregular heart rate End atrial fibrillation

0

Sinus pause End-tidal CO2 low

Apnea

b

IV pump alarm CO2 tubing missing Ventricular rhythm Low Temperature

0

10

20

30

40

50

Observation hour

N patients N alarms

50

0 50 100 150 200

N alarms

Real False Changed Management

30

40

Fig. 2. Types and frequencies of observed monitor alarms. Categorization of alarms as “real” or “false” was based on physical observation of the patient and on previous and/or subsequent vitals and electrocardiographic rhythm strips.

0

10

20

Fig. 1. a. Histogram of the distribution of alarms for the 146 patients studied. b. Number of alarms and number of studied patients by hour of observation.

Limitations

Our study has important limitations. First, this is a small, single- institution study, chiefly intended to inform monitoring changes in the study ED. Thus, our findings may not apply to EDs with different populations, staffing levels, layout, or monitoring practices. Second, all observation was completed by a single observer not blinded to study hypotheses and aims. While this helped reduce variability for the

Table 2

Most common chief complaints as recorded from the electronic health record.

Chief complaint	N (%)
Chest pain	22 (15)
Shortness of breath	19 (12)
Trauma	14 (8.9)
Weakness	11 (7.5)
Altered mental status	7 (4.8)
Syncope	7 (4.8)
Fall	7 (4.8)
Abdominal pain	6 (4.1)
Back pain	6 (4.1)
Headache	5 (3.4)

subjective measures in the study (classification of alarms, observing for changes in clinical management), it could have biased classification of alarms or resulted in some missed alarms. Notably, our “false alarm” rate was similar to the rate found in a study in a UK ED which used three clinician observers [1]. Third, we limited our categorization of staff responses to alarms to whether a staff member overtly acted or did not seem to act upon the alarm as it would be difficult for a single observer to accurately determine if a staff member had seen or noticed an alarm through so much as a glance at a monitor. Therefore, it is pos- sible that the lack of overt responses to many alarms is because the alarm was indeed noticed and judged to be clinically unimportant. Fourth, observation hours were not chosen at random, but for peak ED census hours in order to maximize the number of patients and potential alarms the single observer would track in the study area. This may have affected staff responses to alarms. Fifth, we used a conservative ap- proach in associating alarms with changes in clinical management. For example, if an alarm immediately preceded a change in clinical manage- ment (e.g. hypertensive drip adjusted), we categorized the alarm as being associated with the change in clinical management even though that management change could have potentially occurred without an alarm as well. Finally, while staff were not told of the study’s true aims, the presence of an observer may have still affected staff behavior and made them more aware of alarms [12].

Discussion

In this prospective, observational study of monitor alarms in an urban, academic ED, we found monitor alarms usually represented the true status of patients, but most alarms were of questionable clinical value, and alarms rarely led to changes in clinical management.

Our study builds on and is consistent with prior literature in on ED monitor alarms. The low rate of alarms that changED patient care or clin- ical management we noted is consistent with a study of alarms in a

Table 3

Characteristics of 8 alarms in 5 patients that were followed by changes in clinical management.

Patient	Alarm	Clinical context	Intervention
1	Blood pressure 212/93	Hemorrhagic stroke	Antihypertensive drip titrated
1	Blood pressure 186/86	Hemorrhagic stroke	Antihypertensive drip titrated
1	Blood pressure 163/79	Cerebral hemorrhage	Antihypertensive drip titrated
2	Blood pressure 177/61	Subarachnoid hemorrhage	Antihypertensive drip titrated
3	Respiratory rate 35	Intoxication	Verbal reassurance
3	Respiratory rate 47	Intoxication	Sedatives administered
4	Expiratory CO2 61	Intoxication	Verbal and physical stimulation
5	Blood pressure 87/36	Altered mental status	Administration of IV fluids

cohort of 30 ED patients with chest pain, which found b1% of alarms changed clinical management [2]. Similarly, our 76% rate of “true” alarms is consistent with a study in a UK ED that found 75% of alarms reflected true clinical circumstances [1].

The fact that only 0.8% of alarms led to clinical management changes suggest that it may be safe to restrict continuous ED cardiopulmonary monitoring to patients at high risk of clinical deterioration. Indeed, all of the alarms that led to changes in clinical management were in pa- tients with diagnoses that put them at obvious high risk for clinical de- cline and altered mental state: subarachnoid hemorrhage, hemorrhagic stroke, and severe intoxication. Also notable is that all the critical alarms observed, including alarms for asystole, ventricular tachycardia, and apnea, were “false” alarms.

Current practice and monitor settings in many EDs emphasize the benefits of monitoring as improving patient safety given its ability to de- tect acute physiologic decline. This purported benefit has also led to in- creasing use of monitoring in non-ICU inpatient settings [8]. Similarly, the default parameters set by the monitor manufacturers for most mon- itors (which tend to remain unchanged in many hospitals) follow the dictum of “the more that is monitored, the safer the patient.” Essentially, as a diagnostic test, monitoring was set up to be highly sensitive, but poorly specific. Predictably, the resulting high false-positive rate led to the emergence of alarm fatigue, ultimately reducing the sensitivity of monitoring. The challenge, then, is to reduce the frequency of false- positives, the clinically-meaningless alarms.

Studies in Inpatient units have successfully reduced the number of alarms using a variety of strategies including changing default monitor thresholds [9], customizing parameters to patients’ needs [9,13], more frequent changes of ECG leads [13], and requiring a physician order to initiate continuous Pulse oximetry monitoring [14]. The studies

No overt response Silenced Changed lead/setting Intervened clinically

Real False

N alarms

200 400

600

800

Fig. 3. Staff response by whether the alarm was categorized as “real” or “false”. Non- responded-to alarms were either not noticed by staff or noticed but not acted upon.

0

reported reductions in alarm frequency ranging from 43% to 77%, though no clinical outcomes were measured. Measures such as chang- ing default monitor thresholds and requiring physician orders for mon- itoring could be easily implemented in the ED environment. The alarm changes being implemented in the study ED are included in eTable 2 in the Supplement. Such settings changes would not, however, reduce lead-off alarms, the most common type of alarm in our study. A poten- tial solution to these alarms may be single-lead, wireless ECG monitor- ing [15]. However, the technology is not widely available, is costly, and by itself does not address the many clinically-meaningless alarms. Algo- rithms that combine data from multiple leads and time-trends to reduce false-positive alarms have been proposed [16-18], but, to our knowl- edge, the use of such software has not yet been evaluated in the clinical setting.

Until a combination of hardware and software technologies emerge to produce a highly sensitive and highly specific monitoring system, more selective monitoring may be key to reducing alarm frequency and the resulting alarm fatigue. For example, many patients monitored in the ED or admitted to telemetry beds are low-risk patients with chest pain, and multiple studies have shown little to no benefit of monitoring in this group [19-23]. EDs could implement protocols to restrict contin- uous monitoring only to patients with significant potential for clinical decline based on the treating clinician’s judgement.

In summary, in the study ED monitor alarms rarely led to observable changes in patient care and most alarms did not trigger overt responses by staff. Efforts to reduce alarms in EDs should take into account local factors and may include configuring alarm thresholds customized to the ED environment, disabling alarms unlikely to result in changes in clinical management, and more selective monitoring of patients.

Author contribution statement

WF, SLB, NR, and BC conceived and designed the study. VP provided critical input on study design. SLB supervised the conduct of the study. WF conducted data collection, processing and analysis. WF drafted the manuscript, and all authors critically revised it. WF takes responsibility for the paper as a whole.

Declaration of Competing Interest

None.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2019.158370.

References

1. Way RB, Beer SA, Wilson SJ. Whats that noise? bedside monitoring in the emergency department. Int Emerg Nurs 2014;22(4):197-201.
2. Atzema C, Schull MJ, Borgundvaag B, Slaughter GRD, Lee CK. ALARMED: adverse events in low-risk patients with chest pain receiving continuous

   electrocardiographic monitoring in the emergency department. A pilot study. Am J Emerg Med 2006;24(1):62-7.

   The Joint Commission. New alert promotes medical device alarm safety in hospitals. Jt Comm Perspect-Sentin event [internet]. [cited 2013 Aug 3]; Available from: http:// goo.gl/ZIwFkp; 2013.

3. Edworthy J. Designing effective alarm sounds. Biomed Instrum Technol 2011;45(4): 290-4.
4. Hasanain B, Boyd AD, Edworthy J, Bolton ML. A formal approach to discovering si- multaneous additive masking between auditory medical alarms. Appl Ergon 2017; 58:500-14.
5. Bliss JP, Gilson RD, Deaton JE. Human probability matching behaviour in response to alarms of varying reliability. Ergonomics 1995;38(11):2300-12.
6. Momtahan K, Hetu R, Tansley B. Audibility and identification of auditory alarms in the operating room and intensive care unit. Ergonomics 1993;36(10):1159-76.
7. Feder S, Funk M. Over-monitoring and alarm fatigue: for whom do the bells toll? Heart Lung J Acute Crit Care 2013;42(6):395-6.
8. Graham KC, Cvach M. Monitor alarm fatigue: standardizing use of physiological monitors and decreasing nuisance alarms. Am J Crit Care 2010;19(1):28-34.
9. Chopra V, McMahon LF. Redesigning hospital alarms for patient safety: alarmed and potentially dangerous. JAMA 2014;311(12):1199-200.
10. Meredith C, Edworthy J. Are there too many alarms in the intensive care unit? An overview of the problems. J Adv Nurs 1995;21(1):15-20.
11. Campbell JP, Maxey VA, Watson WA. Hawthorne effect: implications for prehospital research. Ann Emerg Med 1995;26(5):590-4.
12. Dandoy CE, Davies SM, Flesch L, et al. A team-based approach to reducing cardiac monitor alarms. Pediatrics 2014;134(6):e1686-94.
13. Stephens Amanda R, Piepenbrink Jim, Whalen Deborah, Sloan Karin A. A multidisci- plinary initiative to reduce alarm fatigue from unnecessary pulse oximetry alarms [internet]. B102. Strategies to improve care quality in pulmonary and critical illness.

    American Thoracic Society; 2016. p. A4359 cited 2016 May 8. Available from: http:// www.atsjournals.org/doi/abs/10.1164/ajrccm-conference.2016.193.1_ MeetingAbstracts.A4359.

    Pollack Jr CV. Wireless cardiac event alert monitoring is feasible and effective in the emergency department and adjacent waiting areas. Crit Pathw Cardiol 2009; 8(1): 7-11.

14. Aboukhalil A, Nielsen L, Saeed M, Mark RG, Clifford GD. Reducing false alarm rates for critical arrhythmias using the arterial blood pressure waveform. J Biomed Inform 2008;41(3):442-51.
15. Baumgartner B, Rodel K, Knoll A. A data mining approach to reduce the false alarm rate of patient monitors. Conf Proc Annu Int Conf IEEE Eng Med Biol Soc IEEE Eng Med Biol Soc Conf 2012; 2012. p. 5935-8.
16. Imhoff M, Kuhls S, Gather U, Fried R. Smart alarms from medical devices in the OR and ICU. Best Pract Res Clin Anaesthesiol 2009;23(1):39-50.
17. Hollander MD, Judd E, Valentine RN, et al. Are monitored telemetry beds necessary for patients with nontraumatic chest pain and normal or nonspecific electrocardio- grams? Am J Cardiol 1997;79(8):1110-1.
18. Hollander JE, Sites FD, Pollack CV, Shofer FS. Lack of utility of telemetry monitoring for identification of cardiac death and life-threatening ventricular dysrhythmias in low-risk patients with chest pain. Ann Emerg Med 2004;43(1):71-6.
19. Durairaj L, Reilly B, Das K, et al. Emergency department admissions to inpatient car- diac telemetry beds: a prospective cohort study of risk stratification and outcomes. Am J Med 2001;110(1):7-11.
20. Estrada CA, Prasad NK, Rosman HS, Young MJ. Outcomes of patients hospitalized to a telemetry unit. Am J Cardiol 1994;74(4):357-62.
21. Snider A, Papaleo M, Beldner S, et al. Is telemetry monitoring necessary in low-risk suspected acute chest pain syndromes? Chest J 2002;122(2):517-23.