BACKGROUND: Concerns about alarm fatigue prompted The Joint Commission to issue a Sentinel Event Alert urging hospitals to minimize alarms. We previously conducted a quality improvement project on a single unit that reduced time on continuous pulse oximetry, a common source of physiologic monitor alarms, for patients with wheezing (ie, asthma and bronchiolitis, wheezing-associated respiratory infections).
OBJECTIVE: To study the impact of our improvement work on overall physiologic monitor alarm frequency for these patients.
METHODS: This was a retrospective cohort study at a freestanding children’s hospital over an 8-week period. We compared alarm count, including respiratory, cardiac, and pulse oximetry alarms, for patients admitted to the intervention unit with the alarm count for similar patients on a control unit by using the Wilcoxon rank sum test. We used negative binomial regression to evaluate differences in alarm count between the units, adjusting for age, medical comorbidity, and length of stay.
RESULTS: There were 101 patients on the intervention unit and 46 patients on the control unit. The percentage of patients with medical comorbidities was significantly higher on the intervention unit (P = .01). Median alarm count per day for patients on the intervention unit was lower; however, this difference was not statistically significant (71 vs 76 alarms per patient-day, P = .29). The multivariable model estimated a nonsignificant 6.4-count decrease in alarms for patients on the intervention unit.
CONCLUSIONS: Reducing continuous pulse oximetry use alone may not make substantial reductions in overall alarm counts. Even on our intervention unit, alarm burden remained quite high.
When hospitalized children with lower respiratory tract illness have improved and no longer require supplemental oxygen, continuous pulse oximetry (CPOx) has been associated with increased length of stay (LOS), and no demonstrated improvements in outcomes.1,2 The American of Academy of Pediatrics Bronchiolitis Care Practice Guidelines discourages routine use of CPOx in this patient population due to concerns about its lack of value added in these patients’ care.3 The Pediatric Hospital Medicine Subcommittee of the American Board of Internal Medicine’s Choosing Wisely Campaign also recommends against CPOx use in patients admitted with lower respiratory tract illness when they do not require supplemental oxygen due to concerns about prolonged LOS.4 Pursuant to this, we performed a quality improvement (QI) project that reduced clinically unnecessary CPOx monitoring for pediatric patients with wheezing (ie, asthma, bronchiolitis, and wheezing-associated respiratory infection) by >70% on a single unit.5
Alarm fatigue has been highlighted as a common contributing cause to serious in-hospital morbidity and mortality as health care workers become desensitized to alarms, leading to delayed or absent response to critical patient events.6 Despite studies suggesting questioning the effect of continuous monitoring on general hospital units,7 and recommendations from multiple organizations to reduce CPOx use, its effect on overall physiologic monitor alarm frequency on a general pediatric ward has not been quantified. Previous studies have focused primarily on changing monitor parameters as a means of reducing pulse oximetry alarms.8–10 The goal of this study was to investigate the effect of our pulse oximetry reduction QI project on overall physiologic monitor alarm frequency for patients with wheezing when compared with a non-QI unit.
Cincinnati Children’s Hospital Medical Center (CCHMC) is a 522-bed freestanding children’s hospital. General hospital medicine patients are primarily admitted to 2 units that are staffed by separate nursing groups. All patient beds are equipped with bedside physiologic monitors (Dash 3000 and 4000 models General Electric, Fairfield, CT) that transmit alarm data to a central alarm collection system. Per hospital policy, any patient placed on CPOx must also be placed on continuous heart rate and respiratory rate monitoring. The alarm threshold for low oxygen saturation alarms is 90%.
We previously performed a controlled, QI study that reduced medically unnecessary CPOx monitoring on a single general pediatric unit, referred to hereafter as the intervention unit.5 Our other general pediatric unit served as a control during the QI project. Our outcome measure was the time on CPOx after meeting predefined goals (eg, oxygen saturations on room air >90% for 2 hours or weaned to every 2-hour albuterol treatments). For the intervention unit, during the QI study period, we identified 2 special cause variation points associated with a shift in the median time on CPOx after meeting goals from 10.7 hours to 3.1 hours on the intervention unit. We did not target any interventions for the control unit but noted 1 special cause variation point associated with a shift in the median time on CPOx after meeting goals from 11.5 hours to 6.9 hours, likely due to unintentional spread of some interventions.5
Study Design and Study Population
This was a retrospective cohort study of children hospitalized with wheezing between October and December 2013; a 2-month sample was taken after initiation of process improvement work geared toward reducing medically unnecessary CPOx monitoring on a single unit.
Patients were included if they had a primary diagnosis of lower respiratory tract illness (asthma, 493.xx; acute bronchiolitis, 466.1x, and wheezing, 786.07) and were admitted to the intervention unit or the control unit during the study time period. Patients with these diagnoses are not preferentially admitted to 1 unit; however, other specific populations of patients (eg, adolescents with eating disorders) are preferentially admitted to the control unit, leaving fewer available beds for other diagnoses. We excluded patients with a home oxygen requirement because they are kept on CPOx per standard practice.
This study was reviewed by the CCHMC Institutional Review Board and was determined not to be research on human subjects.
Alarm data were collected from a central alarm collection system (Hill-Rom, Inc., Batesville, IN) that links all physiologic monitor alarms to a patient bed. Data are collected for all audible alarms occurring at the bedside monitor, including clinical alarms (eg, for heart rate or respiratory rate) and technical alarms (eg, leads fail or artifact). This system is not capable of identifying the lead (eg, pulse oximetry probe) or type (eg, low oxygen level) of the alarm; however, alarms from pulse oximeters comprise ∼36% of all alarms at CCHMC.11 Alarm data were included only from the time that patients were on the intervention or control units; alarm data from the emergency department and the ICU were excluded. Patient-level information was collected via structured chart review from the electronic health record and included age, gender, history of prematurity, presence of other significant medical comorbidities, need for ICU admission at any point during hospitalization, and LOS. Alarm count per patient-day was computed for each patient by dividing their total number of alarms by their LOS in days.
Descriptive statistics including count and percentage for categorical variables and median and interquartile range (IQR) for continuous variables were used to describe patient characteristics by unit and compared between units using χ2 test and Wilcoxon rank sum test, respectively. Unadjusted differences in total alarm count and alarm count per patient-day between intervention and control units were evaluated using Wilcoxon rank sum test. To adjust for patient age and presence of medical comorbidity in the evaluation of alarm count rate differences between units, we used a zero-truncated negative binomial model with total alarm count as the dependent variable and LOS in days as an offset in the model. Due to the difference in percentage of patients with medical comorbidities between the 2 units, we conducted a sensitivity analysis by excluding patients with comorbidities and evaluated alarm count per patient-day between intervention and control units by using Wilcoxon rank sum test. SAS version 9.4 (SAS Institute, Inc, Cary, NC) was used for all data analyses.
There were 101 patients on the intervention unit and 46 patients on the control unit (Table 1). The percentage of patients with medical comorbidities was significantly higher on the intervention unit compared with the control unit (P = .01). The most common comorbidities were prematurity, chronic lung disease, cerebral palsy, and congenital heart disease.
Median alarm count per day for patients on the intervention unit was 71 (IQR 24–155) vs 76 (27–191) on the control unit; however, this difference was not statistically significant (P = .29, Fig 1). The multivariable model, adjusting for age, presence of medical comorbidity, and LOS, estimated a nonstatistically significant 6.4-count average decrease per patient per day in alarms for patients on the improvement unit compared with the control unit (95% confidence interval –42 to 29). A decrease in 6.4 alarms for a nurse caring for 4 patients is equivalent to reducing alarm count by 1 alarm per hour. Among those subjects with no comorbidities, the median alarm count per day for patients (n = 74) on the intervention unit was 57 (IQR 22–143) versus 76 (26–225) on the control unit (n = 42) (P = .12).
Given that we did not find a statistically significant difference between groups, we computed a post hoc power analysis based on the observed variability in alarm count data with a coefficient of variance of 1.24. Assuming a log-normal distribution, the study sample size of 101 in the intervention group and 46 in the control group has 80% power at α level of 0.05 to detect the difference between 76 alarms per patient per day in the control group versus 47 alarms per patient per day in the intervention group.
Our QI intervention successfully reduced median excess time on CPOx in patients with wheezing in the intervention unit as indicated by special cause variation; however, the process improvement work did not make a meaningful impact on alarm count when compared with a control unit. This study adds to literature examining the impact of the Choosing Wisely Campaign recommendations.12–15 Although alarms from pulse oximeters have been stated as a concern regarding CPOx use,3 we did not see a significant difference between our intervention unit, where we used interventions aimed at reducing length of CPOx monitoring, and our control unit. However, in the subset of patients with no medical comorbidities, our observed difference in alarm count between the 2 units was 19 alarms per patient per day. Thus, our results appear to be complicated by the imbalance of patients with comorbidities.
Our work adds to a growing literature illustrating the difficulties of moving outcomes even with successful bronchiolitis QI interventions. McCulloh and colleagues16 in a recently published randomized controlled trial found that intermittent (versus continuous) pulse oximetry did not reduce LOS. Our team also failed to demonstrate a reduced LOS coincident with our QI intervention that decreased time on CPOx after clinical improvement.5 Collectively, this reinforces the challenge of evaluating the effectiveness of improvement interventions in complex and adapting systems. Our single-center study performed over a 2-month period does not rule out the existence of a clinically important effect that reduced CPOx use would have on total alarm counts; it does demonstrate that QI interventions often may not have as large an effect on outcomes as we would intuit or hope.
Previous work attempting to reduce alarms from CPOx has focused on altering alarm thresholds or alarm delays and has occurred largely in the adult population.8–10 Although these interventions reduced alarm rates in adult populations, continuously monitored children may be more subject to motion artifact due to activity17 and may have different reasons for alarming than adults because of different underlying pathology. A recent commentary proposed a framework for thinking about alarm fatigue that included strategies to reduce monitoring time as well as reduce the number of alarms that occur during monitoring.18 Although interventions aimed at earlier discontinuation of monitoring, as in our study, may potentially decrease the number of alarms overall, they do not address the excessive numbers of alarms occurring during times of monitoring. More recently, improvement work focusing on alarm rates in pediatric units has looked at both earlier discontinuation of continuous monitoring and addressing false alarms as a means of successfully reducing alarm rates.19 It is likely that interventions are needed to address both length of monitoring and reducing alarms during monitoring to achieve clinically significant reductions in alarm rates.
This was a single-center study with a limited sample size conducted during a QI intervention. It is possible that in a larger study in a different context we may have seen a clinically meaningful difference. The study was also limited in its ability to detect reductions in specific alarm types; the alarm collection software did not distinguish among heart rate, respiratory rate, and pulse oximetry alarms. This study was also limited in that, although we did not target any interventions at our control unit, we did note a reduction in time on CPOx likely due to unintentional spread of our interventions. This may have biased our results toward the null of no difference in alarm counts between units. Finally, our study was limited in that there was a significant difference between the units with respect to comorbidities. Despite adjusting for several covariates, it is possible residual confounding remains. Frontline care providers may use different monitoring strategies depending on the disease, patient age, or unit culture.
Reducing CPOx use alone may not make substantial reductions in overall alarm counts, and even on our intervention unit, alarm burden remained quite high. Hospitals should consider monitoring policies that address alarms from all types of continuous monitors.
The authors acknowledge the work of Caitlin Clohessy and Alero Olomajeye in acquisition of data for this project.
Dr Schondelmeyer conceptualized and designed the study, and drafted the initial manuscript; Drs Brady and Simmons contributed to design of the study and interpretation of data, and critically revised the manuscript for important intellectual content; Dr Sucharew and Ms Huang contributed to the design of the study, performed the primary analysis, and critically revised the manuscript for important intellectual content; Ms Hofacer contributed to the design of the study and the acquisition and management of data, and critically revised the manuscript for important intellectual content; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FINANCIAL DISCLOSURE: Dr Brady is supported by the Agency for Healthcare Research and Quality under award K08HS023827. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality.
FUNDING: This study was funded by the Arnold W. Strauss Fellow Grant, Cincinnati Children’s Hospital Medical Center.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Shay DK,
- Holman RC,
- Roosevelt GE,
- Clarke MJ,
- Anderson LJ
- Ralston SL,
- Lieberthal AS,
- Meissner HC,
- et al
- Schondelmeyer AC,
- Simmons JM,
- Statile AM,
- et al
- ↵The Joint Commission. Sentinel Event Alert Issue 50: medical device alarm safety in hospitals. April 8, 2013. Available at: www.jointcommission.org/sea_issue_50. Accessed March 23, 2015
- Welch J
- Rheineck-Leyssius AT,
- Kalkman CJ
- Rheineck-Leyssius AT,
- Kalkman CJ
- Schondelmeyer AC,
- Goel VV,
- Blake N,
- et al
- Hines JZ,
- Sewell JL,
- Sehgal NL,
- Moriates C,
- Horton CK,
- Chen AH
- Karnik A,
- Bonafide CP
- Dandoy CE,
- Davies SM,
- Flesch L,
- et al
- Copyright © 2016 by the American Academy of Pediatrics