Reducing Night‐Time Sound Levels Through Bed‐Level Quiet ICU Bundle: A Randomised Counterbalanced Crossover Study in Adult Intensive Care

ABSTRACT

Background

Elevated night‐time noise in intensive care units (ICUs) may reduce alarm salience, increase cognitive workload and contribute to alarm desensitisation during bedside surveillance. Most evaluations of Quiet ICU interventions have used uncontrolled or quasi‐experimental designs, limiting causal inference.

Aim

To evaluate whether a structured bed‐level Quiet ICU bundle reduced nocturnal acoustic burden in an adult ICU while preserving alarm visibility for clinical surveillance.

Study Design

This prospective, randomised counterbalanced crossover study was conducted in an adult intensive care unit in Singapore. Eligible bed‐episodes were assigned to one of two 4‐night sequences: AABB or BBAA, where A represented usual care and B represented Quiet ICU. The Quiet ICU bundle included muting bedside ventilator and monitor alarm audio while maintaining central alarm visibility, alongside alarm‐hygiene measures and quieter communication. Acoustic monitoring was performed overnight from 10:00 PM to 6:00 AM. The primary outcome was equivalent continuous sound level (LAeq); secondary outcomes were maximum sound level (Lmax) and time above 80 dBA. Repeated bed‐night observations were analysed using linear mixed‐effects models.

Results

Sixty‐nine monitored bed‐episodes contributed 220 analysed nights. Compared with usual care, Quiet ICU reduced LAeq (adjusted mean difference −4.56 dBA, 95% CI −5.69 to −3.43; p < 0.001), Lmax (−7.77 dBA, 95% CI −9.47 to −6.06; p < 0.001) and time above 80 dBA (−2.43 min, 95% CI −3.63 to −1.23; p < 0.001). The LAeq reduction was likely noticeable in practice. Bedside alarm audio was reactivated on 2 of 113 Quiet ICU nights (1.8%). No safety incidents were identified.

Conclusions

A structured bed‐level Quiet ICU bundle reduced night‐time acoustic burden while preserving central alarm visibility. Effects on alarm response, staff workload and patient outcomes remain uncertain.

Relevance to Clinical Practice

In ICUs with reliable central monitoring and clear alarm‐governance processes, a Quiet ICU bundle may offer a feasible, low‐resource strategy to reduce bedside noise.

Study Registration

OSF Registries; DOI: 10.17605/OSF.IO/AM59F.

Impact Statements

• What is known about the topic

  • ○
    Intensive care units remain acoustically challenging environments, particularly at night, with alarms, equipment and routine workflow contributing substantially to bedside noise burden.
  • ○
    Excessive background noise may reduce alarm salience, impair signal discrimination and increase cognitive workload during bedside surveillance.
  • ○
    Many Quiet ICU and noise‐reduction evaluations have used before‐and‐after or quasi‐experimental designs, limiting causal inference.
• What this paper adds

  • ○
    A structured bed‐level Quiet ICU bundle was associated with lower measured night‐time acoustic burden, including a lower A‐weighted equivalent continuous sound level (LAeq), a lower maximum A‐weighted sound level (Lmax) and less time above 80 dBA.
  • ○
    The observed reduction in night‐time LAeq was of a magnitude likely to be noticeable in practice, distinguishing this intervention from studies reporting smaller statistically significant reductions that may have limited perceptual relevance.
  • ○
    These reductions were achieved without altering alarm‐generation algorithms or replacing monitoring systems, using a pragmatic workflow‐embedded approach with maintained central alarm visibility.
  • ○
    The findings support a measurable environmental effect relevant to nursing surveillance, while highlighting the need for future studies to evaluate direct clinical outcomes such as alarm response, workload, sleep and safety.

1. Introduction

Intensive care units (ICUs) are acoustically complex environments in which alarms, life‐support equipment, staff communication, bedside procedures and routine clinical workflow generate sustained sound exposure throughout the 24‐h cycle [1, 2, 3, 4]. At night, this soundscape remains dense despite reduced lighting and efforts to promote rest, and clinically important alarms must compete with background noise and concurrent bedside activity for clinicians' attention. In this setting, ICU noise is not merely an environmental comfort issue but a potential systems issue affecting bedside surveillance, communication, workflow and the delivery of safe care.

2. Background/Justification for the Study

Growing evidence suggests that ICU noise exposure may affect healthcare professionals as well as patients. Among staff, higher acoustic burden has been associated with impaired concentration, communication difficulty, interruption load and increased perceived workload [1, 5]. In alarm‐dense environments, excessive background noise may reduce alarm salience and increase the cognitive effort required to distinguish, interpret and prioritise auditory cues during bedside surveillance. This is especially relevant to nursing practice, where timely recognition and interpretation of alarms are part of continuous patient monitoring, care escalation and coordination with the wider multidisciplinary team. The immediate bedside acoustic environment may therefore influence clinicians' ability to detect and respond to clinically meaningful signals [6].

The ICU acoustic environment is also relevant to patient experience and recovery. Night‐time noise has been linked to sleep disruption, stress and impaired rest, while interventions such as earplugs, eye masks, circadian hygiene strategies and broader environmental modifications have shown potential to improve sleep‐related outcomes, although implementation in high‐acuity settings remains challenging [7, 8, 9]. In response, many ICUs have introduced ‘Quiet ICU’ initiatives that combine behavioural, environmental and educational strategies to reduce unnecessary noise exposure [2, 6, 10].

However, the current evidence base has important methodological limitations. Many evaluations of Quiet ICU interventions have used before‐and‐after or quasi‐experimental designs, which are vulnerable to temporal confounding, secular workflow changes and other contextual influences. In addition, much of the prior work has focused on unit‐level sound exposure rather than the immediate bedside environment, where alarm interpretation and clinical surveillance actually occur. A randomised counterbalanced crossover design offers a stronger methodological approach because it enables comparison of study conditions within the same clinical environment while reducing the influence of between‐bed and between‐period variability.

3. Aim and Hypothesis

This study examined whether a structured bed‐level Quiet ICU bundle could reduce night‐time acoustic burden in a mixed adult ICU. Using a prospective, randomised, counterbalanced crossover design, we hypothesised that, compared with usual care, Quiet ICU exposure would be associated with lower night‐time equivalent continuous sound levels (LAeq), lower maximum sound levels (Lmax) and fewer high‐intensity noise events above 80 dBA.

4. Design and Methods

4.1. Study Design

This study used a prospective, randomised, counterbalanced bed‐level crossover design to evaluate the effect of a structured Quiet ICU bundle on night‐time acoustic burden. Eligible ICU bed‐episodes were allocated in a 1:1 ratio to one of two crossover sequences, AABB or BBAA, where A represented usual care and B represented the Quiet ICU condition. Each condition was delivered for two consecutive nights, yielding a four‐night protocol for each enrolled bed‐episode. The prespecified intervention window for each study night was 10:00 PM to 6:00 AM.

A counterbalanced crossover design was chosen to reduce the influence of temporal variation in ICU activity, staffing, workflow and patient condition. Because the intervention was embedded within routine care, a formal washout period was not feasible. Potential sequence and period effects were therefore addressed through the AABB/BBAA design and prespecified analytic adjustment. The intervention was assigned at the bed‐episode level, whereas the primary analysis was performed at the bed‐night level with repeated observations clustered within bed‐episodes (Figure 1).

FIGURE 1.

Study flow and bed‐level AABB/BBAA crossover design.

4.2. Setting and Sample

The study was conducted in Singapore in a 34‐bed adult mixed medical–surgical intensive care unit at a tertiary academic hospital, with data collected between 1 January 2026 and 31 March 2026. Isolation rooms were excluded because room‐specific environmental controls could materially influence bedside acoustic measurements. The ICU used integrated bedside and central monitoring, with physiological and ventilator alarms generated at the bedside and displayed simultaneously at the central nursing station. During Quiet ICU periods, bedside monitor and ventilator alarm audio were muted while central alarm visibility was maintained.

Adult patients occupying participating ICU beds were screened by the multidisciplinary clinical team. Bed‐episodes were eligible if the anticipated ICU stay was at least four nights and were excluded if a short stay or isolation precautions were anticipated.

Eligible bed‐episodes were randomised in a 1:1 ratio to the AABB or BBAA sequence using a computer‐generated allocation schedule, with allocation concealed until implementation. Bed location and ventilation status were not used as stratification variables because both were determined by routine clinical assignment and patient condition. Condition changes occurred at 6:00 AM, aligned with ICU shift change and routine workflow transition.

Based on prior ICU acoustic studies reporting night‐time sound‐level variability of approximately 3 to 6 dBA, and assuming a conservative standard deviation of 6 dBA, 32 protocol‐complete bed‐episodes were estimated to provide 80% power to detect a 3 dBA difference in night‐time LAeq at a two‐sided alpha of 0.05. To allow for incomplete crossover completion under routine ICU conditions, recruitment of up to 69 bed‐episodes was planned.

4.3. Data Collection Tools and Methods

4.3.1. Intervention Delivery

The Quiet ICU bundle was designed to reduce night‐time bedside acoustic burden without altering alarm thresholds, alarm‐generation algorithms, hardware or escalation pathways. The principal intervention component was muting bedside monitor and ventilator alarm audio during Quiet ICU periods while maintaining continuous central alarm visibility at the nursing station. Additional elements included daily review and individualisation of alarm limits, reduction of duplicate alarm sources where feasible, quieter bedside communication, relocation of non‐essential conversations away from the bedside and use of visual cues to indicate designated quiet periods.

Bedside nursing teams continued standard monitoring and could immediately restore bedside alarm audio whenever clinically necessary. Nursing staff and respiratory therapists received brief orientation before implementation, and fidelity was monitored through routine supervisory checks documenting alarm reactivation, protocol deviations and operational concerns.

4.3.2. Acoustic Measurement

Continuous acoustic monitoring was performed using IEC Class 2 sound‐level meters positioned at the head of the bed at a standardised height and orientation approximating patient ear level. Devices were calibrated before deployment using a CENTER 326 acoustic calibrator with NIST‐traceable, ISO 17025:2017‐accredited calibration in accordance with IEC 942 Class 2 standards. The microphone was positioned 60 ± 10 cm from the patient's ear and directed towards the patient. Placement was standardised using a measuring guide and fixed floor markings (Figure S1).

Sound pressure levels were recorded at 7‐s intervals and aggregated into 1‐min epochs for analysis. Night‐time was prespecified as 10:00 PM to 6:00 AM, and data recorded outside this window were excluded from the primary analysis. Devices were checked daily for placement, battery status and data integrity. Acoustic monitoring measured sound pressure levels only and did not record speech or identifiable audio content.

4.3.3. Outcomes and Covariates

The primary outcome was night‐time equivalent continuous sound level (LAeq, dBA). Secondary outcomes were night‐time maximum sound level (Lmax, dBA) and time above 80 dBA (minutes per night). Because minutes above 80 dBA were expected to be right‐skewed with many zero values, a prespecified sensitivity analysis dichotomised this variable as any threshold exceedance (> 0 vs. 0 min).

Prespecified covariates were bed location, ventilation status, crossover sequence and study period. Bed location was included to account for structural differences in baseline acoustic exposure, ventilation status to account for greater alarm‐ and equipment‐related noise among mechanically ventilated patients and crossover sequence and study period to account for order and temporal effects. Bed locations were classified as high‐traffic or low‐traffic according to proximity to staff activity areas.

4.4. Data Analysis

Acoustic outcomes were analysed at the bed‐night level using linear mixed‐effects models with bed‐episode included as a random intercept to account for repeated observations within the same crossover episode. Fixed effects included study condition, crossover sequence and study period, with bed location and ventilation status entered as prespecified covariates. Adjusted mean differences with 95% confidence intervals were reported, and two‐sided p < 0.05 were considered statistically significant. Inclusion of incomplete crossover sequences in the primary analysis was prespecified.

Distributional characteristics of continuous outcomes were assessed using descriptive statistics, histograms and quantile–quantile plots. Model assumptions for linear mixed‐effects models were evaluated by visual inspection of residual plots to assess approximate normality, linearity and homoscedasticity. Because minutes above 80 dBA were sparse and right‐skewed, a prespecified sensitivity analysis dichotomised this outcome as any threshold exceedance and analysed it using logistic generalised estimating equations clustered by bed‐episode.

A secondary paired analysis was conducted in the protocol‐complete subset that completed the full four‐night crossover sequence. Within each bed‐episode, outcomes were averaged across the two usual‐care nights and the two Quiet ICU nights, and paired differences were analysed using paired t‐tests. For the outcome of minutes above 80 dBA, the prespecified sensitivity analysis dichotomised the variable as any threshold exceedance and analysed it using logistic generalised estimating equations clustered by bed‐episode. Exploratory interaction analyses examined whether intervention effects varied by study period, crossover sequence or bed location. All analyses were performed using Stata/SE 18.0 (StataCorp LLC, College Station, TX, USA).

4.5. Ethical and Institutional Approvals

This study was approved by the National Healthcare Group Domain Specific Review Board on 31‐12‐2025 (reference no. 2025/1210). Informed consent was waived because the study was considered minimal risk. Acoustic monitoring recorded sound pressure levels only and did not capture speech or identifiable patient information. No patient‐level clinical data were analysed.

4.6. Study Registration

The study protocol and analysis plan were registered on OSF Registries (DOI: 10.17605/OSF.IO/AM59F). The registration record, including the protocol and prespecified analysis plan, is available on the OSF registration site.

4.7. Declaration of Generative AI and AI‐Assisted Technologies in the Manuscript Preparation Process

During the preparation of this manuscript, the authors used OpenAI ChatGPT (GPT‐5.5 Thinking) for language refinement, manuscript organisation and editorial support. The tool was not used to generate research data, conduct statistical analyses, interpret study findings independently or make scientific decisions. All AI‐assisted outputs were reviewed, edited and verified by the authors. All authors take full responsibility for the content, accuracy, integrity and final submitted version of the manuscript.

5. Results

5.1. Study and Protocol‐Complete Sample

A total of 69 monitored ICU bed‐episodes were included during the study period. Of these, 37 (53.6%) completed the full 4‐night randomised AABB/BBAA crossover protocol, whereas 32 (46.4%) were discontinued early because of routine clinical events, including discharge, transfer, extubation, clinical deterioration or death. Among protocol‐complete episodes, 18 were assigned to AABB and 19 to BBAA.

The acoustic dataset comprised 1 721 163 individual sound measurements, including 854 064 during usual care and 867 099 during Quiet ICU periods. The prespecified night‐time window of 10:00 PM to 6:00 AM yielded 220 bed‐nights for the primary analysis, including 107 usual‐care nights and 113 Quiet ICU nights.

5.2. Episode‐Level Contextual Characteristics

Episode‐level characteristics of the 37 protocol‐complete episodes are summarised in Table 1. Respiratory‐related diagnoses were the most common admission category (19/37, 51.4%). Overall, 28 of 37 episodes (75.7%) involved invasive ventilation, including 14 of 18 episodes (77.8%) in the AABB sequence and 14 of 19 episodes (73.7%) in the BBAA sequence. Bed location was similarly distributed between sequences.

TABLE 1.

Episode‐level characteristics of completed randomised ICU bed‐episodes by crossover sequence.

Variables	Total completed episodes (n = 37)	AABB (n = 18)	BBAA (n = 19)	p
Primary admission diagnosis, n (%)
Respiratory‐related	19 (51.4)	10 (55.6)	9 (47.4)	0.618
Sepsis/shock	3 (8.1)	1 (5.6)	2 (10.5)	0.580
Neurological	3 (8.1)	0 (0.0)	3 (15.8)	0.079
Metabolic/renal	4 (10.8)	2 (11.1)	2 (10.5)	0.954
Cardiac arrest/cardiovascular	3 (8.1)	3 (16.7)	0 (0.0)	0.063
Other	5 (13.5)	2 (11.1)	3 (15.8)	0.677
Ventilation status, n (%)
Intubated	28 (75.7)	14 (77.8)	14 (73.7)	0.772
Bed location, n (%)
High‐traffic	17 (45.9)	9 (50.0)	8 (42.1)	0.630

Note: Values are presented as n (%). Percentages are column percentages. A total of 37 ICU bed‐episodes completed the full 4‐day randomised crossover protocol, including 18 assigned to the AABB sequence and 19 assigned to the BBAA sequence. Primary admission diagnoses were grouped into clinically relevant categories for descriptive presentation. AABB indicates usual care on days 1–2 followed by Quiet ICU on days 3–4; BBAA indicates Quiet ICU on days 1–2 followed by usual care on days 3–4.

5.3. Night‐Level Contextual Characteristics

The primary analysis included 220 bed‐nights, comprising 107 usual‐care nights and 113 Quiet ICU nights. Night‐level contextual characteristics are summarised in Table 2. Invasive ventilation was present in 127 of 220 bed‐nights (57.7%), including 61 of 107 usual‐care nights (57.0%) and 66 of 113 Quiet ICU nights (58.4%). Bed location, weekend distribution, protocol period, ICU occupancy and patient‐to‐nurse ratio were similar across study conditions.

TABLE 2.

Night‐level contextual characteristics of analysed bed‐nights by study condition.

Variables	Total (n = 220)	Usual care (A) (n = 107)	Quiet ICU (B) (n = 113)	p
Ventilation status, n (%)
Intubated	127 (57.7)	61 (57.0)	66 (58.4)	0.834
Bed location, n (%)
High‐traffic	116 (52.7)	56 (52.3)	60 (53.1)	0.881
Weekend night, n (%)
Yes	89 (40.5)	43 (40.2)	46 (40.7)	1.000
Protocol period, n (%)
Night 1	69 (31.4)	35 (32.7)	34 (30.1)	0.771
Night 2	65 (29.5)	32 (29.9)	33 (29.2)	1.000
Night 3	49 (22.3)	22 (20.6)	27 (23.9)	0.628
Night 4	37 (16.8)	18 (16.8)	19 (16.8)	1.000
ICU occupancy, median (IQR)	17 (13–24)	16 (13–25)	17 (13–24)	0.787
Patient‐to‐nurse ratio, median (IQR)	1.8 (1.4–2.5)	1.7 (1.4–2.5)	1.8 (1.3–2.5)	0.821

5.4. Night‐Time Acoustic Outcomes

All prespecified night‐time acoustic outcomes were lower during Quiet ICU than during usual care. Mean night‐time LAeq decreased from 59.57 dBA (SD 4.57) to 55.00 dBA (SD 4.59), mean Lmax from 78.97 dBA (SD 6.84) to 71.21 dBA (SD 5.90) and mean time above 80 dBA from 2.92 min (SD 5.20) to 0.50 min (SD 1.35; Figure 2).

FIGURE 2.

Comparison of night‐time acoustic outcomes. Box plots comparing night‐time equivalent continuous sound levels (LAeq) and maximum sound levels (Lmax) between usual‐care and Quiet ICU periods during the standardised 10:00 PM to 6:00 AM monitoring window. Boxes represent the interquartile range (25th–75th percentiles), with the horizontal line indicating the median. Whiskers extend to the most extreme non‐outlier values. LAeq = equivalent continuous sound level; Lmax = maximum sound level.

5.5. Safety Outcomes

Safety override metrics supported protocol adherence and operational feasibility. Bedside alarm audio was reactivated in 2 of 113 Quiet ICU nights (1.8%), once for planned maintenance (30 min) and once for system downtime (88 min). No missed critical alarms or adverse safety events were reported through institutional surveillance systems during the study period. Throughout Quiet ICU periods, 100% of alarms remained continuously visible at the central monitoring station, so the intervention functioned as a redundant surveillance architecture rather than a withdrawal of surveillance. No intervention‐related changes to nurse staffing, respiratory therapy coverage or standard clinical responsibilities occurred during monitored periods. These findings support operational feasibility in this setting, although the study was not designed or powered to definitively establish safety.

5.6. Mixed‐Effects Analysis

In bed‐night level mixed‐effects models with bed‐episode included as a random intercept, Quiet ICU exposure was associated with lower values across all prespecified acoustic outcomes (Table 3). Compared with usual care, Quiet ICU was associated with a 4.56 dBA reduction in night‐time LAeq (95% CI −5.69 to −3.43; p < 0.001), a 7.77 dBA reduction in night‐time Lmax (95% CI −9.47 to −6.06; p < 0.001) and a 2.43‐min reduction in time above 80 dBA (95% CI −3.63 to −1.23; p < 0.001; Figure 3). In the sensitivity analysis, Quiet ICU was also associated with lower odds of any threshold exceedance above 80 dBA (odds ratio 0.16; 95% CI 0.07 to 0.35; p < 0.001).

TABLE 3.

Descriptive and mixed‐effects model estimates for night‐time acoustic outcomes during usual‐care and Quiet ICU periods.

Outcome	Usual care (A), mean (SD)	Quiet ICU (B), mean (SD)	Raw mean difference (B − A), 95% CI	Adjusted mean difference, 95% CI	Wald z	p
Night‐time LAeq, dBA	59.57 (4.57)	55.00 (4.59)	−4.57 (−5.79 to −3.35)	−4.56 (−5.69 to −3.43)	−7.91	< 0.001*
Night‐time Lmax, dBA	78.97 (6.84)	71.21 (5.90)	−7.76 (−9.46 to −6.06)	−7.77 (−9.47 to −6.06)	−8.93	< 0.001*
Minutes > 80 dBA	2.92 (5.20)	0.50 (1.35)	−2.42 (−3.44 to −1.40)	−2.43 (−3.63 to −1.23)	−3.97	< 0.001*

Note: Values are mean (SD). Night‐time was defined as 10:00 PM to 6:00 AM. Raw mean differences and 95% confidence intervals represent unadjusted descriptive comparisons between Quiet ICU and usual‐care periods and do not account for clustering of repeated bed‐night observations. Adjusted mean differences, 95% confidence intervals, Wald z statistics and p values were estimated using night‐level linear mixed‐effects models with bed‐episode included as a random intercept. Models were adjusted for crossover sequence, study period, bed location and ventilation status. *p‐value < 0.05.

Abbreviations: LAeq = equivalent continuous sound level; Lmax = maximum sound level.

FIGURE 3.

Forest plot of mixed‐effects model estimates. Forest plot showing adjusted mean differences and 95% confidence intervals for the association between Quiet ICU exposure and night‐time LAeq, Lmax and minutes above 80 dBA. Negative estimates indicate lower acoustic burden during Quiet ICU periods. CI = confidence interval; dBA = decibels A‐weighted; LAeq = equivalent continuous sound level; Lmax = maximum sound level.

In an exploratory interaction analysis, the association between Quiet ICU exposure and night‐time LAeq did not differ significantly by bed location (treatment‐by‐bed‐location interaction p = 0.668; Table S1). Estimated reductions in LAeq were observed in both low‐traffic beds (−5.02 dBA, 95% CI −8.06 to −1.98; p = 0.001) and high‐traffic beds (−4.20 dBA, 95% CI −6.42 to −1.98; p < 0.001).

5.7. Within‐Episode Paired Analysis

Within the protocol‐complete subset, paired comparisons were consistent with the primary analysis (Table S2). Mean paired differences (B − A) were −4.60 dBA for LAeq (95% CI −6.82 to −2.38; p < 0.001), −8.33 dBA for Lmax (95% CI −11.46 to −5.20; p < 0.001) and −3.50 min for time above 80 dBA (95% CI −6.03 to −0.97; p < 0.001).

5.8. Carryover and Sequence Analyses

In exploratory sequence analyses of the 220 analysed bed‐nights, a statistically significant treatment‐by‐sequence interaction was observed for LAeq (p = 0.033), indicating that the magnitude of sound reduction varied by crossover order. In the AABB sequence, mean LAeq decreased from 60.68 dBA during usual‐care nights to 51.90 dBA during Quiet ICU nights, whereas in the BBAA sequence, it decreased from 63.01 to 54.79 dBA. In the later usual‐care phase of the BBAA sequence, LAeq rose to 63.99 and 62.02 dBA on Nights 3 and 4, indicating that quieter conditions were not sustained after return to usual care. Corresponding treatment‐by‐sequence interactions were not statistically significant for Lmax (p = 0.069) or minutes above 80 dBA (p = 0.928). Supplementary between‐sequence comparisons of phase‐specific LAeq were also not statistically significant for either the A phase (p = 0.290) or the B phase (p = 0.341) (Table S3). These sequence‐specific findings indicate that Quiet ICU was associated with lower LAeq in both crossover orders, but the magnitude of reduction differed by sequence. In the BBAA sequence, LAeq increased after return to usual care, indicating that lower sound levels were not sustained during the later usual‐care phase.

6. Discussion

6.1. Principal Findings

In this randomised counterbalanced crossover study, a structured bed‐level Quiet ICU bundle was associated with lower night‐time acoustic burden than usual care. Quiet ICU exposure was associated with lower equivalent continuous sound levels, lower peak sound levels and fewer high‐intensity noise events above 80 dBA. These findings were directionally consistent across descriptive, mixed‐effects and protocol‐complete paired analyses, supporting a measurable intervention effect on the overnight bedside acoustic environment.

6.2. Interpretation and Clinical Relevance

The observed reduction in night‐time LAeq is clinically and acoustically important because its magnitude was likely to be noticeable in the ICU environment, rather than representing only a statistically significant but practically small change. This is an important distinction from many noise‐reduction studies in which reductions below approximately 3 dBA may reach statistical significance but are less likely to be clearly perceived by patients or staff [2, 4, 11]. In the present study, the adjusted LAeq reduction of 4.56 dBA and Lmax reduction of 7.77 dBA suggest a meaningful reduction in both sustained background sound and peak sound exposure during the overnight period. Although the signal‐to‐noise ratio was not measured directly, a lower background acoustic burden may improve the contrast between clinically important cues and surrounding noise, potentially supporting alarm salience and auditory discrimination during overnight bedside surveillance [1, 3].

Noise perception is also important because staff experience of ICU sound is not determined by decibel level alone. At night, alarms and equipment sounds may be perceived as more intrusive because staff are required to maintain vigilance during periods of reduced lighting, lower ambient activity, circadian fatigue and potentially fewer immediate team interactions [5, 6, 10]. The meaning, urgency, unpredictability, frequency and controllability of sound may therefore influence whether noise is experienced as tolerable background activity or as disruptive sensory load [1, 3]. From a nursing practice perspective, reducing bedside acoustic clutter may help create more favourable surveillance conditions by decreasing non‐essential sensory competition within the auditory workspace, particularly during overnight monitoring, communication and escalation of care [3, 6]. However, because alarm response behaviour, clinician workload, communication performance and subjective noise perception were not directly measured, these findings should be interpreted as evidence of improved acoustic conditions rather than confirmed improvement in behavioural or clinical performance.

Importantly, these reductions were achieved without structural renovation, replacement of monitoring systems or modification of alarm thresholds or alarm‐generation algorithms. This suggests that pragmatic, workflow‐embedded strategies may improve bedside acoustic conditions while preserving alarm availability through existing central monitoring systems.

6.3. Implementation Considerations for Critical Care Practice

The intervention appeared feasible within this ICU context. No implementation‐related safety incidents were identified through routine operational review, and the need to restore bedside alarm audio was limited. However, these findings should not be interpreted as conclusive evidence of safety. Rather than removing surveillance, the intervention redistributed alarm notification from bedside audibility to continuous central visual monitoring supported by bedside clinical care.

This has practical implications for critical care nursing. A Quiet ICU strategy of this type is likely to be most suitable in units with reliable central alarm visibility, clearly defined escalation pathways and staff who are trained to maintain active bedside and central surveillance. Safe implementation would require explicit local governance, including clear criteria for muting and restoring bedside alarm audio, documentation of deviations, defined responsibility for central alarm monitoring and contingency plans for monitoring downtime or workflow disruption. Accordingly, this intervention should be understood as a structured surveillance redesign rather than a simple noise‐reduction measure.

6.4. Comparison With Previous Literature

These findings are consistent with prior ICU noise literature from different healthcare settings. In the United Kingdom, Tahvili et al. [12] reported persistently high ICU sound levels and identified bedside care activities, medication administration, suctioning, alarms, equipment, staff activity and workflow as important contributors to acoustic burden. In China, Wang et al. [3] similarly reported that environmental noise interfered with nurses' recognition of clinical alarms and that nuisance alarms could disrupt patient care and reduce trust in alarms. In Germany, Witek et al. [4] and Armbruster et al. [5] evaluated ICU noise‐management approaches and staff noise exposure, while Vreman et al. [2, 11] summarised and tested noise‐reduction interventions in European critical‐care settings. The present study extends this literature by using a randomised counterbalanced crossover design in Singapore and by focusing on the immediate bedside environment, where alarm interpretation, bedside communication and overnight surveillance occur.

A key difference between the present study and some previous ICU noise‐reduction evaluations is the magnitude of the observed change. While smaller reductions in sound level may be statistically significant, reductions below approximately 3 dBA may have limited perceptual relevance in a complex clinical soundscape [2, 4, 11]. The larger reduction observed in this study, therefore, strengthens the practical relevance of the finding and suggests that the intervention altered the bedside acoustic environment in a way that staff or patients may plausibly notice during night‐time care. This perceptual relevance is particularly important in ICUs, where alarm audibility, staff communication and sustained vigilance occur within a dense and unpredictable sound environment [1, 3, 5, 6].

6.5. Methodological Considerations

A key strength of this study is the use of a randomised AABB/BBAA counterbalanced crossover design, which offers a more rigorous evaluation of a bedside acoustic intervention than uncontrolled or conventional before‐and‐after designs. Confidence in the overall finding is strengthened by consistency across multiple analytic approaches.

However, the significant treatment‐by‐sequence interaction for the primary outcome indicates that the magnitude of the LAeq reduction varied according to crossover order. Sequence‐specific results showed lower LAeq during Quiet ICU periods in both sequences, but no sustained reduction was observed after return to usual care in the BBAA sequence. This suggests that contextual or sequence‐related factors may have influenced the magnitude of the observed effect rather than indicating a stable carryover benefit.

Possible explanations include behavioural adaptation during early exposure to the intervention, temporal variation in workload or staffing, or regression towards usual baseline conditions after cessation of Quiet ICU exposure. Because a formal washout period was not feasible in this pragmatic design, residual carryover or other time‐dependent influences cannot be excluded. Although similar interactions were not observed for the secondary outcomes, the primary outcome should be interpreted with appropriate caution.

6.6. Limitations

Several limitations should be considered. First, this was a single‐centre study in an adult mixed medical‐surgical ICU with established central monitoring infrastructure, which may limit generalisability to units with different layouts, staffing models, alarm systems or monitoring capabilities. Second, the study evaluated acoustic outcomes only and did not directly measure alarm response, missed alarms, clinician workload, communication, subjective noise perception, sleep, delirium or other downstream staff‐related and patient‐centred outcomes. In particular, the study cannot determine whether the measured reduction in sound level was perceived by nurses, respiratory therapists, physicians, patients or families as less disruptive, less stressful or more supportive of overnight surveillance. The findings therefore support an effect on the measured bedside acoustic environment rather than a demonstrated behavioural, safety or clinical effect.

Third, not all monitored bed‐episodes completed the full 4‐night crossover sequence due to routine clinical events. Although incomplete observations were retained in the mixed‐effects analysis and the paired analysis showed directionally consistent findings, incomplete crossover completion may still have influenced the analysed sample. In addition, the significant treatment‐by‐sequence interaction indicates that sequence‐related effects cannot be fully excluded. Because the intervention was behavioural and embedded within routine care, washout was not feasible and contextual or sequence‐related influences may have contributed to the observed pattern.

Fourth, time above 80 dBA was sparse and right‐skewed, particularly during Quiet ICU periods. Although the prespecified sensitivity analysis supported the same directional finding, future studies with larger samples should consider modelling approaches better suited to zero‐inflated outcomes. Finally, acoustic measurements were obtained from a standardised fixed position near the patient's ear level and may not have captured the full spatial variability of bedside sound exposure. In addition, the study did not directly assess how reduced bedside alarm audibility affected perception or response to alarms in practice.

6.7. Recommendations or Implications for Practice and/or Further Research

A structured bed‐level Quiet ICU bundle may offer a feasible workflow‐embedded approach to reducing night‐time bedside acoustic burden in ICUs with reliable central monitoring and clearly defined alarm governance. Further investigations are warranted to assess whether these reductions in acoustics contribute to improved alarm recognition, decreased staff workload, enhanced patient sleep quality and safer overnight care. The exploratory analysis did not demonstrate a significant effect variation based on bed location, thus highlighting bed‐location prioritisation as an area for subsequent research.

7. Conclusion

In this randomised counterbalanced crossover study, a structured bed‐level Quiet ICU bundle was associated with lower objectively measured night‐time acoustic burden while preserving central alarm visibility. These findings support an objectively measured and likely perceptible improvement in bedside acoustic conditions relevant to critical care surveillance, but they do not establish improved alarm response, reduced workload or better patient outcomes. Further evaluation is needed to determine whether these acoustic improvements translate into measurable benefits for nursing practice, patient experience and safety across ICU settings.

Author Contributions

Wei Jun Dan Ong: conceptualisation, methodology, investigation, formal analysis, data curation, visualisation, project administration, writing – original draft, writing – review and editing. Adrian Ujin Yap: methodology, formal analysis, supervision, writing – review and editing. Woon Hean Keenan Chong: conceptualisation, methodology, supervision, writing – review and editing. Lawrence Ace Azul: investigation, resources, project administration. Patricia Leong: resources, project administration. Faheem Ahmed Khan: conceptualisation, methodology, supervision, writing – review and editing.

Funding

The authors have nothing to report.

Ethics Statement

This study was approved by the National Healthcare Group Domain Specific Review Board (2025/1210) on 31‐12‐2025.

Consent

Informed consent was waived because the study was considered minimal risk. Acoustic monitoring recorded sound pressure levels only and did not capture speech or identifiable patient information. No patient‐level clinical data were analysed.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: ICU monitoring environment, central alarm visibility and standardised bedside acoustic measurement set‐up.

Table S1: Exploratory subgroup analysis of Quiet ICU effect on night‐time LAeq by bed location.

Table S2: Within‐episode paired comparison of night‐time acoustic outcomes between A nights and B nights in the protocol‐complete subset.

Table S3: Sequence‐specific descriptive night‐time LAeq values by crossover order and study period.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.