Delirium Variability is Influenced by the Sound Environment (DEVISE Study): How Changes in the Intensive Care Unit soundscape affect delirium incidence

Abstract

Quantitative data on the sensory environment of intensive care unit (ICU) patients and its potential link to increased risk of delirium is limited. We examined whether higher average sound and light levels in ICU environments are associated with delirium incidence. Over 111 million sound and light measurements from 143 patient stays in the surgical and trauma ICUs were collected using Quietyme® (Neshkoro, Wisconsin) sensors from May to July 2018 and analyzed. Sensory data were grouped into time of day, then normalized against their ICU environments, with Confusion Assessment Method (CAM-ICU) scores measured each shift. We then performed logistic regression analysis, adjusting for possible confounding variables. Lower morning sound averages (8 am–12 pm) (OR = 0.835, 95% OR CI = [0.746, 0.934], p = 0.002) and higher daytime sound averages (12 pm–6 pm) (OR = 1.157, 95% OR CI = [1.036, 1.292], p = 0.011) were associated with an increased odds of delirium incidence, while nighttime sound averages (10 pm–8 am) (OR = 0.990, 95% OR CI = [0.804, 1.221], p = 0.928) and the ICU light environment did not show statistical significance. Our results suggest an association between the ICU soundscape and the odds of developing delirium. This creates a future paradigm for studies of the ICU soundscape and lightscape.

Introduction

Intensive Care Unit (ICU) delirium, defined as “an acute and fluctuating disturbance of consciousness and cognition,” affects between 20 and 80% of ICU patients [1]. Delirium is an independent risk factor for adverse outcomes in the ICU including increased mortality, ICU and hospital length of stay, and post-ICU cognitive dysfunction [1–3]. There are many risk factors that influence the development of ICU delirium that are not modifiable, such as age, severity of illness, and baseline cognitive status [1]. Thus, any prospective modifiable risk factors become even more important.

Noise is a pollutant, and prolonged exposure to it can have detrimental effects on health, including sleep disturbances and cardiovascular disease [4, 5]. Poor sleep quality and quantity may increase the incidence of delirium [6–8]. Light also has a key role in synchronizing the circadian rhythm, and ICU survivors have reported that it can disrupt sleep, although to a lesser extent than noise [9, 10]. The National Institute for Occupational Safety and Health (NIOSH) recommends a noise exposure level in the workplace of 85 dB (dB) or less [11]. The ICU is considered a workplace, not in a separate category, per NIOSH recommendations. However, more specific to the intensive care unit, the World Health Organization (WHO) and the Environmental Protection Agency (EPA) recommend that, in hospitals, sound levels should not surpass 40 dB and 45 dB respectively, which is comparable to normal conversation or light rain [4, 5]. The International Electrotechnical Commission (IEC) creates the standard of auditory medical alarms, but they do not provide further details that supersede those of the NIOSH, WHO, or EPA [12]. Numerous studies have reported noise levels in ICUs as high as 65–70 dB, comparable to a vacuum cleaner or hair dryer [13, 14].

There are no previous studies comparing sound level measurements and the incidence of ICU delirium in adult patients. A time series analysis in the Netherlands explored ICU sound and delirium and found that implementation of a nocturnal sound reduction initiative decreased delirium incidence, but did not measure real-time sound levels in a patient-specific manner to study their association with delirium [15]. Recently, Kawai et al. conducted a quality improvement project at an academic hospital pediatric ICU that successfully used a nonpharmacological delirium bundle to reduce nocturnal noise to an hourly nighttime average of 35 dB and promote better quality of sleep, using Quietyme® (Neshkoro, Wisconsin) noise sensors to measure sound levels in ICU rooms [16, 17]. However, they did not measure the effects of this sound reduction on delirium incidence in their pediatric ICU.

The primary goal of our exploratory study was to examine whether sound levels are independently associated with delirium incidence in the ICU. We hypothesized that higher average sound levels in ICU patient rooms are associated with increased incidence of delirium among ICU patients. The secondary goal was to evaluate ICU light levels. Our exploratory hypothesis was that higher light levels in ICU patient rooms are associated with increased incidence of delirium.

Material and methods

Study design

We conducted a prospective cohort study in 297 adult patients admitted to the Surgical ICU and the Trauma ICU of Vanderbilt University Medical Center (VUMC), a US academic quaternary care center, from May 2018 to July 2018. This study received VUMC IRB approval, though patients were not formally enrolled as the sensors did not interfere with patient care nor record protected health information.

Data collection

Sensor measurements

We used 37 Quietyme sensors to measure sound and light levels (per second) in adult ICU patient rooms in the Surgical and Trauma ICUs. Quietyme sensors have been used in research studies in 37 hospitals across the US, for practical noise reduction efforts in healthcare settings, and have a measurement range of 25 to 120 dB with an accuracy of ±1 dB [17]. Sound was measured in decibels, while light levels were measured in a relative fashion designed to detect on or off in a known space. Each sensor was placed 18 in. off the ground and plugged into the 120 V AC outlet closest to the patient’s head. To address variability in sensor measurements due to sensor placement, we added an offset to a patient’s sensor data so that their average sensor value for the interval of 1 am to 2 am was now equal to the calculated hourly average from the entire ICU in that hour, creating a set of ICU-normalized sensor data.

Confounding variables

The ages, sexes, length of stays, Diagnostic Related Group (DRG) weights, periodic Glasgow Coma Scale (GCS) scores, and periodic Richmond Agitation-Sedation Scale (RASS) scores of patients were also recorded [18, 19]. DRG weight refers to the level of resources involved in care and is designed to capture the acuity and complexity of the hospital admission - which is a reasonable surrogate for acuity of hospitalization [20].

Outcomes

Incidence of delirium in ICU patients was measured using Confusion Assessment Method (CAM-ICU) scores recorded by bedside nurses during each shift [21].

Statistical analysis

Generated temporal sound and light features

Python version 3.7.5 from the Python Software Foundation (Wilmington, Delaware) was used to analyze sensor measurements. First, overall sound and light averages were calculated for each patient stay. Then, to better capture the temporal characteristics of sound and light in the ICU, we generated nineteen further light or sound features for each patient stay, which are described below and conceptually depicted in Fig. 1. The different definitions for peaks and thresholds between sound and light attempted to account for differences in the distributions of sound and light data.

Fig. 1.

A sample sound signal over the course of a day is analyzed using the methods described. The day is divided into morning, daytime, evening, and nighttime periods, in which the average values are calculated. The signal’s peaks are labeled and counted, and the various periods of rapid changes are measured. Finally, threshold analysis is done by calculating the ratio of signal data that exceeds a certain threshold (shown). dB = decibels

Periodic averages

The average of collected sensor data across each patient stay was calculated during different intervals throughout the day: morning (8 am–12 pm), daytime (12–6 pm), evening (6–10 pm), and nighttime (10 pm–8 am). To compare data between different ICUs, we measured the difference between a patient’s ICU-normalized data and the average of all hourly averages for that ICU (referred to as the ‘Grand Average’) and used this measurement in logistic regression analysis. For light, the Grand Averages for the Surgical ICU and the Trauma ICU were 17.7 and 59.6 respectively; for sound they were 52.8 dB and 54.4 dB respectively.

Rapid changes

To capture large fluctuations in sound, we defined a concept of rapid change for sound as any rolling window of 60 s in which the maximum recorded sound exceeded 15 dB greater than the ICU’s Grand Average and the difference between the maximum and minimum recorded sounds in the window exceeded 10 dB (roughly corresponding to the volume at least doubling). Similarly, for light, rapid change was any window of 60 s in which the maximum recorded light signal exceeded 15 units greater than the ICU’s Grand Average and the difference between the maximum and minimum recorded light signals in the window exceeded 10 units. The average daily number of seconds in which a patient was in a window of rapid change for light, sound, or both sound and light signals simultaneously were recorded.

Peaks

To capture disturbances related to light or sound, we counted the average number of daily peaks of sound and light using the signal processing package from the Python library SciPy. Sound peaks needed to be at least 15 dB above the ICU’s Grand Average with at least one minute in between peaks, while light peaks needed to be above the ICU’s Grand Average with at least one minute in between peaks.

Threshold features

These features were created by calculating the ratio of signal data that exceeded a certain threshold. For sound, the thresholds used were 100%, 105%, and 110% of the ICU’s Grand Average. For light, the thresholds used were 100%, 125%, and 150% of the ICU’s Grand Average.

Inclusion criteria

Patients who stayed in an ICU for longer than one day, had recorded DRG weights, had at least 70% of sound and light data recorded for their stay, and had the necessary nighttime sensor data needed for data normalization were included in this analysis.

Statistical approach

The ages, sexes, length of stays, DRG weights, RASS scores, and GCS scores were compared between patients with and without any delirium during their stay using Welch’s t-test. Logistic regression analysis conducted with R (version 3.6.1) from the R Foundation for Statistical Computing (Vienna, Austria) was used to quantify the effects of sound and light on the odds that a patient will experience delirium at any point during their ICU stay. An initial logistic regression analysis was done that adjusted for the average sound and light levels along with potential confounding variables (patient age, sex, DRG weight, length of stay, and ICU unit). Subsequently, a more detailed logistic regression analysis that adjusted for all the nineteen generated sound and light features described in 2.3.1 along with potential confounding variables was performed. That analysis was followed-up with a sensitivity analysis that repeated the same logistic regression analysis after omitting all variables that were previously found to be non-statistically significant.

Results

Patient characteristics

There were 143 patients (44.8% with delirium) who met the criteria for inclusion, with approximately 111 million sensor measurements. Descriptive statistics of these patients along with their average, minimum, and maximum recorded RASS and GCS scores from their ICU stays are displayed in Table 1. 92 patients were from the Surgical ICU (41.3% with delirium) and 51 patients were from the Trauma ICU (51.0% with delirium).

Table 1.

Patient characteristics and descriptive statistics

Variables	Patients without Delirium (n=79)	Patients with Delirium (n=64)
Age (years)	56.6 [38.2, 62.9]	56.8 [38.0, 65.8]
Sex	Male: 57.0% (45/79) Female: 43.0% (34/79)	Male: 65.6% (42/64) Female: 34.4% (22/64)
Length of Stay (days)*	2.9 [2.0, 5.1]	5.0 [2.8, 8.9]
DRG Weight*	3.0 [2.2, 4.3]	5.0 [2.2, 8.7]
Max GCS Score	15 [15, 15]	15 [15, 15]
Avg GCS Score*	14.9 [13.4, 14.9]	12.7 [10.5, 13.9]
Min GCS Score*	14 [7, 14]	3 [3, 7.8]
Max RASS Score*	0 [0, 1]	1 [0, 2]
Avg RASS Score*	−0.1 [−0.3, 0.0]	−0.6 [−1.1, −0.2]
Min RASS Score*	−1 [−3, −1]	−4 [−5, −2.2]

Values are median [interquartile range] or percentage (counts)(* indicates p < .001 for Welch’s T-Test). DRG = Diagnostic Related Group, GCS = Glasgow Coma Scale, RASS = Richmond Agitation-Sedation Scale, Min = minimum, Max = maximum, Avg = average.

Daily ICU sound and light profiles

In both ICUs, the average sound level oscillated several times during the morning and daytime before a more noticeable increase around 7 PM that could reflect a nursing shift change and/or increased visitor presence (Fig. 2). The average light level remained relatively stable from approximately 8 AM to 7 PM.

Fig. 2.

Average sound and light levels over the course of a full day in the Surgical and Trauma ICUs. db = decibels, ICU = intensive care unit

Overall sound and light averages with delirium

After accounting for potential confounders such as age, DRG weights, sex, length of stay, and ICU unit, an adjusted odds ratio test indicated that both the overall average sound level and the overall average light level had no statistically significant effects on the odds of delirium incidence (Table 2). This then led to examination of the more granular, temporal relationships in sound and light exposure with regard to delirium incidence, as those nuanced differences in the data were lost when comparing the averages in broad strokes [22].

Table 2.

Overall average sound and light levels in the surgical and trauma ICU for patients with and without delirium

	Odds Ratio (OR)	95% OR Confidence Interval (CI)	P value	Value (Patients Without Delirium)	Value (Patients With Delirium)
Overall Avg. Sound Level (dB)	0.953	[0.885, 1.026]	0.315	Surgical ICU: 52.7 [52.4, 53.3] Trauma ICU: 54.3 [53.3, 55.4]	Surgical ICU: 52.8 [52.2, 53.6] Trauma ICU: 53.7 [53.6, 54.0]
Overall Avg. Light Level (relative intensity)	1.004	[0.998, 1.011]	0.211	Surgical ICU: 14.5 [9.7, 20.2] Trauma ICU: 55.0 [39.0, 68.8]	Surgical ICU: 14.5 [10.9, 22.7] Trauma ICU: 60.2 [47.1, 83.4]

Values are median [interquartile range]. ICU = intensive care unit, dB = decibels, Avg = average.

Association of Morning and Daytime Sound Averages with delirium

For both patients with and without delirium, the calculated values for the average sound and light levels during different time intervals across a full day, average daily time spent experiencing a period of rapid change for light or sound (or both simultaneously), average number of daily sound and light peaks, and proportion of sound and light that exceeded various thresholds are included in Table 3. After performing an adjusted odds ratio test to collectively examine the associations between these temporal sound and light features (along with potential confounding variables such as patient age, sex, DRG weight, length of stay, and ICU unit) and delirium, we found that lower morning sound averages (OR = 0.835, 95% OR CI = [0.746, 0.934], p = 0.002) and higher daytime sound averages (OR = 1.157, 95% OR CI = [1.036, 1.292], p = 0.011) were significantly associated with an increased odds of delirium incidence. Of the potential confounding variables, both the DRG weight (OR = 1.035, 95% OR CI = [1.013, 1.058], p = 0.002) and length of stay (OR = 1.030, 95% OR CI = [1.006, 1.056], p = 0.016) were significantly associated with increased odds of delirium incidence. No other sound and light features were found to have a statistically significant association with delirium incidence (Table 3).

Table 3.

Results of an adjusted odds ratio test for delirium incidence with temporal sound and light features and possible confounding variables

	Odds Ratio (OR)	95% OR Confidence Interval (CI)	P value	Value (Patients Without Delirium)	Value (Patients With Delirium)
Sound
Morning (8 am–12 pm) Sound Average (dB)	0.835	[0.746, 0.934]	0.002	Surgical ICU: 53.2 [52.5, 54.1] Trauma ICU: 54.6 [53.3, 56.2]	Surgical ICU: 53.0 [51.8, 54.6] Trauma ICU: 54.0 [53.6, 54.9]
Daytime (12 pm–6 pm) Sound Average (dB)	1.157	[1.036, 1.292]	0.011	Surgical ICU: 53.0 [52.3, 54.2] Trauma ICU: 54.6 [53.5, 56.3]	Surgical ICU: 53.1 [52.6, 54.5] Trauma ICU: 54.0 [53.8, 55.0]
Evening (6 pm–10 pm) Sound Average (dB)	1.002	[0.916, 1.097]	0.962	Surgical ICU: 53.3 [52.5, 54.1] Trauma ICU: 54.6 [53.9, 56.0]	Surgical ICU: 53.1 [52.5, 53.9] Trauma ICU: 54.1 [53.7, 54.5]
Nighttime (10 pm–8 am) Sound Average (dB)	0.990	[0.804, 1.221]	0.928	Surgical ICU: 52.1 [51.9, 52.6] Trauma ICU: 53.6 [53.3, 54.1]	Surgical ICU: 52.2 [51.9, 52.4] Trauma ICU: 53.3 [53.2, 53.5]
Average Daily Period of Rapid Change for Sound (sec)	1.000	[1.000, 1.000]	0.339	3594.0 [1644.9, 6642.0]	3604.3 [1715.9, 6213.6]
Average Number of Daily Sound Peaks	1.001	[0.991, 1.029]	0.304	49.7 [23.9, 89.4]	53.6 [25.7, 85.6]
Proportion of Sound>ICU Grand Average	0.454	[0.171, 1.211]	0.117	41.1% [31.4%, 54.4%]	38.9% [29.6%, 49.6%]
Proportion of Sound>105% of the ICU Grand Average	5.352	[0.908, 31.559]	0.066	16.4% [10.2%, 25.1%]	15.6% [12.0%, 21.1%]
Proportion of Sound>110% of the ICU Grand Average	0.163	[0.013, 2.063]	0.164	5.9% [3.7%, 13.1%]	6.8% [4.0%, 10.2%]
Light
Morning (8 am–12 pm) Light Average (relative intensity)	1.004	[0.997, 1.012]	0.268	Surgical ICU: 20.5 [13.6, 33.3] Trauma ICU: 76.8 [56.0, 118.1]	Surgical ICU: 23.9 [13.6, 35.2] Trauma ICU: 95.8 [59.7, 111.8]
Daytime (12 pm–6 pm) Light Average (relative intensity)	0.996	[0.986, 1.007]	0.470	Surgical ICU: 18.6 [11.8, 26.0] Trauma ICU: 83.5 [55.2, 107.2]	Surgical ICU: 17.6 [13.4, 31.6] Trauma ICU: 88.5 [68.9, 101.5]
Evening (6 pm–10 pm) Light Average (relative intensity)	1.000	[0.992, 1.008]	0.936	Surgical ICU: 14.3 [10.2, 22.8] Trauma ICU: 47.0 [32.6, 59.8]	Surgical ICU: 13.7 [10.2, 19.7] Trauma ICU: 48.7 [33.4, 59.3]
Nighttime (10 pm–8 am) Light Average (relative intensity)	1.000	[0.987, 1.013]	0.982	Surgical ICU: 7.0 [5.7, 9.1] Trauma ICU: 29.2 [23.7, 37.8]	Surgical ICU: 7.7 [6.5, 8.8] Trauma ICU: 33.0 [25.7, 38.3]
Average Daily Period of Rapid Change for Light (sec)	1.000	[1.000, 1.000]	0.620	7867.8 [3856.5, 12,401.3]	7702.5 [2433.9, 13,311.2]
Average Number of Daily Light Peaks	0.999	[0.997, 1.001]	0.461	233.1 [155.5, 316.6]	261.8 [146.9, 327.9]
Proportion of Light>ICU Grand Average	1.184	[0.276, 5.075]	0.820	30.8% [20.3%, 47.4%]	36.3% [23.2%, 52.2%]
Proportion of Light>125% of the ICU Grand Average	2.815	[0.305, 25.956]	0.363	21.8% [10.7%, 42.3%]	25.4% [15.0%, 46.3%]
Proportion of Light>150% of the ICU Grand Average	0.451	[0.051, 4.022]	0.477	16.0% [7.8%, 34.6%]	20.0% [10.1%, 34.9%]
Sound and Light
Average Daily Period of Rapid Change for Sound and Light (sec)	1.000	[1.000, 1.000]	0.432	2010.8 [589.4, 4240.1]	1541.4 [555.0, 4601.6]
Potential Confounding Variables
Age (years)	1.003	[0.998, 1.008]	0.206	56.6 [38.2, 62.9]	56.8 [38.0, 65.8]
Sex	0.980	[0.826, 1.162]	0.816	Male: 57.0% (45/79) Female: 43.0% (34/79)	Male: 65.6% (42/64) Female: 34.4% (22/64)
Length of Stay (days)	1.030	[1.006, 1.056]	0.016	2.9 [2.0, 5.1]	5.0 [2.8, 8.9]
DRG_Weight	1.035	[1.013, 1.058]	0.002	3.0 [2.2, 4.3]	5.0 [2.2, 8.7]
ICU Unit	1.034	[0.699, 1.529]	0.868	Surgical ICU: 68.4% (54/79) Trauma ICU: 31.6% (25/79)	Surgical ICU: 59.4% (38/64) Trauma ICU: 40.6% (26/64)

Bolded rows had statistical significance (p < 0.05). dB = decibels, ICU = intensive care unit, sec = seconds, DRG = Diagnostic Related Group.

A follow-up sensitivity analysis that included only the two statistically significant sound features (morning sound average and daytime sound average) along with the listed potential confounding variables in an adjusted odds ratio test also found that the morning sound average (OR = 0.863, 95% OR CI = [0.792, 0.942], p = 0.001) was significantly negatively associated with delirium incidence while the daytime sound average (OR = 1.141, 95% OR CI = [1.045, 1.247], p = 0.004) was significantly positively associated with delirium incidence.

Discussion

An important aim of this study was to thoroughly characterize the soundscape of the ICU in more detail than previously done using computational techniques, providing the opportunity to intervene in a more focused, educated manner. We also sought to create a paradigm for studying large amounts of sound and light data, moving beyond quality improvement projects into high-powered, big data analysis. Sound is more than just volume; acoustic features such as the amplitude envelope of a sound, or the shape of sound over time, including pitch, sharpness, and frequency, impact one’s perception of the sound and its influence [23]. Although our various novel, detailed sound features did not significantly impact the odds of delirium incidence, they describe a chaotic sound environment for all patients, delirium or not.

The overall average sound levels for the Surgical ICU and Trauma ICU were above the WHO and EPA recommendations of 40 and 45 dB respectively - consistent with previous ICU studies [13, 14]. Furthermore, patients with delirium in this sample had higher maximum RASS scores and lower minimum RASS scores, suggesting the presence of both hypoactive and hyperactive subtypes of delirium - which likely has effects on morbidity and mortality [24–27]. We also found that DRG weight and length of stay were significantly associated with increased odds of delirium incidence, consistent with previous research [1–3].

The only sound features that had statistically significant effects on the odds of delirium incidence were the morning sound average (negatively associated with delirium) and the daytime sound average (positively associated with delirium), meaning that those patients with lower morning (8 am–12 pm) sound averages and higher daytime (12 pm–6 pm) sound averages had an increased odds of delirium incidence. This suggests that an increase in sound from 8 am to 12 pm could be protective against delirium. One can imagine that care teams that are acutely aware of the importance of maintaining sleep/wake cycles in the ICU would attempt to awaken the patient in the morning and encourage patient participation in their care goals for the day - all of which would increase the morning sound average. Intensivists often comment on “opening the curtains” and other interventions to increase light in the room to begin the day. However, our results showed that the morning light average had no significant effects on the incidence of delirium. Meanwhile, the reason for the association between increased daytime sound average and delirium incidence is likely multifactorial, but it could be related to components of the ICU workflow. Patients who need, for example, additional procedures, replacement of central venous lines, blood cultures drawn, trips to and from the operating room, or increased clinical presence in the afternoon, may be more at risk for delirium. Notably, this analysis adjusted for illness severity by incorporating DRG weights, suggesting that it is the sound component from 12 pm–6 pm, and not just increased severity of illness itself, that is associated with increased odds of delirium incidence.

The overall sound averages, evening sound averages, and nighttime sound averages did not have statistically significant effects on the odds of delirium incidence. This is important because it highlights the principle that delirium prevention may not necessitate large adjustments; rather, delirium prevention may hinge on small, targeted modifications to clinical behavior and environment. This may be a result of efforts of the last decade to actively try to minimize delirium, especially with regard to sleep in critically ill patients. In the critical care ABCDEF bundle, a key action toward delirium prevention is better sleep hygiene to minimize sleep disruption [28]. While noise reduction in the ICU improves sleep quality and leads to better neuropsychological outcomes [29–31], our results indicate that interventions in the ICU soundscape should also be more targeted towards the 8 am–6 pm time frame.

The limitations of our study and the future direction are intimately tied. Quietyme is a basic class II sound meter measuring sound pressure levels. Further investigation will utilize a class I sound meter for higher precision and accuracy within a wider frequency range to create even more nuanced data [32]. Our exploratory study did not attempt to discriminate between different types of sounds, though the effects of abrasive artificial sounds such as alarms are likely very different from the effects of organic, harmonious sounds [23, 33]. While noise emitted from a patient is a potential confounder, the overwhelming acoustic spectra from the data collected was outside of the human speech frequency range and the spectral characteristics were not consistent with human speech, both of which would argue against this [34]. As the anticipated effect of this hypothesis-generating study appeared to be unknown, no a priori sample size estimates were performed, nor did prospective randomization appear to be appropriate. As such, these results should be used to guide future prospective, randomized interventions, appropriately designed and powered to demonstrate causality. We used the gold standard of CAM-ICU, but emerging research of utilizing CHART-DEL-ICU for retrospective chart delirium analyses may uncover a stronger link between sound and delirium [21, 35]. We chose confounding variables that are known and validated, but as with any observational study, there may have been confounding from unaccounted variables [1–3]. We chose DRG as a potential confounder, rather than other markers of illness severity (for example Charlson or Elixhauser scoring systems) as those are specifically designed to assess individual risk of mortality [36–40].

Medical systems could implement sound sensors in a more systematic fashion to address patient care. Specifically, future directions include exploring incidence of delirium and immediately preceding soundscape characteristics using a class I sound meter and influences of environmental or workflow modifications on the ICU soundscape [41, 42]. Medical system utilization of a refined class I sound meter could lead to the identification of other acoustic features that may be more modifiable than loudness, such as amplitude envelope, sharpness, roughness, attack, decay, and frequency. Current research on the effects of smart alarm systems on ICU delirium could be combined with future research on the ICU soundscape and delirium [41]. Auditory masking is the principle that tonal differences in harmonics between different sounds in the same space may make sounds indistinguishable from one another due to human sensory limitations; placement of additional sensors in the same room may give more information on auditory masking, which can contribute to alarm fatigue and have negative implications on patient care [43–45]. Architectural design and patient location within the ICU environment have been tied to hospital mortality for sicker patients, and future studies should investigate the relationship between the ICU soundscape and mortality [46].

Conclusions

Our study characterized the sound and light environments of the ICU to a degree not previously done. Our analysis indicated that the sound average in the morning (8 am–12 pm) is negatively associated with an increased odds of delirium incidence while the sound average in the daytime (12 pm–6 pm) is positively associated with an increased odds of delirium incidence. We found no such association between light and delirium incidence. Ultimately, these results are important first steps towards describing the relationship between sound and delirium and helping guide future targeted interventions to improve patient safety and outcomes.

Data availability

By request to the corresponding author.