Knowledge about the impact of sound intensity in ICU is very limited, and our study aims to analyze sleep and sound objectively. A high prevalence of noise complaints from ICU patients has been reported in several previous studies [12–14, 22, 23]. Poor sleep is considered a major concern in the ICU because of its potential interaction with other psychological and somatic diseases and its impact on rehabilitation [14, 22, 23].
Sleep may be disturbed by multiple factors in ICU patients: pain [24, 25], high temperature [26], lighting [27], stress [24, 25], metabolic functions [28], the impact of mechanical ventilation [24, 29], adverse effects of some medications [30], and also by noise [31, 32]. However, it is unclear which sources of sound intensity are linked to the pathogenesis of sleep disturbance in ICU [33].
Using our miniaturized multi-channel ambulatory recording device, we succeeded in recording appropriately the different parameters necessary to score sleep in 30-s intervals in all our subjects. This miniaturized system is better tolerated than other standardized and larger PSG systems previously used in ICU and the shorter electrodes reduce interference.
Overall this study showed a fairly normal TST in subjects aged 64.2 ± 13.6 years—with a median of 502.2 [283.2–718.9] min and with a median nocturnal sleep of 356.9 min [188.6–590.9] and median daytime sleep of 168.5 [142.5–243.3] This TST is quite similar to the natural level (around 360 min per night) found in a meta-analysis of 65 PSG studies including 3577 healthy subjects (ranging from 5 to 102 subjects, mean age 65 years old) [33]. Of note, this meta-analysis gave no reference to 24-h sleep, which has only been estimated subjectively by epidemiological surveys at around 440 min in the general adult population [33].
Twenty-four hour TST is arguably a better measurement to consider in ICU patients as they are on continuous bed rest (except when activated or mobilized by staff) and their sleep may not follow a mono-episodic profile. In our group as mentioned above, patients slept an average of 6 h at night, but also a median of 2.5 h during the day (168.5 min [142.5–243.3]). One previous study assessed day and night sleep over 24 h and found an altered circadian rhythm [23]. Assessing 24 h sleep is crucial to a better understanding of the impact of sleep on ICU prognosis as subjects clearly take daytime naps and promoting sleep in ICU patients may take advantage of this opportunity.
Sleep quality was disturbed more than sleep quantity in our patients. The median N3 percentage was 6.5% [0–23.6] compared to 26.1% [34], and median REM sleep 3.9% [0–10.1] vs. 20% [34]. This poor quality has been reported previously [10, 35, 36]. In our study, we observe a significant link between sleep changes and arousals and noise, suggesting a partial explanation for lack of N3 and REM sleep.
REM sleep represented only 4% of TST during the 24 h, which is lower than the usual rate in normal sleepers (15–20%) [3]. N3 sleep represented 9% of TST, also lower than the usual percentage in normal sleepers (15–20%) [2].
We used sleep analysis according to the American Academy Sleep of Medicine (AASM). Other studies have shown that AASM was often an insufficient monitor for sleep stages in critically ill patients. Therefore, a modified AASM has been proposed by Watson et al. which can be used in critically ill patients [37]. However, in our study, all the sleep stages in these ventilated patients could be pre-categorized as REM or non-REM, N1, N2, N3 and awake. No atypical patterns were observed, and absent signal was only 0.2%. One possible explanation for this finding could lie in the category of patients: the 11 patients included in the present study were highly selected. They were non-sedated, undergoing weaning and post-resolution of disease. In other words, they were long-term ventilated patients without a cute illness. They were ventilated and in the ICU, but not critically ill. The fact that these patients were not critically ill might be an explanation for the lack of atypical sleep patterns. Secondly, the ActiWave itself, with electrodes positioned close to the sleeping brain, may have delivered more accurate quality of sleep signals than previous studies.
Heightened sound intensity levels are considered to be a major sleep-disturbing factor, whatever the context. Sound levels in hospital should not exceed 35 dBA LAeq in areas where patients are being treated or observed, with a corresponding LAmax of 40 dBA [38].
Previous studies have already observed that sound is a significant sleep disturber in ICU [29, 33], but the authors [29, 33] reported that this was only responsible for <20, and 17% of awakenings. Our study shows that sound levels above 77 dBC are associated with awakenings 60% of the time during the night (Fig. 2b). The median noise level at night (70.2 dB [65.1–80.3]) is considerably greater than the WHO limit, and sound levels >59 dBC at night (10 pm–8 am) and >63 dBC during the day were significantly linked with awakenings in ICU patients. Interestingly, monitor alarms and mechanical ventilator alarms, integral to the ICU environment and used for patient safety, also significantly disturbed sleep. Currently, there are suggestions that “noise-alarms” be replaced with other emergency signals such as light at nursing stations. Our study only demonstrates the negative aspect of the “noise-alarm”. This study identifies that these ICU and sound levels are above the limit recommended by the WHO and result in a higher incidence of disturbance of sleep patterns. Apart from alarms, other sounds coming from people around are implicated in the discussion of sleep continuity and quality.
We acknowledge that there are limitations to our study. For example, the evidence is based on a single 24-h PSG recording. It is known that the first PSG night is more disturbed than regular sleep, and two nights are required for pharmaceutical trials. Also we did not take into account the light environment of ICU rooms. Light at night is also known to alter the biological clock cycle and to disturb sleep. We recognize that concentrating on light and sleep in ICU is also a major issue in ICU [27, 29]. In our study, we did not assess the circadian factors that may influence the biological clock. Further research could record wrist actigraphy data from patients over a longer period.
Another potential limitation is that our study was limited to a group of 11 subjects who did not represent all patients hospitalized in the ICU. They had not received sedative medication recently and were not critically ill. However, in this preliminary study we deliberately concentrated on patients with an average of 22.1 ± 18.5 days stay in ICU, avoiding patients in the immediate post-diagnosis emergency period. We believe that our results reinforce and extend previous understanding of the fact that poor sleep is a persistent problem in ICU patients and that this is partly due to the negative effects of the immediate environment. Improving the physical and sensory aspects of the ICU environment represents an important and still under-researched potential for improving the sleep quality, and therefore the rehabilitation potential, of vulnerable patients.