Practice guidelines discourage continuous pulse oximetry (SpO2) monitoring of patients with bronchiolitis who are not receiving supplemental oxygen.1,2 Overuse of SpO2 monitoring in this patient population has been associated with increased length of stay, unnecessary oxygen therapy, and excess hospital costs, without measurable patient benefit.3-5 In spite of this evidence base and expert guidance, nearly half of the more than 100,000 infants admitted for bronchiolitis each year receive excess SpO2 monitoring.6,7
Bronchiolitis guidelines suggest that guideline-discordant SpO2 monitoring may result in excess alarms, which disrupt families’ sleep and engender alarm fatigue among staff.1 Pediatric nurses receive up to 155 alarms per monitored patient per day.8,9 Frequent alarms are associated with slower nurse response times10,11 and increased nurse subjective workload.12The rate of excess alarms occurring during guideline-discordant, continuously SpO2 monitored time, compared to the rate of alarms occurring during guideline-concordant (intermittently measured SpO2) time, has not been evaluated. The magnitude of this difference in alarm rates, if such a difference exists, will inform prioritization of guideline-discordant continuous SpO2 measurement de-implementation. The objective of this study was to quantify the alarm burden associated with excess SpO2 monitoring of bronchiolitis patients not receiving supplemental oxygen.
We retrospectively evaluated SpO2 monitoring patterns and alarm rates of children 0 to 24 months old admitted to a general pediatrics service at a tertiary care children’s hospital. We included patients who had a discharge diagnosis of bronchiolitis (International Classification of Diseases, Tenth Revision codes J45x, T17.2x, T17.3x, T17.4x, T17.5x, T17.8x, T17.9x, A37xx, J04x, J05x, J05.1x, J69.0x, J69.1x, J69.8x) between November 24, 2019, and January 21, 2020, the period of time during which alarm data and monitor data were concurrently available for analysis. In order to conservatively assure applicability of clinical practice guidelines, we excluded patients with discharge diagnoses that included other respiratory conditions (eg, reactive airway disease), patients with complex chronic conditions (CCC) as defined by the CCC version 2 classification system,13 and patients with intensive care unit (ICU) stays during the admission.
Flowsheet data detailing nursing respiratory assessments were extracted from the electronic health record (EHR) database (Clarity, Epic Systems). Using previously validated methodology,14 we identified minutes during which patients received supplemental oxygen or high-flow nasal cannula (supplemental oxygen) based on the documented fraction of inspired oxygen (FiO2), flow rate, and support devices. We then identified the final discontinuation of respiratory support during the hospital admission, and censored the 60 minutes after final discontinuation of supplemental oxygen based upon recent monitoring guidelines.2 Minutes up to an hour after supplemental oxygen discontinuation were classified as receiving supplemental oxygen and not included in our analysis. Only minutes between the end of the censored period and hospital discharge were used in the analysis. For patients who never received respiratory support during the admission, we censored the first 60 minutes of the admission and analyzed the remainder of their stay.
We used device-integrated, physiologic-monitor, vital sign data sent each minute from the General Electric monitor network to the EHR to identify minutes during which patients were connected to physiologic monitors and transmitting signals from SpO2 sensors. We extracted minute-level SpO2 data from the hospital clinical data warehouse (CDW). Minutes in which SpO2 data were present were classified as “monitored,” an approach previously validated using in-person observation.14
To categorize time as “not receiving supplemental oxygen and continuously monitored (guideline-discordant monitoring),” or “not receiving supplemental oxygen and not continuously monitored (guideline-concordant intermittent measurement),” we evaluated the percent of minutes within an hour during which the patient received SpO2 monitoring and applied an a priori conservative rule. Hours during which ≥90% of minutes had SpO2 monitoring data were classified as “continuously monitored.” Hours during which ≤10% of minutes had SpO2 monitoring data were classified as “intermittently measured.” Hours during which 11% to 89% of minutes included monitor data were excluded from further analysis. The number of continuously monitored hours was tabulated for each patient. The median number of continuously monitored hours was computed; results were stratified by prior receipt of respiratory support.
Minute-level monitor alarm counts (the absolute number of abnormal vital signs that triggered a monitor to alarm) were extracted from the CDW. Alarm counts were tabulated for each patient hour. For each patient, the alarm rate (total number of alarms divided by time) was computed for continuously monitored and intermittently measured time. Results were stratified by prior receipt of respiratory support.
The study was reviewed by the institutional review board and determined to meet exemption criteria.
Our cohort included 201 admissions by 197 unique patients (Table). We evaluated 4402 hours that occurred ≥60 minutes following final discontinuation of supplemental oxygen, the time period during which guidelines discourage routine use of continuous SpO2 monitoring. This represented a median of 19 hours (interquartile range [IQR], 14-25) per admission. We excluded 474 hours (11%) that could not be classified as either continuously or intermittently measured.
During time ≥60 minutes following discontinuation of supplemental oxygen, our cohort experienced 1537 hours of guideline-discordant continuous monitoring, a median of 6 hours (IQR, 3-12) per admission. Patients experienced a median of 12 hours (IQR, 5-17) of intermittent measurement. Among patients who received supplemental oxygen, 91% experienced guideline-discordant continuous SpO2 monitoring, as compared to 68% of patients who did not receive supplemental oxygen. Among those who received guideline-discordant continuous SpO2 monitoring, the duration of this monitoring did not differ significantly between those who had received supplemental oxygen during the admission and those who had not.
During classifiable time ≥60 minutes following discontinuation of supplemental oxygen, our cohort experienced 14,371 alarms; 77% (11,101) of these alarms were generated during periods of guideline-discordant continuous monitoring. The median hourly alarm rate during these periods of guideline-discordant continuous monitoring was 6.7 alarms per hour (IQR, 2.1-12.3), representing a median of 35 (IQR, 10-81) additional alarms per patient. During periods of guideline-concordant intermittent measurement, the median hourly alarm rate was 0.5 (IQR, 0.1-0.8), with a median of 5 (IQR, 1-13) alarms per patient.
Those who received supplemental oxygen earlier in the admission had higher alarm rates during continuously monitored time (7.3 per hour [IQR, 2.7-12.7]) than patients who had not received supplemental oxygen (3.3 per hour [IQR, 0.6-11.8]), likely reflecting clinical differences between these patient populations. The most frequent alarm type among continuously monitored patients who had previously received supplemental oxygen was “SpO2 low.”
Across 4402 patient hours, guideline-discordant continuous SpO2 monitoring of patients with bronchiolitis resulted in 11,101 alarms, at a rate of approximately 1 additional alarm every 9 minutes. Patients in our cohort received a median of 6 hours of guideline-discordant monitoring, which imposes a significant alarm burden that is potentially modifiable using targeted reduction strategies.15
Transient, self-resolved hypoxemia is a common feature of bronchiolitis and likely of little clinical consequence.16 Therefore, this rate of hypoxemia alarms is not unexpected. Though we evaluated only the period of time following final discontinuation of respiratory support, this finding is in keeping with the literature associating excess physiologic monitoring of patients with bronchiolitis with unnecessary oxygen therapy and increased length of stay,3-5 largely because clinicians may feel compelled to respond to hypoxemia alarms with either supplemental oxygen or longer monitoring.
Our findings must be contextualized in light of the limitations of our approach. We did not evaluate nurse workload associated with guideline-discordant continuous SpO2 monitoring. Prior work conducted by our lab has demonstrated that when nurses experience more than 40 alarms within a 2-hour period, their measured subjective workload increases to a degree associated with missing important tasks, threatening the quality and safety of the care they deliver.12,17 Given that nurses care for multiple patients, it is likely that the excess alarms introduced by guideline-discordant continuous monitoring contribute to increased nurse workload and alarm fatigue.
Similarly, we could not evaluate whether the alarms nurses experienced were actionable. Although our lab has previously reported that ≥99% of alarms occurring on non-ICU pediatric wards are nonactionable,10,11 it is possible that some of the alarms during guideline-discordant monitoring periods required action. However, it is unlikely that any life-sustaining actions were taken because (1) we only evaluated time >60 minutes after final discontinuation of supplemental oxygen, so by definition none of these alarms required treatment with supplemental oxygen, and (2) none of the patients we included received ICU care during their admission.
The avoidable alarm burden identified in our analysis suggests that eliminating continuous SpO2 monitoring overuse in bronchiolitis will likely reduce nurses’ workload and alarm fatigue in hospital settings that care for children with bronchiolitis.