A total of 203 patients completed the questionnaire (Table 1). Overall, there were significantly more males than females in the study (p < 0.05), although there were no significant (p > 0.05) gender differences among ICUs. The MICU patients were significantly younger than patients in other units (p = 0.003). Otherwise, there were no age-dependent differences (p > 0.05) among the ICUs. The durations of stays in the CICU and MICU were longer than in other units (p = 0.02). The MICU had significantly more ventilated patients than did other units (p = 0.001).

Sleep quality in the ICU was perceived as significantly poorer than sleep at home (p = 0.0001) by all subjects (Figure 2). Neither quality nor perceived daytime sleepiness changed significantly over the course of any patient's ICU stay (p > 0.05). There were no significant differences between genders or individual ICUs in perceived ICU sleep quality or daytime sleepiness, nor were there any significant correlations of perceived ICU sleep quality or daytime sleepiness with patient age or duration of ICU stay (p > 0.05). Perceived ICU sleep quality was not significantly different for ventilated and nonventilated patients (p > 0.05). These data indicate that poor sleep quality and sleep disruption are problems common to all types of ICUs and to many ICU patients.

Fig. 2. Bar graph highlighting differences in perceived sleep quality between home and the ICU. *Sleep quality in the ICU was perceived as significantly poorer than sleep at home (p = 0.0001) by all subjects.

Download Figure | Download Powerpoint

[More] [Minimize]

There were statistically significant differences between the individual environmental factors (p < 0.0001) in terms of their perceived degrees of ICU sleep disruption (Figure 3). Checking of vital signs and phlebotomy were perceived as significantly more disruptive (p = 0.006) to sleep than was the mean of the other factors. Environmental light and medication administration were perceived as significantly less disruptive (p = 0.003) to sleep than was the mean of the other factors. Environmental noise, nursing interventions, and diagnostic tests were perceived as not significantly different (p > 0.05) in their disruptive effect on sleep than was the mean of the other disruptive factors. These data indicate that multiple environmental factors disturb sleep in the ICU, with noise being first among a number of important sleep-disruptive factors. However, there were no significant associations between individual environmental factors (noise, light, nursing interventions [bathing, etc.], diagnostic tests [i.e., chest radiography], recording of vital signs, blood sampling, and the administration of medications) and perceived sleep disruption in relation to patient age, gender, length of stay, ventilator status, or respective unit (p > 0.05).

Fig. 3. Bar graph highlighting perceived sleep disruption from the ICU environment. There were significant differences between the individual environmental factors (p < 0.0001) in terms of perceived ICU sleep disruption. *Recording of vital signs was perceived as significantly more disruptive (p = 0.006) to sleep than the mean of the other factors. **Light and medication administration were perceived as significantly less disruptive to sleep than the mean of the other factors. Noise, nursing interventions, and diagnostic testing were not significantly differently disruptive to sleep (p > 0.05) than the mean of the other disruptive factors.

Download Figure | Download Powerpoint

[More] [Minimize]

One hundred and sixty ICU patients completed the final section of the questionnaire, which evaluated the relationship between specific ICU noises and perceived sleep disruption. There were significant differences (p < 0.0001) among the specific ICU noises in terms of their perceived disruptive effect on sleep in the ICU. Communication between staff members (talking) and telemetry alarms were significantly more disruptive (p = 0.003) to sleep than was the mean of the other factors. Noises from telephones, television, and doctors' beepers were significantly less disruptive (p = 0.0001) to sleep than was the mean of the other factors. Sounds of pulse oximetry and intravenous pump alarms were perceived as not significantly different (p > 0.05) than the mean of the other disruptive factors (Figure 4). There were no significant associations between patient age or duration of stay and sleep disruption by specific noises. There were no significant differences between the respective ICUs and perceived sleep disruption by specific noises (p > 0.05).

Fig. 4. Bar graph highlighting perceived sleep disruption by specific environmental noises. There were significant differences between the individual environmental noises (p < 0.0001) in terms of perceived ICU sleep disruption. *Talking and telemetry alarms were perceived as significantly more disruptive (p = 0.003) to sleep than the mean of the other factors. **Noises from telephones, television, and doctor's beepers were perceived as significantly less disruptive (p = 0.0001) to sleep than the mean of the other factors. Pulse oximetry and intravenous pump alarms were not significantly different in terms of sleep distruption (p > 0.05) than the mean of the other disruptive factors.

Download Figure | Download Powerpoint

[More] [Minimize]

Factor analysis was used to extract patterns of association from the data set and to reduce the large number of variables from the questionnaire to a smaller number of reference factors. Table 2 presents the factor loadings for the four-factor model after orthogonal rotation. The loadings refer to the correlation of each variable with each factor. A cutoff value of 0.40 was used to determine whether a particular variable was strongly associated with a specific factor. The loadings confirmed a definite factor structure; each of the questions had a large loading (strong association) on only one factor. The four factors were interpreted as follows: Factor 1, sleep disruption secondary to interruptions by hospital staff and diagnostic tests; Factor 2, ICU sleep quality; Factor 3, daytime sleepiness; and Factor 4, sleep disruption by environmental factors (light and noise). The total variance explained by the four-factor model was 54.9%. The proportions of the total variance that were explained by Factors 1 through 4 were 18.2%, 16.2%, 11.7%, and 8.8%, respectively.

Factor analysis confirmed several findings from the primary analysis. None of the four factors showed significant (p > 0.05) associations with patient age or gender. The duration of stay was not (p > 0.05) associated with sleep disruptions caused by human intervention (Factor 1), sleep quality (Factor 2), or daytime sleepiness (Factor 3). A patient's ventilator status was not (p > 0.05) associated with perceived sleep disruptions caused by human intervention (Factor 1), noise/ light (Factor 4), or sleep quality (Factor 2). Also, neither sleep quality nor daytime sleepiness was (p > 0.05) associated with a patient's respective unit.

Factor analysis allowed the detection of specific associations with respect to sleep quality, daytime sleepiness, and sleep disruption caused by the ICU environment that were not evident in the primary analysis. Daytime sleepiness (Factor 3) was significantly different in ventilated and nonventilated patients (p = 0.008), with ventilated patients experiencing more daytime sleepiness than nonventilated patients. Sleep disruption caused by environmental light and noise (Factor 4) showed a weak but significant positive correlation (r = 0.19; p = 0.006) with duration of ICU stay (i.e., patients who had longer stays in the ICU were more likely to have their sleep disturbed by environmental light and noise). Sleep disruptions caused by human interventions/diagnostic tests and environmental light/noise (Factors 1 and 4) were significantly different in the different units (p = 0.0365 and p = 0.0028, respectively) (Table 3).

Table 3 shows that patients in the MICU perceived interruptions caused by human interventions and diagnostic tests (Factor 1) to be significantly more disruptive to their sleep than did patients in the SICU and CICU (p = 0.012 and 0.016, respectively). There were no significant differences among the other groups with respect to Factor 1 (p > 0.05). MICU patients also perceived environmental light and noise (Factor 4) to be significantly more disruptive to their sleep than did patients in the SICU and CCU (p = 0.009 and 0.001, respectively). Patients in the CICU perceived environmental light and noise to be significantly more disruptive to their sleep than did patients in the CCU (p = 0.016), with a trend toward significance versus patients in the SICU (p = 0.06). There were no significant differences among the other groups with respect to Factors 1 and 4 (p > 0.05).

Our study design had several limitations. Because the study assessed sleep quality subjectively, we were unable to determine the patients' true sleep architecture and degree of sleep disruption caused by the various environmental stimuli. However, we believe that this investigation was an essential first step in understanding the factors perceived by ICU patients as disturbing sleep. We did not control for patients' severity of illness or medication use, both of which factors may have affected the results of the study. We also did not have a control group of hospitalized non-ICU patients for comparison, since our main objectives were to determine whether sleep quality was disturbed in an ICU setting and to determine the effect of ICU environmental stimuli (specifically noise) on perceived sleep disruption. Thus hospitalized non-ICU patients would not have constituted a suitable control group, since the environmental stimuli to which they are exposed differ from those to which the ICU patient is exposed. Future studies, however, should be performed with polysomnography to accurately quantify total sleep times and define sleep architecture.

The demographics of our ICU patients were different across the ICUs in our study. There were significantly more males than females in the cohort. MICU patients were significantly younger than the patients in other units, and MICU and CICU patients had significantly longer durations of stay than did CCU or SICU patients. However, we do not believe that these demographic differences adversely affected our results, because patient age, sex, and duration of ICU stay as a whole were not significantly associated with differences in perceived sleep quality, daytime sleepiness, or sleep disruptions caused by environmental stimuli.

Because the present study was the first instance of application of our questionnaire to a large ICU population, the test-retest reliability and validity of the questionnaire are unknown, and will require further study. Our questionnaire appears to be internally valid, on the basis of the groupings in the factor analysis. Assessment of the validity of our questionnaire for assessing sleep quality and the etiologies responsible for sleep disruption in the ICU will require future objective (polysomnographic) studies in conjunction with the questionnaire.

Recall bias is another potentially confounding problem for questionnaire studies in general. This did not appear to be a major problem with our questionnaire, however, since the duration of ICU stay was not associated with changes in sleep quality, daytime sleepiness, or perceived environmental disruptions of sleep. Because no large studies have objectively evaluated the accuracy of patient recollections of their ICU stays, and it is clinically apparent that many ICU patients who undergo heavy sedation do not remember much of their ICU stay, our results may have been subject to patient recall bias, especially by those patients with longer durations of ICU stay and in those patients who were heavily sedated for a portion of their ICU stay.

Additionally, there may have been a recruitment bias that potentially prevents application of the results of our study to the entire ICU population. Although all ICU patients were eligible to participate in the study, those patients who were not discharged from their respective ICUs because of increased severity of illness were not included in the study. Therefore, our results cannot be generalized to critically ill ICU patients who die before their ICU discharge.

Our finding that patients perceive their sleep quality to be worse in the ICU than at home is supported by previous investigations (1-8). Studies of sleep in ICU patients indicate that their sleep is abnormal, as demonstrated by a predominance of Stages 1 and 2 sleep, decreased or absent Stages 3 and 4 and REM sleep, shortened periods of REM sleep, frequent arousals, and sleep fragmentation (1-8). Our results also demonstrate that perceived sleep quality and daytime sleepiness did not improve over the course of our patients' ICU stays. These results are consistent with those of previous polysomnographic studies, which showed that 40% to 50% of the total sleep time in an ICU occurred during the day, and that altered sleep patterns may not improve over the course of a patient's ICU stay, and may take several days to normalize after transfer of the patient to a general hospital ward (2, 3, 6, 7).

Although daytime sleepiness appeared to be a problem common to all of the types of ICUs investigated in our study, factor analysis showed that patients who had been mechanically ventilated during their ICU stay perceived that they were significantly more sleepy during the day than were nonventilated patients. This finding may be explained by the typically greater sedation of mechanically ventilated than of nonventilated patients, and their greater severity of illness than that of nonventilated patients. Further studies will be necessary to determine whether sedation and severity of illness are correlated with sleep disruption in the ICU.

Interestingly, there were no differences among any of the individual ICUs in perceived sleep quality, nor were there any associations between sleep quality and patient age, gender, duration of stay, or ventilator status. This suggests that poor sleep quality and disrupted sleep are problems common to all ICUs and to many ICU patients. It also suggests that factors disturbing sleep are generic to all ICUs.

Our findings support the assertion that environmental causes of sleep disruption in the ICU are multifactorial. Our data indicate that human interventions and diagnostic testing appear to be as important to disrupting sleep as is environmental noise. Our findings do not support our primary hypothesis that ICU noise is the most disruptive environmental stimulus to sleep for most ICU patients. In fact, the effect of specific ICU noises on sleep disruption appears to be low, as shown by the low mean sleep-disruptive scores in our study. This finding contradicts much of the current literature, which suggests that noise is the major ICU factor responsible for sleep disruption (1, 6, 10, 12, 13, 15-18, 21, 27-29).

Why was noise not perceived as the most sleep-disruptive factor in the ICU? A number of possible explanations exist for this finding. First, noise in the ICU setting may not be as disruptive to sleep as is currently thought. Patients in intensive care may adapt quickly to environmental noise. Previous studies have shown that patients habituate to sound by increasing their arousal threshold with time, with some individuals being able to increase their arousal threshold for noise to more than 80 dB (30, 31). Second, noise may have caused arousals but not awakenings during sleep among our patients, leading to sleep fragmentation and poor sleep quality. Patients may not have been able to recall the etiology of their fragmented sleep because they were never fully awakened from sleep. Third, noise may have caused awakenings during delta sleep, when it is more difficult for individuals to recall the etiology of an awakening. This last explanation is an unlikely one, since the current literature indicates that ICU patients show decreased or absent delta sleep (2, 6).

Our results support the postulate that interruptions of sleep in the ICU are multifactorial, and that those caused by human interventions and diagnostic tests are important. This finding is consistent with the results described earlier in a study done by Woods and colleagues (23), who found that postcardiotomy patients had their sleep interrupted by human interventions (recording of vital signs, phlebotomy, etc.) on an average of 59.5 times per night, with the greatest amount of uninterrupted sleep time being 43 min on postoperative Day 1. The number of nocturnal awakenings decreased and the uninterrupted potential sleep time increased on subsequent hospital days. Meyer and coworkers monitored sleep interruptions in ICU patients over a 24-h period and showed that interruptions caused by human interventions occurred at least hourly throughout the day and night (15). Future research should focus on methods for decreasing frequent interruptions of sleep in ICU patients, especially at night, to allow them to achieve longer periods of consolidated sleep.

Patient characteristics such as age, gender, or ventilator status were not related to perceived sleep quality or disruptions caused by any specific environmental stimulus in our study. That older age was not associated with poorer sleep quality is not necessarily consistent with findings reported in the literature. Typically, older individuals are more likely to be aroused from sleep than are younger individuals (30, 31), yet the mean age of our patients was 58.6 ± 15.4 yr. We may not have sampled enough younger patients (age < 40 yr) to show a significant association between age, sleep quality, and sleep disruptions. One would also assume that patients who were mechanically ventilated and thus exposed to the additional noise of ventilators and alarms would perceive the ICU environment, and specifically noise, as more disruptive to their sleep than would nonventilated patients. This was not the case in our study. Our results suggest that all ICU patients, and not specific subgroups, are at risk for sleep disturbances.

Our results support the assertion that sleep disruption in the ICU is multifactorial, and suggest that other, nonenvironmental factors may be responsible for causing sleep disruptions in the ICU patient population. In general, the mean disruption scores for the environmental factors and specific ICU noises causing sleep disruption in our study were relatively low, although the perceived sleep quality among the vast majority of patients was poorer than that of sleep at home. It is clear that the ICU environment is not solely responsible for disturbed sleep–wake patterns. Several nonenvironmental factors, such as medications (32), pain, fever (9, 33), and a patient's underlying chronic disease may have adverse effects on sleep quality (34, 35). Studies with surgical ICU patients that have decreased the number of environmental interruptions of sleep have not demonstrated improvement in the patients' altered sleep patterns. Other studies have shown continued alterations in sleep duration and other sleep parameters when patients were transferred from the ICU to a hospital ward (3). It is possible that the severity of illness or underlying disease states contributed to the perceived differences in sleep quality in our study. Future studies should examine the effects of relationships between disease severity, age, medications, and time in the ICU on disruptions of sleep–wake patterns in the ICU, although this may prove difficult, given the number of potentially confounding factors.