Article Text

Download PDFPDF

Under-recognition of alarms in a neonatal intensive care unit
  1. Pablo E Brockmann1,2,
  2. Cornelia Wiechers1,
  3. Tobias Pantalitschka1,
  4. Jessica Diebold1,
  5. Jan Vagedes1,3,
  6. Christian F Poets1
  1. 1Department of Neonatology, University Children's Hospital, Tuebingen, Germany
  2. 2Department of Paediatrics, School of Medicine, Pontificia Universidad Catolica de Chile, Santiago, Chile
  3. 3Filderklinik, ACRIM Research Institute, Filderstadt, Germany
  1. Correspondence to Professor Christian F Poets, Department of Neonatology, University Children's Hospital, Calwerstr. 7, 72076 Tuebingen Germany; christian-f.poets{at}


Background Treatment decisions for apnoea of prematurity (AOP) are usually based on nursing staff's documentation of pulse oximeter and heart rate alarms.

Objective In an observational study, to compare the accuracy of oxygen saturation (SpO2) and heart rate alarm documentation, and the resulting interventions by nursing staff, with objectively registered events using polysomnographic and video recording.

Methods Data on 21 preterm neonates (12 male) with a diagnosis of AOP were analysed. Nursing staff's desaturation (<80% SpO2) and bradycardia (<80/min) alarm documentation was compared with events registered objectively using simultaneous polysomnography. Interventions by nursing staff were evaluated using 24 h video recordings and compared with their chart documentation. Nursing staff had been unaware that the polygraphic and video recordings would be used subsequently for this purpose.

Results Median (minimum–maximum) postnatal age was 15.5 (3–65) days. 968 SpO2 desaturation events and  415 bradycardias were documented by polysomnography. Nursing staff registered 23% of these desaturation events, and 60% of bradycardias (n=223, and n=133, respectively). Intraclass correlation coefficient (95% CI) between objectively measured desaturation events and those documented by nursing staff was 0.14 (−0.31 to 0.53); and for bradycardias 0.51 (0.11 to 0.78). 225 nursing staff interventions were registered on video, of which 87 (39%) were documented.

Conclusions The alarm documentation by neonatal intensive care unit staff does not appear to be sufficiently accurate to permit further understanding and treatment of AOP. It is unclear if the alarms missed here would have led to clinical consequences had they been documented.

  • Monitoring
  • Neonatology
  • Nursing Care
View Full Text

Statistics from

What is already known on this topic

  • With traditional neonatal intensive care unit monitoring, many alarms are not documented in nursing charts.

  • It is unclear whether this situation has changed with the introduction of artefact rejection methods used with modern pulse oximeters.

What this study adds

  • Nursing staff failed to document most bradycardia and desaturation events in these preterm neonates.

  • This was even true for events that resulted in an intervention or lasted >20 s.


Apnoea of prematurity (AOP) is present in almost every infant born at <29 weeks of gestational age.1 ,2 Monitoring for AOP involves the continuous measurement of arterial oxygen saturation via pulse oximetry (SpO2), breathing movements and heart rate.

Although AOP is a self-resolving condition, it leads to serious challenges in monitoring. No SpO2 or heart rate threshold or duration has yet been defined above which the risk of adverse neurodevelopment is increased.

Also, while treatment decisions often depend on nursing staff's documentation of desaturation and bradycardia alarms, the validity of this documentation has rarely been evaluated.3 ,4 By the 1980s, neonatal intensive care unit (NICU) staff were found to recognise only one-third of apnoea alarms occurring,4 but since then many innovations have taken place, including pulse oximetry.

Previous studies were unable to determine whether a discrepancy between alarm documentation via nursing charts and objectively registered alarms is associated with a failure to respond to alarms.3 ,4 The consequences of intermittent hypoxia in neonates with AOP remain unknown; their investigation is made difficult by the lack of an accurate identification of SpO2 desaturation events in the NICU. Here, we aimed to investigate (i) the accuracy of alarm documentation for desaturation events and bradycardias conducted by nursing staff compared with continuous polysomnographic recording, and (ii) the interventions performed by nursing staff after such alarms.



This study used data obtained during the investigation of the effects of different nasal continuous positive airway pressure (NCPAP) generators.5 Inborn neonates admitted to the NICU at Tuebingen University Hospital were included if (i) their gestational age at birth was <34 weeks, (ii) their postmenstrual age and body weight at time of study were ≤38 weeks and >1000 g, respectively, and (iii) they required NCPAP to treat AOP. Neonates with congenital abnormalities or any acute medical condition were excluded. All infants were receiving caffeine. They were studied for 24 h using polysomnography and infrared video recording. Nurses were asked to perform their standard care, including their alarm documentation; they were aware that a study was being carried out but had no knowledge of the nature of the investigation. Nurses were asked to keep a registration of feeding and intervention periods. Written informed parental consent was obtained. The study protocol had been approved by the ethics committee of Tuebingen University Hospital.

Polysomnographic and video recording

Recordings were conducted using a computerised polysomnographic system (Embla N7000 and Somnologica Studio 3.0; Embla, Broomfield, Colorado, USA). The following signals were recorded: chest and abdominal wall movements (inductance plethysmography; Embla), SpO2 and pulse waveform (Radical in 4 s averaging mode; Masimo, Irvine, California, USA), electrocardiography and beat-to-beat heart rate (Embla). Video recordings were captured with an infrared camera (Panasonic, Osaka, Japan) and total and artefact-free recording times determined. The latter were defined as all quiet resting periods minus nursing and feeding times. For this study, data on desaturation events and bradycardias were investigated manually. A desaturation event was defined as a fall in SpO2 to <80% for more than 1 s to identify all possible desaturation events.6–8

Desaturation events with a distorted pulse waveform signal within 7 s before their onset were considered artefactual and excluded (the 7 s corresponding to the oximeter's signal processing time). A bradycardia was defined as a fall in heart rate to ≤80 beats/min for more than one beat. To exclude spurious events caused by motion, bradycardias with a distorted electrocardiography signal immediately before their onset were discarded. To assess nursing and feeding periods objectively, and interventions by nursing staff, infrared video recordings were analysed visually and classified as (i) gentle or vigorous tactile stimulation, (ii) body position changes, (iii) aspiration of nasopharyngeal secretions, (iv) provision of supplemental oxygen and (v) changes in NCPAP settings.

Alarm documentation and interventions

The routine paper chart documentation of each infant's vital parameters (done at least hourly) and the monitor alarms (entered into the same bedside sheet) were carried out by trained NICU nurses; it was important that this documentation was made at the time, in order to coincide with the timing of the 24 h polysomnographic recording. Continuous SpO2 monitoring is routinely performed in our unit. Nurses registered all alarms due to desaturation events and bradycardias. The recognition of these alarms was based on the standard clinical monitoring in place in the NICU. Bradycardias were assessed using the same electrocardiographic electrodes as used for polysomnography. For the recognition of alarms due to oxygen desaturation events, a standard pulse oximetry module (Radical, Masimo) in a 10 s averaging mode was used. Interventions conducted in response to an alarm were also routinely documented by nursing staff using the same classification method mentioned above. All recordings were analysed after the study was finished, to keep the blinding to nurses’ documentation. To improve the coincidence between polysomnographic events and those registered by nursing staff, a time lag of up to 2 min was accepted.

Comparison of pulse oximetry data

SpO2 obtained by polysomnography used an averaging mode of 4 s, while that obtained during standard NICU monitoring used 10 s. Thus, we transformed the number of desaturation events using a previously validated transformation algorithm.9 ,10 As bradycardias were identified by both systems without any averaging, no conversion was necessary for these latter alarms.


Descriptive statistics as numbers and percentages, mean and SD, and median (minimum–maximum) were used to summarise the results. The comparison of objectively assessed SpO2 desaturation events and bradycardias with events registered by nursing staff was performed using scatterplots, linear R2 and paired t test. The interindividual variation between detection of SpO2 desaturation events and bradycardias using polysomnography, compared with nurses’ registry, was calculated using intraclass correlation coefficients with 95% CIs. A p value <0.05 was considered significant. Agreement was considered good if intraclass correlation coefficients were >0.7 and <0.9; and excellent if >0.9. All analyses were performed using statistical software (IBM SPSS, release V.18.0 for Windows; IBM, Chicago, USA).


Demographic and clinical characteristics are given in table 1. Median (minimum–maximum) postnatal age was 15.5 (3–65) days. Mean (SD) gestational age at birth and birth weight were 27.5 (1.8) weeks and 1018 (264.5) g, respectively. Artefact-free recording time was 19.9 (12.8–21.7) h. Based on polysomnography, we identified 2627 desaturation events and 415 bradycardias. After correcting for differences in averaging time between polysomnography and clinical monitoring, 968 desaturation events were determined polysomnographically. Forty-five (11%) and 20 (5%) bradycardias lasted for >10 or >20 s, respectively. For desaturation events, the corresponding figures were 141 (>10 s; 5%) and 110 (>20 s; 4%), respectively.

Table 1

Demographic characteristics of included subjects (n=21)

Nursing staff registered 223 desaturation events (77% fewer than the total converted polysomnography events), and 133 bradycardias (40% fewer). Nursing staff registered only 61 of the 141 desaturation events lasting >10 s (57% fewer), and 39 of the 110 events lasting >20 s (65% fewer). Nursing staff registered 13 of the 45 bradycardias that lasted >10 s (71% fewer). The corresponding figure for bradycardias >20 s was 7 out of 20 (65% fewer).

In 92 of the 356 above mentioned events (ie, 223 desaturation events and 133 bradycardias), the following potential causes were documented by nursing staff: (i) nasogastric tube insertion/replacement (n=35), (ii) transient problems with the NCPAP system (n=21), (iii) agitation of the neonate (n=14), (iv) changes in positioning (n=12) and (v) aspiration of secretions (n=10). In the remaining 264 events, no details on the intervention were documented by nursing staff. Data on the comparison between the identification of prolonged desaturation events and bradycardias by polysomnography and nursing staff's documentation are given in table 2. Scatterplots between desaturation events and bradycardia identification based on polysomnography and nursing staff's documentation are shown in figures 1 and 2, respectively.

Table 2

Comparison of prolonged events (>10/20 s) found in polysomnography and those documented by nursing staff

Figure 1

Scatterplot of SpO2 desaturation events identified by polysomnography (after mathematical conversion) and those documented by nursing staff. PSG, polysomnography.

Figure 2

Scatterplot of bradycardias identified by polysomnography and those documented by nursing staff. PSG, polysomnography.

The intraclass correlation coefficient (95% CI) between objectively documented events and those noted by nursing staff was 0.14 (−0.31 to 0.53), p=0.275 for desaturation events and 0.51 (0.11 to 0.78), p=0.008 for bradycardias.

A total of 225 interventions in response to an alarm were carried out by nursing staff and documented on video (ie, mean interventions of 0.44/h per neonate), compared with 87 (39%) documented in the corresponding nursing charts. Of the latter, 57 involved tactile stimulation, 23 supplemental oxygen administration and aspiration of nasopharyngeal secretions, five aspiration of nasopharyngeal secretions only and two body position changes.


This study confirms the inconsistency between alarm detection by nursing staff and objectively measured SpO2 desaturation events and bradycardias. It also provides new information on the failure of nursing staff to document the interventions actually performed in response to such alarms. While identification of bradycardias by nursing staff showed at least moderate correlation to events recorded objectively, identification of desaturation events showed no significant correlation at all.

Our findings are in line with earlier work: Southall et al4 performed 24 h recordings of heart rate and abdominal wall movements in 14 preterm neonates and showed that 67% of apnoeas >20 s were not recorded by nursing staff. However, it was not possible to decide whether this was due to a failure to recognise or to respond to alarms. Graff et al3 showed that nursing documentation failed to detect alarms in 11 of 61 neonates (18%) with prolonged cardiorespiratory events. Another study compared computer-based vital sign measurements with nurses’ observation in neonates, and showed that nurses registered significantly higher values for these parameters.11

We show that nursing staff documented only 39% of interventions actually carried out. In another study, nursing staff often did not respond directly to alarms, but, rather, used them as an additional source of information.12 Razi et al,13 however, showed that 36% of nursing staff's alarm registry showed no objective correlate.

The discrepancy between conducted and registered interventions may possibly be explained by (i) lack of time (ie, nursing staff's high work load in the NICU), (ii) the fact that not all alarms are considered clinically relevant, or (iii) nurses using other sources of information (eg, clinical assessment). We suggest that this, final, explanation is probably the most important reason for the non-recording of interventions and even for the failure to recognise alarms, as the nursing team involved in this study has previously been shown to be extremely competent in monitoring their patients.14

The duration of a desaturation or bradycardia event may also influence decision-making. It may be impossible to detect very short desaturation events (ie, those lasting 1–2 s) that are recorded on polysomnography, and their importance is also unknown. However, nursing staff also failed to detect most events of >10 or >20 s duration. We hypothesise that these longer bradycardias and desaturation events were noticed by nursing staff, but had not been considered clinically important. Sadly, written evidence of artefacts or clinical situations that might have justified such interpretation, was missing in the nursing staff's registry.

Our study has several limitations. We aimed to compare nursing staff's alarm documentation with an objective method.15 The averaging time of pulse oximeters used for polysomnography, however, is much shorter than that used with NICU monitors. We attempted to solve this problem by using a recently developed and validated formula to transform recordings performed with one averaging time into another one; this produced an approximation to the real number of alarms that would have been observed with regular NICU monitoring. Nursing staff were encouraged to document events as soon as possible; nonetheless, they might have documented some only beyond the 2 min time window provided.

The consequences of under-recognition of NICU monitor alarms remain unknown. Therefore, there is an urgent need for studies that assess if neonates presenting with intermittent hypoxaemia or bradycardias have an increased risk of neurodevelopmental impairment. However, such studies would need a prior improvement of alarm detection.


The alarm detection by NICU nursing staff does not appear to be sufficiently accurate to permit further understanding and treatment of AOP. There is an urgent need to develop algorithms that improve decision-making; this would allow future studies to investigate more fully the consequences of hypoxaemia and bradycardias.


View Abstract


  • Contributors PEB analysed the data and wrote a first draft of the manuscript; CW, JD and TP were involved in study design, performing the recordings and data analysis; JV was involved in data analysis, and CFP supervised the study and was involved in study design and manuscript preparation.

  • Competing interests None.

  • Ethics approval Ethics committee of Tübingen University Hospital.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.