Sources of Sound Exposure in Pediatric Critical Care

Abstract

Background:

Pediatric intensive care unit (PICU) sound is frequently above recommended levels. Few researchers have identified sound sources contributing to high levels.

Objectives:

Identify sources of PICU sound exposure during day (7:00–18:59) and night (19:00–6:59) shift, times of high (decibels [dB]≥45) and low (dB<45) sound levels, and during sound peaks (≥70 dB).

Methods:

Secondary analysis of continuous bedside video and dosimeter data (n=220.7hrs). Data were uploaded to Noldus Observer XT^® and time synchronized. A reliable coding scheme developed to identify sound sources in the adult ICU was modified for pediatrics. Sound sources (e.g., clinician/family/child vocalization, medical equipment) were identified via instantaneous sampling at each minute of recording. Proportions of sampling points with each sound source were compared between times of high and low sound, during day and night shift, and during sound peaks.

Results:

Overall, family vocalizations (38% of observation time, n=83.9hrs), clinician vocalizations (32%, n=70.6hrs), and child non-verbal vocalizations (29.4%, n=64.9hrs) were main human sound sources. Media (57.7%, n=127.3hrs), general activity (40.7%, n=89.8hrs), and medical equipment (31.3%, n=69.1hrs) were main environmental sound sources. Media sounds occurred in over half of video hours. Child non-verbal (71.6%, n=10.2hrs) and family vocalizations (63.2%, n=9hrs) were highly prevalent during sound peaks. Although there was decreased nighttime sound, general activity (32.1%, n=33.2hrs), clinician vocalizations (22.5%, n=23.3hrs), and medical equipment (20.6, n=21.3hrs) were prevalent during night shift.

Conclusions:

Clinicians should partner with families to limit nighttime PICU noise pollution. Large-scale studies using this highly reliable coding scheme are needed to understand the PICU sound environment.

Introduction

Sleep is crucial for healing during critical illness, however the pediatric intensive care unit (PICU) is not conducive to restorative sleep.^1,2 Sound exposure in the PICU is consistently above recommended levels.^3,4 The Environmental Protection Agency recommends hospital sound levels below 45 decibels (dB),⁵ while the World Health Organization (WHO) recommends levels less than 40 dB in hospital hallways, 35 dB at the bedside, and 30 dB at night.⁶ In contrast, children in the PICU experience an average of 115 minutes per day of sound levels greater than 100 dB,⁴ with average sound levels ranging from 46 dB to 79.5 dB.^4,7 Sleep disturbance can occur at levels above 30 dB, and the EPA and WHO recommend hearing protection for sustained exposure to sound levels greater than 85 dB.^5,6

Parents frequently report that the PICU is too loud to allow their children to rest.⁸ Clinicians and parents identify medical equipment, monitor alarms, and clinical staff conversations as main contributors to high PICU sound levels.^4,9,10 Sound levels associated with mechanical ventilation, monitor alarms, and staff conversations have been recorded at 71, 81, and 91 dB, respectively.¹¹ However, few researchers have combined measurement of PICU sound with continuous identification of the various sources contributing to sound levels. Recently, Naef et al.¹² created a protocol for measurement of sound levels and sound sources in the adult ICU, which provides a framework for identifying sources of high sound levels in the PICU. Identification of the sources of high PICU sound levels can help clinicians and researchers provide targeted interventions to improve sleep quality in the PICU.

The purpose of this observational study was to identify sources of PICU sound exposure during day (i.e., 7:00–18:59) and night (i.e., 19:00–6:59) shift, times of high (i.e., dB≥45) and low (i.e., dB<45) sound levels, and during sound peaks (i.e., ≥70 dB).

Methods

The Strengthening the Reporting of Observational Studies in Epidemiology guidelines were used as a framework for reporting study methods and results.¹³ See Kalvas et al.¹⁴ for full protocol details.

Study Design

This is a secondary analysis of continuous bedside video and dosimeter data (n=220.7 hours) collected between December 2020 and July 2021 during an observational pilot study exploring associations among modifiable PICU environmental exposures (light and sound levels, caregiving patterns), sleep disruption, and delirium.¹⁴ These data were used to identify sources of PICU sound exposure in the present study.

Setting and Participants

After obtaining Institutional Review Board approval (IRB #00000034), data were collected from a convenience sample of 12 critically ill children 1 to 4 years of age admitted to three PICUs (i.e., medical, surgical, cardiothoracic) at a large academic children’s hospital in the Midwestern United States. See Kalvas et al.¹⁴ for inclusion/exclusion criteria. Average age was 20.1 (SD=9) months. Most participants (83.3%, n=10) were admitted for respiratory failure. Children had a low severity of illness, with a median Pediatric Risk of Mortality III¹⁵ score of 1 (IQR 0, 4.3) and mean PICU length of stay of 41.4 (SD=25.3) hours. All children were hospitalized in private rooms. Enrollment occurred within 48 hours of PICU admission. All parents opted-in to use of their child’s research data for future secondary analyses.

Measures

Video and dosimeter recordings were initiated at enrollment and continued until PICU discharge, with signs posted inside and outside the hospital room as reminders that recording was ongoing. Equipment position and power supply were checked every 12 hours by a study team member. At conclusion of the observation period, data were downloaded onto a secure research drive and then deleted from the research equipment.

Sound Levels

Sound levels were measured with an Etymotic Research Dosimeter, Model ER-200DW7 (Etymotic Research, Inc., Elk Grove Village, IL) attached to the head of the hospital crib/bed. Sound levels were measured in A-weighted dB (dBA) in 0.22-second intervals, then a continuous equivalent sound level was calculated in 3.75-minute intervals. The accuracy of this dosimeter is ±2.5 dBA.

Sound Source

A Sony Handycam^® video recorder (Sony, New York, NY) was attached to the top of the foot of the hospital crib/bed with the camera directed towards the child. Recordings were coded using the Noldus Observer XT^® system (Noldus Information Technology, Netherlands). A coding scheme was developed using an instantaneous sampling method to identify sources of sound exposure in 1-minute intervals. Given the 3.75-minute dosimeter measurement interval, a 1-minute instantaneous sampling interval was deemed adequate to identify the sound sources contributing to each dosimeter measurement.^16,17 At the end of each 1-minute interval, each potential source of sound was marked as present or not present in the previous 15 seconds of recording.

The coding scheme is based on a template developed by Naef et al.¹² to identify sources of sound in the adult ICU through in-person observation. The first author (LBK) adapted the template for video observation in the pediatric population and broadly separated sounds into human and environmental sources (Table 1). Categories and codes were simplified because most verbal speech and many clinical activities occurred off camera. Vocalizations were identified by source (i.e., clinician, family, child), but the number of people contributing to conversation was not quantified. Given the young age of the children, their vocalizations were classified as verbal or non-verbal while adult (i.e., clinician, family) vocalizations were not differentiated as verbal or non-verbal. Patient care activities were separated into caregiving that involved physical touch with the child (i.e., patient touch care)¹⁴ and all other activities (i.e., general activity). Toys were added as a potential sound source. Initially the sound of PICU machines (e.g., intravenous medication pump, telemetry) and the alarms associated with them were collapsed into one category. During coding, it became clear that medical equipment was continuously creating sound. Therefore, the equipment code was redefined as only pertaining to acute, short-lasting (e.g., beeps, alarms) sounds, and videos were re-coded. If a sound source could not be identified, even with contextual cues (e.g., clinician explaining procedure to family), or was not included in the coding scheme, it was coded as Other.

Table 1.

Sound Source Instantaneous Sampling Codes

Code	Definition
Human Sources of Sound
Clinician Vocal	Verbal (e.g., talking) or non-verbal (e.g., laughing, coughing).
Family Vocal	Verbal (e.g., talking) or non-verbal (e.g., laughing, coughing).
Child Verbal	Verbal (e.g., talking).
Child Non-Verbal	Non-verbal (e.g., laughing, coughing, crying).
Child Activity	Child movements made without adult assistance (e.g., kicking).
Environmental Sources of Sound
Patient Touch Care	Care from any adult that involves physical contact with the child (e.g., chest physiotherapy, oral care).
General Activity	Adult activities that do not involve physical contact with the child (e.g., opening/closing doors or drawers, emptying trash can).
Medical Equipment	Acute, short-lasting sounds (e.g., beep, alarm, alert) from medical equipment (e.g., non-invasive ventilation, telemetry, intravenous medication pump, clinical communication devices). Does NOT include background, continuous sounds of medical equipment (e.g., air flow of non-invasive ventilation).
Bed	Use of hospital crib or bed (e.g., raising/lowering bed rails, object banging against rails).
Media	Non-medical electronic devices (e.g., television, mobile phone, tablet).
Toys	Toys (e.g., music, talking, beeping).
Other	Problem code for off screen sounds that cannot be identified, even with contextual clues, or are not included in coding scheme.

The coding scheme was developed iteratively through review of study videos and discussion with TMH, who has expertise in observational video coding. All video coding was performed by a research assistant trained in use of the Noldus Observer XT^® software and coding scheme and oriented to the PICU environment. Interrater reliability was established prior to independent coding. Reliability was measured with Cohen’s kappa (κ) statistic.^18,19 Given the high level of interrater reliability identified by Naef et al.¹² for the original coding scheme, a cut-off value of 0.70 was considered acceptable. To ensure ongoing agreement, LBK independently coded 13.7% (n=30.2 hours) of randomly selected video segments.²⁰ During these checks, LBK identified systemic discrepancies in coding and reviewed these with the research assistant to ensure ongoing accuracy in video coding.

Statistical Analysis

Video and dosimeter data were imported into the Noldus Observer^® system and time aligned for video coding and analysis. Videos were coded in 2-to-3-hour segments. The proportion of intervals in which each source of sound exposure was present was calculated. Videos were separated into day (7:00–18:59) and night (19:00–6:59) shift and proportions compared between periods of high (i.e., ≥45 dBA and low (i.e., <45 dBA) sound levels. Video segments which contained both day and night shift were assigned to the shift which comprised the majority of the video segment. Finally, proportions were calculated for intervals in which sound levels peaked, reaching a level of 70 dBA or higher.

Results

Overall, the most common human sources of sound in the PICU were family vocalizations (38% of observation time, n=83.9 hours), clinician vocalizations (32%, n=70.6 hours), and child non-verbal vocalizations (29.4%, n=64.9 hours; Figure 1; Table 2). The most common environmental sources of sound in the PICU were media (57.7%, n=127.3 hours), general activity (40.7%, n=89.8 hours), and medical equipment (31.3%, n=69.1 hours). Other sources of sound were identified in 1.5% (n=3.3 hours) of observation time, and commonly included traffic sounds from outside the hospital.

Figure 1:

Sources of Sound Exposure in Pediatric Critical Care

Table 2.

Percentage of Observation Time with Each Sound Source

Sound Source	Day Shifta		Night Shiftb		Sound Peak ≥ 70 dBA	Overall
	≥ 45 dBA	< 45 dBA	≥ 45 dBA	< 45 dBA
Clinician Vocal	50.3%	34%	39.9%	17.7%	55%	32%
Family Vocal	66.6%	35.5%	54.4%	20.8%	63.2%	38%
Child Verbal	12.5%	4.5%	9.3%	2.1%	12%	5.9%
Child Non-Verbal	61%	19%	56.9%	14%	71.6%	29.4%
Child Activity	20.7%	4.3%	18.5%	3.1%	20.3%	8.6%
Patient Touch Care	11.9%	2.4%	8.6%	2%	16.3%	4.8%
General Activity	57.6%	42.6%	52.8%	26.4%	53.7%	40.7%
Medical Equipment	46.5%	30.7%	41.3%	21.1%	51%	31.3%
Bed	5.5%	0.5%	7.1%	0.8%	8.2%	2.3%
Media	68.1%	68.3%	53.5%	43.2%	59.2%	57.7%
Toys	7.4%	2.6%	8.9%	6.4%	8.6%	5.5%

^a7:00 – 18:59

^b19:00 – 6:59

Day Shift

Video recordings included 116.1 (52.9%) day shift hours. Of these hours, 39.3% (n=45.7 hours) had sound levels of at least 45 dBA and 8.1% (n=9.4 hours) had sound peaks of at least 70 dBA. When sound levels were 45 dBA or higher, family vocalizations (66.6%, n=30.4 hours), child non-verbal vocalizations (61%, n=27.9 hours), and clinician vocalizations (50.3%, n=23 hours) were the main sources of human sound (Table 2). When sound levels were below 45 dBA, all human sound sources decreased in prevalence. However, child non-verbal vocalizations (19%, n=13.4 hours) and family vocalizations (35.5%, n=41.2 hours) decreased the most.

When sound levels were 45 dBA or higher, media (68.1% n=31.1 hours), general activity (57.6%, n=26.3 hours), and medical equipment (46.5%, n=21.3 hours) were the main sources of environmental sound. When sound levels were below 45 dBA, all environmental sound sources decreased in prevalence, except media which remained at a similarly high percentage of observation time (68.3%, n=48.1 hours). However, when sound levels were low, environmental sound sources did not decrease in prevalence as much as child non-verbal vocalizations and family vocalizations.

Night Shift

Video recordings included 103.4 (47.1%) night shift hours. Of these hours, 21.9% (n=22.7 hours) had sound levels of at least 45 dBA and 4.7% (n=4.9 hours) had sound peaks of at least 70 dBA. When sound levels were 45 dBA or higher, child non-verbal vocalizations (56.9%, n=12.9 hours), family vocalizations (54.4%, n=12.3 hours), and clinician vocalizations (39.9%, n=9.1 hours) were the main sources of human sound (Table 2). When sound levels were below 45 dBA, all human sound sources decreased in prevalence. However, child non-verbal vocalizations (14%, n=11.3 hours) and family vocalizations (20.8%, n=16.8 hours) decreased the most.

When sound levels were 45 dBA or higher, media (53.5%, n=12.1 hours), general activity (52.8%, n=12 hours), and medical equipment (41.3%, n=9.4 hours) were the main sources of environmental sound. When sound levels were below 45 dBA, all environmental sound sources decreased in prevalence, but not as much as child non-verbal vocalizations and family vocalizations.

Day vs. Night Shift

Compared to night shift, there was a higher percentage of hours with sound levels above 45 dBA (21.9% vs. 39.3%) and 70 dBA (4.7% vs. 8.1%) during the day shift. During the day, clinician vocalizations (40.4% vs. 22.5%), family vocalizations (47.7% vs. 28.2%), media (68.2% vs. 45.4%), and general activity (48.6% vs. 32.1%) were more frequent than at night. Although prevalence decreased at night, general activity (32.1%, n=33.2hrs), clinician vocalizations (22.5%, n=23.3hrs), and medical equipment (20.6, n=21.3hrs) sounds were common during night shift.

Sound Peaks

Video recordings included 14.3 (6.5%) hours of sound peaks above 70 dBA. During sound peaks, child non-verbal vocalizations (71.6%, n=10.2 hours), family vocalizations (63.2%, n=9 hours), and clinician vocalizations (55%, n=7.9 hours) were the main sources of human sound (Table 2). Media (59.2% n=8.5 hours), general activity (53.7%, n=7.7 hours), and medical equipment (51%, n=7.3 hours) were the main sources of environmental sound. During sound peaks, child non-verbal vocalizations (71.6% vs. 59.2%), clinician vocalizations (55% vs. 46.6%), medical equipment (51% vs. 44.1%), and patient touch care (16.3% vs. 10.7%) increased the most in prevalence compared to when sound levels were 45 dBA or greater.

Interrater Reliability

At the completion of video coding, average overall interrater reliability between LBK and the research assistant was near perfect (κ=0.98, SD=0.02). See Table 3 for agreement on individual codes, which ranged from substantial (e.g., media; κ=0.82, SD=0.40) to near perfect (e.g., toys, κ=1, SD=0).

Table 3.

Instantaneous Code Inter-Rater Reliability

Code	κ (SD)
Clinician Vocal	0.91 (0.13)
Family Vocal	0.96 (0.08)
Child Verbal	0.98 (0.04)
Child Non-Verbal	0.85 (0.23)
Child Activity	0.90 (0.13)
Patient Touch Care	0.83 (0.34)
General Activity	0.89 (0.12)
Medical Equipmenta	0.86 (0.12)
Bed	0.97 (0.07)
Media	0.82 (0.40)
Toys	1 (0)

^aCalculated for 28.3 video hours (12.5%).

Discussion

Similar to previous studies of PICU sound exposure,^3,4 sound levels rose frequently above the recommended maximum of 45 dBA. Regardless of shift or sound level, family vocalizations, clinician vocalizations, and child non-verbal vocalizations were the main human sources of PICU sound, while media, general activity, and medical equipment were the main environmental sources of PICU sound. Of all sound sources, child non-verbal vocalizations and family vocalizations decreased the most when sound levels were below 45 dBA, compared to when sound levels rose above 45 dBA. This suggests that families and children were main sources of PICU sound exposure in the current study. This is in contrast to previous studies in which clinicians and parents identified family as an infrequent contributor to high PICU sound levels.^4,9,10 Family vocalizations may be a therapeutic source of PICU sound, as this indicates family presence and potential interaction with the child for care or comfort. Child non-verbal vocalizations include coughing and crying, so the high frequency of these non-verbal cues may be related to the child’s illness or discomfort. Clinicians should empower families to provide comforting interventions to manage distressing symptoms (e.g., discomfort, delirium) in their children.^21,22

When sound levels peaked above 70 dBA, child non-verbal vocalizations, clinician vocalizations, medical equipment, and patient touch care increased in prevalence the most, compared to when sound levels were 45 dBA or higher. This cluster of sound sources suggests that child discomfort (e.g., crying), clinician vocal and physical interventions, and medical equipment alarms are contributing to the highest sound levels reached in the PICU. This is in alignment with previous studies in which clinicians and parents identified medical equipment, monitor alarms, and clinician conversation as main contributors to high PICU sound levels.^4,9,10

There was evidence of day/night sound cycling in the PICU, as there was a lower percentage of hours at 45 dBA or greater during night shift compared to day shift. Clinicians decreased levels of general activity (e.g., emptying trash, opening/shutting drawers), medical equipment alarms, vocalizations, and patient touch care during nighttime hours. However, general activity still occurred during one third of night shift hours, and medical equipment alarms and clinician vocalizations were present almost a quarter of night shift hours. Interventions to decrease high sound peaks and nighttime PICU sound include limiting clinician and family bedside conversation, grouping care activities, offering ear plugs, closing the child’s door if possible, and setting phones or pagers to silent,²³ as well as ensuring suitable alarm parameters and discontinuing continuous monitoring when appropriate.²⁴ When possible given the critical nature of illness treated in the PICU, these measures can result in significant reductions in nighttime PICU noise pollution.²³

Media (e.g., television, mobile phone, tablet) was the most common source of PICU sound, present in over half of all observed hours. However, the percentage of observation time with media sound was not highly different when sound levels were high or low. While screen media was highly pervasive in the present study, it may not have contributed to high sound levels. Similarly, clinicians and parents report media (e.g., television, radio) as an infrequent source of PICU sound.^4,9 Nevertheless, screen time is associated with sleep disturbance in children, including prolonged sleep latency and reduced sleep duration.²⁵ Researchers have identified a high level of screen time exposure in hospitalized children,^26–28 and this should be investigated in future research as a potential contributor to sleep disruption during pediatric critical illness. Clinicians should encourage families and children to limit screen time during nighttime hours.

Similar to Naef et al.,¹² this modified coding scheme for identifying sources of PICU sound through video observation had a high level of interrater reliability. This indicates transferability of the coding scheme from the adult to the pediatric population. Benefits of bedside video observation for sound source identification include decreased clinician, family, and child reactivity to observation, as the video camera was out of eyesight at the top of the bed, as well as the ability to re-watch video to ensure identification of all sound sources. Limitations of video recording include limited ability to identify sounds occurring off-screen. A simplified coding scheme was developed, compared to the scheme used by Naef and colleagues for in-person observation. A 1-minute instantaneous sampling interval was deemed adequate to sample each 3.75-minute dosimeter measurement interval, but may have missed identification of some sound sources. The timing of sound sources within each sampling point was not recorded, which limits ability to understand how sound sources are temporally related (e.g., family vocalizations frequently follow child non-verbal vocalizations). Placement of the dosimeter at the head of the hospital crib/bed did not account for times when children were out of bed (e.g., held in chair). Finally, data were collected at the bedside of 12 young children with low criticality at one large, academic children’s hospital. Large-scale studies with a wider age range of children with differing diagnoses and severity of illness at multiple institutions may better capture the PICU sound environment. Vehicle traffic outside the hospital was a newly identified source of sound that may apply to hospitals located in urban areas and should be included in future versions of the coding scheme. Future investigations should systematically consider whether identified sources of sound are modifiable and/or therapeutic to better direct targeted interventions to limit PICU noise pollution.

Conclusion

Although PICU sound exposure is known to be consistently above recommended levels, few researchers have combined measurement of PICU sound with continuous identification of the various sound sources contributing to sound levels. In this secondary analysis, bedside video and dosimeter data from an observational pilot study were used to identify sources of PICU sound exposure during day and night shift, times of high and low sound levels, and during sound peaks. Family, clinician, and child non-verbal vocalizations were the main human sources of PICU sound, while media, general activity, and medical equipment alarms were the main environmental sources of PICU sound. Media was a pervasive sound source, occurring in over half of video hours. Of all sound sources, child non-verbal vocalizations and family vocalizations decreased the most when sound levels were low, compared to when sound levels were high. During sound peaks, clinician vocalizations and medical equipment were highly prevalent. Although there was some evidence of day/night sound cycling, general activity still occurred during one third of night shift hours, and medical equipment alarms and clinician vocalizations were present a quarter of night shift hours. Clinicians should partner with families to engage in interventions to limit nighttime noise pollution in the PICU. Large-scale studies using this highly reliable coding scheme are needed to fully understand the PICU sound environment.