A total of 24 ICU patients were enrolled between March 1997 and February 1999. Two patients withdrew from the study prior to its initiation at the request from the patient's family members after the initial consent was given. Twenty-two patients completed the study for a total of 30 24-h periods (8 patients were studied for 48 h continuously and the remaining 14 patients were studied for 24 h). The study population was comprised of 12 males and 10 females with a mean age of 61 ± 16 yr (range 20–83), mean APACHE III score of 57 ± 28 (range 7–132), and a mean duration of ICU stay prior to the study of 18 ± 20 d (range 3–80). Twenty of the 22 patients were mechanically ventilated at the time of the study, and remained mechanically ventilated for the entire study period. Primary reasons for mechanical ventilation included pneumonia (7 patients), sepsis (5 patients), chronic obstructive pulmonary disease (COPD) exacerbation (3 patients), acute respiratory distress syndrome (ARDS) (3 patients), and myasthenia gravis (2 patients). Fourteen patients received no sedation during the study period. Of the remaining 8 patients, 4 patients received intermittent, as needed, intravenous lorezepam and 4 patients received intermittent, as needed, intravenous doses of fentanyl. No patients were on a combination of benzodiazepines and narcotics during the study period. No patients received tricyclic or other types of antidepressant medications during the study period. All patients had a Glasgow coma score of 14 or greater upon entrance into the study.

Seventeen (77.3%) of the 22 patients had scorable EEG data according to the criteria of Rechtschaffen and Kales (18). From this group, there were a total of 21 separate 24-h day/ night periods (13 patients with 24 h of scorable data and 4 patients with 48 h of scorable data). The other 5 patients (22.7%) demonstrated evidence of septic encephalopathy throughout the majority or all of the study period, and were therefore unable to be scored according to standard criteria of sleep versus wake (see section on sepsis and sleep). Of the 17 ICU patients with scorable sleep–wake polysomnograms, all demonstrated abnormal sleep architecture. Although the mean total sleep time per 24 h period was within the normal range (8.8 ± 5.0 h) (17) there were large individual variations in total sleep time (range 1.7 to 19.4 h).

There was a predominance of stage 1 sleep (mean 59 ± 33%) with decreased or absent stages 2 (mean 26 ± 28%), 3/4 (mean 9 ± 18%), and REM sleep (mean 6 ± 9%). Twelve of these 17 patients demonstrated no REM sleep. Of the 8 patients who were studied continuously for 48 h, 5 had scorable EEG data. There were no significant differences between study Day 1 and 2 with respect to mean total time spent in sleep stages 1, 3/4, and REM, day versus night total sleep times, or the overall arousal index. There were no significant differences (p > 0.05) between sexes or window status with respect to TST/24-h period in time in any sleep stage. There were no significant correlations (p > 0.05) between TST/24-h period or time in any sleep stage with age, duration of ICU stay, or APACHE III score.

All 17 patients with scorable sleep–wake stages demonstrated nonconsolidated sleep that was distributed throughout the 24-h day–night study period (see Figure 2). Fifty-seven percent ± 18% of the total sleep time occurred during the daytime (6:00 a.m.–10:00 p.m.) and 43 ± 18% of the TST occurred during the nighttime hours (10:00 p.m.–6:00 a.m.). The mean number of sleep periods per 24-h study period was 41 ± 28 (range 5–100). The mean length of each sleep bout was 15 ± 9 min (range 5.5–40 min). REM sleep, when it occurred, was equally distributed between day (50 ± 5%) and night (50 ± 5%). The overall arousal index (number of arousals per hour of sleep) was normal (mean 11.6 ± 5.0; range 4.6–21.2) (22). There were no significant differences (p > 0.05) between sexes or window status with respect to daytime versus nighttime sleep or the overall arousal index. Age, duration of ICU stay, and APACHE III score were not significantly correlated (p > 0.05) with daytime versus nighttime sleep or the overall arousal index.

Fig. 2. Schematic representation of the redistribution of sleep and wake in five subjects over the 24-h period. Black areas represent episodes of sleep and white areas represent wakefulness.

Download Figure | Download Powerpoint

[More] [Minimize]

The mean, mean maximum, and mean peak ICU environmental noise levels exceeded EPA recommendation during both the daytime and nighttime hours. There were no statistically significant differences (p > 0.05) in mean (59.1 ± 6.1 dB[A] versus 56.8 ± 4.9 dB[A]), mean maximum (68.5 ± 7.7 dB versus 64.6 ± 7.5 dB), or mean peak (85.9 ± 5.1 dB versus 82.8 ± 5.3 dB) noise levels between the day and night, respectively. There were no significant differences (p > 0.05) between mean (58.9 ± 6.0 dB versus 57.1 ± 5.2 dB), mean maximum (68.3 ± 7.5 dB versus 64.9 ± 7.6 dB), and mean peak (85.6 ± 5.0 dB versus 84.9 ± 4.8 dB) noise levels during periods of sleep and wakefulness.

Overall, 11.5% of the arousals from sleep for the entire population studied were secondary to environmental noise. Arousals related to environmental noise comprised an average of 11.5 ± 11.8% of the total arousals from sleep per subject. The mean arousal index specifically related to environmental noise was 1.9 ± 2.1. This was significantly less (p < 0.0001) than the mean spontaneous (nonnoise) arousal index of 9.6 ± 4.9. Overall, environmental noise was responsible for 17% of the awakenings from sleep. Awakenings related to environmental noise comprised an average of 26.2 ± 24.8% (range 0–75%) of the total awakenings from sleep per subject. There were no significant differences (p > 0.05) between sexes or window status with respect to the noise specific arousal index or awakenings secondary to noise. Age, duration of ICU stay, and APACHE III score were not significantly correlated (p > 0.05) with the noise specific arousal index or awakenings secondary to noise.

As stated earlier, 5 patients either developed sepsis and/or positive blood cultures during the study period (4 of 5) or were recovering from sepsis (1 of 5) during the study period. None of the patients was receiving continuous sedative medications immediately before (previous 24 h) or during the study period. All 5 patients demonstrated similar EEG patterns of a baseline of low-voltage mixed-frequency waves with intermittent and variable amounts of theta and delta waveform activity (see Figure 3). In the 4 patients without known sepsis prior to the study period, this EEG pattern appeared up to 8 h prior to these patients demonstrating clinical signs of sepsis (fever, hypotension). This EEG pattern was present both when the patients' eyes were open as well as when they were closed. For this reason, we were unable to define the patients' state of consciousness into definitive sleep or wake states, by current EEG criteria. These 5 patients did not demonstrate any evidence of clearly definable sleep throughout the study period. There was no evidence of spindles, K-complexes, or REM activity, all hallmarks or normal sleep.

Fig. 3. Polysomnographic representation of septic encephalopathy. This EEG pattern demonstrates a baseline of low-voltage mixed-frequency waves with intermittent theta and delta waveform activity.

Download Figure | Download Powerpoint

[More] [Minimize]

Our study design had several limitations, which should be reviewed. Our results are not generalizable to all ICU patient populations as the majority of the patients that we evaluated were nonsedated mechanically ventilated medical ICU patients. However, this is a very important patient population to study since many ICU patients are mechanically ventilated. There may also have been a selection bias as the patients were selected to participate based on their likelihood of remaining in the ICU for a continuous 48-h period and were therefore not randomly selected. We did not control for underlying disease state as we were interested in evaluating disorders of sleep and wake in a heterogeneous population of ICU patients. We excluded patients under heavy sedation in this initial study because we thought it would be difficult to classify a given patient's level of consciousness as wakefulness versus sleep. Also, sedatives themselves can affect sleep and/or the EEG potentially confounding our results (23). However, the majority of our patients were not treated with sedatives. We were unable to determine if all of the patients' sleep–wake cycles changed on a day-to-day basis or improved over time as we evaluated each patient only for 24–48 h and did not control for length of stay in the ICU. However, the abnormalities in sleep architecture were evident regardless of length of stay, suggesting that length of stay may not be an important contributor to sleep alterations in the ICU. Also, there were no significant differences in the total sleep time, time in the various sleep stages, or day versus night total sleep times between study Day 1 and 2 in the 5 patients with scorable EEG data who were studied for 48 continuous hours. This suggests that there is little day-to-day variation in sleep architecture.

Environmental noise clearly plays a role in disrupting sleep in medical ICU patients. The finding that environmental noise was not responsible for the majority of arousals and awakenings from sleep is contrary to the current literature, which considers noise to be the major etiologic factor responsible for sleep disruption in the ICU (5-11). We have objectively demonstrated that environmental noise was responsible for 11.5 and 17% of the arousals and awakenings from sleep, respectively. These findings confirm our previous research that demonstrated that although ICU patients subjectively experienced significantly poorer sleep quality in the ICU than at home, ICU noise was not perceived as the most disruptive environmental stimulus (15).

We believe our results are valid for several reasons. First, our technique of simultaneously monitoring environmental noise and polysomnography allowed for the objective measurement of effects of noise on sleep. Second, there were no significant differences in mean, maximum, or peak noise levels during periods of wake and sleep, so our results cannot be explained by differences in noise levels during sleep versus wake periods. Finally, it cannot be argued that our ICUs were quieter than other ICUs as the noise levels recorded in our ICUs appear to be comparable to those in other publications (7, 24).

The finding that environmental noise was not as disruptive to ICU patient sleep as previously described in the literature was not completely surprising. Previous studies show that there are wide variations in individual sensitivity to sounds during sleep and that the meaning of the sound is also critical (25, 26). Although studies on the acute effect of noise on sleep have consistently demonstrated a sleep disrupting effect, normal individuals rapidly adapt to the disruptive effects of environmental noise on sleep (27). Studies in normals have demonstrated that individuals habituate to sound by increasing their arousal threshold for noise over time, with some individuals being able to increase their arousal threshold for noise to more than 80 dB(A) (28-30). This finding is also consistent with the results of Aurell and Elmquist (3) who demonstrated that sleep architecture and distribution were significantly altered in 9 postsurgical ICU patients, despite a concerted effort by the staff to keep environmental disturbances (noise, interruption, and light) to a minimum.

From the standpoint of the ICU environment, we previously described that ICU patients subjectively perceived human interventions and diagnostic testing to be as disruptive to sleep as noise (15). In this predominantly mechanically ventilated group of patients, it is possible that sleep was disrupted by patient/ventilator dyssynchrony as well as from human interventions such as suctioning and the administration of respiratory treatments. This study was not designed to evaluate the effects of human interventions or other factors (light, noxious odors, etc.) on sleep–wake disturbances. Future research should focus on objectively determining the effects of these factors on sleep disturbances in the ICU population.

It was remarkable that the abnormal sleep–wake patterns and sleep disruptions attributable to noise demonstrated by this group of patients were not affected by patient age, sex, duration of ICU stay, or severity of illness. These findings confirm our previous research (15), which demonstrated that the patient characteristics of age, sex, and duration of stay were not associated with an ICU patient's poor subjective sleep quality. This indicates that all mechanically ventilated medical ICU patients are at risk for sleep disturbances. This also indicates that other factors, environmental (human interventions and light) and/or patient specific (medications, pain/anxiety, underlying disease, inflammatory mediators, and/or circadian rhythm disturbances), must be responsible for sleep–wake disturbances in the ICU.

Although the majority of the critical care literature (1-4) has suggested that ICU patients are sleep deprived, we have demonstrated that mechanically ventilated medical ICU patients are not necessarily quantitatively sleep deprived. Our patients demonstrated large variations in total sleep time per 24-h period, with a mean total sleep time per 24-h period of 8.8 ± 5.0 h. Many patients are selectively deprived of certain stages of sleep, but our patients were not totally sleep deprived for any given 24-h period.

Although ICU patients may not be quantitatively sleep deprived, they all demonstrated fragmentation and nonconsolidation of their sleep–wake cycles as well as a predominance of stage 1 sleep. Thus, they may be functionally sleep deprived. The sleep continuity theory demonstrates that consolidation of sleep is as important as total sleep time, as decrements in daytime function increase as the length of periods of consolidated sleep decreases (28-30). It is likely that many of these patients suffer from chronic daytime and/or nighttime sleepiness as well as neurocognitive performance deficits that are common in individuals experiencing sleep fragmentation (28-30).

ICU patients appear to lose the ability to maintain the normal circadian night/day distribution of sleep and wake. Our data indicate that total sleep time is redistributed over a 24-h period with large individual variations in total sleep time. These data are in agreement with other studies evaluating sleep in the ICU with continuous polysomnography (7, 31). Our results reinforce the importance of performing sleep studies over a 24-h period to adequately characterize sleep–wake patterns in this patient population as it is evident that studying nocturnal sleep alone is insufficient.

Decreased or absent REM sleep as well as its equal distribution across the day and night is abnormal. REM sleep typically occupies 20–25% of nocturnal sleep time in normal individuals and unlike delta sleep, REM distribution remains relatively stable throughout the life span (17). The mechanisms responsible for the absence or abnormal distribution of REM sleep in critically ill patients is unknown, but may be explained by (1) inadequate time during short bouts of sleep to cycle into REM sleep (17), (2) disturbances in the circadian system that normally controls the timing of REM sleep in normal individuals (32, 33), and (3) underlying disease and specifically mediators (endotoxin) that may be released with inflammatory states/sepsis, which may decrease or inhibit REM sleep (34, 35). This may help to explain why our septic subjects did not demonstrate any REM activity. We do not believe that medications were responsible for the REM suppression demonstrated by this group of patients as the majority (84%) of the patients studied were not on REM-suppressing medications.

This study also provides insight into the effects of sepsis on sleep and wake states. The finding of EEG slowing with sepsis states is most consistent with septic encephalopathy (35). Interestingly, our patients with sepsis demonstrated no evidence of clearly definable sleep or wake states by standard monitoring criteria over the 24- to 48-h monitoring period. Patients with underlying sepsis appear to be in a dissociated state of consciousness, deprived of both normal sleep and wake states. Although not previously described in older studies evaluating sleep in ICU patients (2-4), Cooper and coworkers (31) recently described similar findings in a group of mechanically ventilated patients with acute lung injury. It is likely that many ICU patients may demonstrate a similar state of dissociated consciousness.

Although EEG changes with previously diagnosed sepsis have been described in the literature (36-38), the observation that these EEG changes may precede other clinical features of sepsis (fever, tachycardia, hypotension) is a novel finding. Five of the 22 patients demonstrated EEG findings that were consistent with mild to moderate encephalopathy, before other signs of sepsis were present. None of the nonseptic patients demonstrated this EEG pattern. Future studies will need to be prospectively performed to determine if continuous EEG will be a valid, noninvasive marker of early sepsis.

Our data indicate that mechanically ventilated medical ICU patients manifest sleep–wake pattern abnormalities. ICU patients demonstrate large individual variations in total sleep time and, in general, are not quantitatively sleep deprived. Sleep architecture and distribution are severely altered as illustrated by excessive sleep fragmentation, selective sleep stage deprivation, and loss of the normal circadian night–day, sleep–wake cycle distribution. Although all of the mechanisms responsible for these sleep–wake abnormalities are not yet elucidated, it appears that the impact of environmental noise on sleep disruption is much less important than previously described. Our data also indicate that all patients are at risk for sleep–wake disturbances as age, sex, severity of illness, and duration of stay were not associated with sleep–wake abnormalities. Further research is needed to better define the mechanisms responsible for sleep abnormalities in patients who are critically ill.