Introduction

The acute care units (ACU) in a hospital, which include the medical and surgical intensive care units (MICU and SICU), coronary care units (CCU) and post-anesthesia care units (PACU), are areas for patients who need special attention. Patients in ACUs are usually critically ill, making them more susceptible to the unfavorable atmosphere in the hospital. One of these unfavorable factors is sleep disruption and deprivation [1–4]. Frequently, however, sleep disruption and deprivation is not noticed by the medical staff in ACUs, and many of these professionals have limited knowledge regarding sleep disruption in ACUs and its impact on patient health [5]. In ACUs, many organ systems are aggressively monitored for the development of dysfunction or failure, but frequently the function of sleep in acutely ill patients is ignored.

Many factors may affect sleep in the ACU, including therapeutic interventions, diagnostic procedures, medications, the underlying disease process, and noise generated in the ACU environment. Many detrimental physiological effects can occur secondary to noise, including cardiovascular stimulation, increased gastric secretion, pituitary and adrenal stimulation, suppression of the immune system and wound healing, and possible contribution to delirium [6–10].

It is our opinion that physicians and nurses working in ACUs have a low level of awareness regarding sleep disruption in critically ill patients and the impact of this disruption on the health of these patients. Over the past few years, many studies have endeavored to objectively assess sleep in the ACUs, as well as the effect of mechanical ventilation and circadian rhythm changes in critically ill patients. At this time, therefore, it is important to review published data regarding sleep in ACUs to improve the knowledge and recognition of this problem by health care professionals. We have therefore reviewed the methods used to assess sleep in ACUs, factors that may affect sleep in the ACU environment, and the clinical implications of sleep disruption in the ACU.

Normal sleep

Sleep is a physiological state that humans need to pass through every day to repair and restore body functions. Typically, humans adapt to a 24-h circadian pattern, where they sleep at night and are awake during the day. This 24-h internal clock (circadian pattern) is maintained by environmental factors, primarily light exposure, which affects melatonin secretion at night. Several physiological and biochemical body functions follow this circadian pattern. In normal individuals, sleep latency usually lasts about 10–20 min and sleep lasts for 6–9 h on average, although variability among individuals can be significant.

While asleep, an individual passes through four to six cycles, each of which consists of five sleep stages (Fig. 1b). One is called rapid eye movement (REM), and the rest (stages 1 through 4) are non-REM (NREM) sleep. Sleep onset begins with stage 1, which is a transitional stage, usually about 5% of normal sleep, and then progresses into stage 2, which can constitute as much as 50–60% of sleep. Thereafter, sleep progresses to stages 3 and 4, which are known as delta, slow-wave, or deep sleep. These stages occur predominantly during the first half of the night and constitute 15–20% of night sleep. Compared with light sleep, slow-wave sleep is usually deeper and more restful and requires a more intense stimulus to awaken the sleeper. REM sleep occurs approximately every 90 min in 4–6 cycles, increasing in duration during the second half of the night. During REM sleep, there is inhibition of spinal motor neurons, which leads to paralysis of the major muscle groups but spares the diaphragm and ocular muscles. Although brain activity is increased during REM sleep, this stage is considered to be restful, with a variable arousal threshold. Factors disturbing sleep, such as noise, may affect the integrity of sleep. This may result in electroencephalographic (EEG) arousals and awakenings, which may affect sleep architecture and prevent the normal progress into the deeper stages of sleep [11].

Fig. 1
figure 1

Hypnogram of a patient 2 days after acute myocardial infarction (a) and 6 months later (b). The vertical axis represents sleep stages and the horizontal axis time. a Significant sleep disruptions and awakenings and absence of slow-wave and REM sleep. b Clearly defined sleep cycles with progression to different sleep stages. S1 Stage 1, S2 stage 2, S3 stage 3, S4 stage 4, REM rapid eye movement, MVT movement, WK wake (BaHammam, unpublished data)

Full size image

Methods for sleep assessment in ACUs

Before discussing factors that affect sleep in ACUs, it is important to describe the methods used to assess sleep in ACU settings. Among the methods used to assess sleep in ACUs are staff observations [12], self-reporting, polysomnography (PSG), and bispectral analysis.

Self-reporting

Several studies have subjectively assessed sleep in ACUs using qualitative and quantitative methodologies. In an evaluation of physical and psychological stressors in 50 intensive care unit (ICU) patients who completed the ICU Environmental Stressor Scale, insomnia was ranked as the second most important stressor, second only to pain [13]. Of patients interviewed 3 days after discharge from the ICU, 61% reported sleep deprivation and 7% rated insomnia as their worst experience in the ICU [14]. In another study, 27% of ICU patients reported having insufficient sleep [15]. Furthermore, perceived ICU sleep quality was poor compared with baseline and did not change over the course of patient stay in the ICU [16]. However, self-reported data suffer from recall bias and a lack of objective assessment of sleep quality.

Polysomnography

The standard diagnostic test for sleep orders is polysomnography, during which neuro-cardio-respiratory parameters are usually monitored. Although PSG studies are usually performed in a sleep laboratory, PSG has been used to assess sleep in critically ill patients in ACUs [2–4], [17–26]. In general, these studies were conducted in different ACUs and involved small groups of patients with different underlying medical and surgical problems, including those with acute coronary syndrome [3, 23], patients recovering from open heart surgery [4], mechanically ventilated patients [22], trauma patients [19], patients with neurological and respiratory disorders [24], and mixed groups of critically ill patients. Previous studies have described the techniques for performing PSG in critically ill patients in ACUs attended by ICU nurses, but not by sleep technicians or in a laboratory setting attended by a sleep technician [23–24]. Overall, sleep recording on ACU patients revealed sleep fragmentation, increases in stages 1 and 2 sleep, reductions in slow-wave and REM sleep, and reductions in total sleep time and sleep efficiency. Most of the previous studies only evaluated sleep during the night, rather than over a 24-h period. However, those studies that did monitor PSG continuously for ≥24 h showed that 40–50% of the total sleep time in ACUs occur during the day [2, 18, 25, 26]. This disturbed sleep pattern may not improve over the course of a patient’s stay in the ICU and may take several days to normalize after discharge from the ICU [21]. For example, repetitive whole-night in-hospital EEG showed that normal sleep patterns did not return to patients with acute myocardial infarction until 9 days after discharge from the ICU [3].

Many factors in the ACUs may disrupt the typical EEG architecture seen during normal sleep, making the interpretation of sleep quality difficult. In a recent study, 24-h PSG monitoring of critically ill mechanically ventilated patients in the ICU revealed that eight patients had disturbed sleep architecture [22]. The remaining patients had either coma (seven patients) or atypical sleep (five patients), in which their EEGs were intermediate between sleep and coma and characterized by a virtual absence of stage 2 and REM sleep [22]. The authors thought that these EEG changes were likely due to sedation.

The bispectral index (BIS)

Because of the difficulties of performing full PSG in the ACUs environment and the need for certain expertise, an alternative method, the bispectral index (BIS), has been used to objectively assess sleep in ACU settings. Previously, the BIS was used extensively by anesthetists to measure the depth of anesthesia and sedation. The BIS monitor displays a real-time EEG trace, acquired from a frontotemporal montage. It combines power spectral analysis with interrogation of the phase relationship between component waves of different frequencies. The monitor generates a dimensionless number on a continuous scale of 0–100, with 100 representing normal cortical electrical activity and 0 indicating cortical electrical silence [27].

Several studies have used the BIS to assess sleep in ACUs (Fig. 2). In one study, the investigators compared the changes in the BIS with the conventional EEG stages of sleep in five subjects during the early part of the night. They reported a good correlation between the BIS and the level of natural sleep [28]. Light sleep occurred at BIS values of 75–90, slow-wave sleep occurred at BIS values of 20–70, and rapid eye movement sleep occurred at BIS values of 75–92 [28]. In another study, which assessed sleep patterns in 27 ICU patients using the BIS and submental electromyogram, none of the patients showed a completely normal sleep pattern [29]. In comparing the BIS and spectral edge frequency (SEF) with PSG in ten patients with mild apnea/hypopnea syndrome to test whether these signals could predict physiologic sleep stages, neither the BIS nor SEF reliably indicated conventionally determined sleep stages [30]. Further studies are needed to access the utility of BIS as a monitor of sleep in patients in ACUs.

Fig. 2
figure 2

Normal changes in sleep stage and BIS in a sleeping healthy volunteer throughout the night (−1, REM sleep; 0, awake; 1 and 2, light sleep; 3 and 4, slow-wave sleep) [29]. Reproduced with permission from the publisher

Full size image

Skin potential levels of critically ill patients in the ICU have been recorded to monitor the sleep-wake cycle of these patients [31]. Recently, actigraphy, a portable device that records movement over extended periods of time and worn most commonly on the wrist, [32] has been used to assess agitation and sleep in critically ill patients [33, 34]. In one study, actigraphy correlated well with observed and subjective scores on agitation and sedation scales [34]. Further studies are needed to assess the usefulness of this technique in assessing circadian patterns in critically ill patients.

Factors affecting sleep in ACUs

Environmental factors

Environmental factors in ACUs include noise produced by phones, pagers, alarms, and staff and light, which may disturb the normal light-dark cycle.

Noise

Several studies have shown that the noise level in ACUs is quite high [17, 35–41]. Excess noise may increase the mistakes by [42, 43] as well as impair the concentration and mental efficiency of the staff [44, 45]. Noise levels in ACUs range from 60 to 84 dB throughout the day and night [36]. As a reference, a busy office has an average noise level of 70 dB, and a pneumatic drill heard from around 15 m away has a noise level of 80 dB [16, 37]. As noise measurements use a logarithmic scale, an increase of 10 dB represents a doubling of the noise level. The US Environmental Protection Agency (EPA) has recommended that hospital noise levels not exceed 45 dB during the day and 35 dB at night [46]. A study in an intermediate respiratory care unit reported a very strong correlation between the number of patient arousals and the number of sound peaks ≥80 dB [17]. In a study of 26 adult patients in a PACU, the mean integrated sound pressure was 67.1 dB [40]. Staff conversation caused 56% of the recorded noise above 65 dB, while other sources of noise, such as alarms, telephones, and nursing care, accounted for less than 10% each [40]. This finding agrees with that of two other studies, both of which reported that staff conversation was one of the most disruptive environmental noises [16, 25]. Synchronous audio, video, and polysomnographic recording in seven patients showed that 20% of the recorded arousals and awakenings were related to noise and 10% to patient care activities, whereas the causes of the remaining 70% were not identified, indicating that other as yet unknown factors are involved [25]. The above findings concur with those obtained by Freedman et al. [26] in a study of 22 critically ill patients. By simultaneously monitoring environmental noise and PSG, they found that environmental noise was responsible for 11.5 and 17% of the arousals and awakenings from sleep, respectively.

Staff-patient interactions

Despite the increased sophistication of monitoring systems in ACUs, which should decrease hands-on manipulations of sleeping patients, there are still frequent and repetitive staff interactions with critically ill patients, thus reducing patient sleep time in ACUs. Several studies have found that the mean number of staff-patient interactions ranged from 40 to 60 per night [47–49]. Using a 24-h PSG monitoring, it was found that there were eight staff-patient interactions per hour of patient sleep, with most of these interactions due to nursing activities, such as wound dressing, adjustment of intravenous drips, and administration of medications [25]. Nurses were found to routinely provide daily baths for patients between 02:00 and 05:00 on 55 of the 147 study nights [48]. Sleep disruptions caused by human interventions/diagnostic tests were found to be significantly different in different ACUs [16]. For example, patients in the MICU perceived interruptions caused by staff interactions and diagnostic tests to be significantly more disruptive to their sleep than did patients in the SICU and the cardiac intermediate care unit. Based on these results, it has been recommended that nocturnal staff-patient interactions be clustered, thus allowing patients uninterrupted periods of sleep [50–52]. In addition, although most of these studies assessed sleep interruptions and staff interactions at night, future studies should assess these variables during both the day and night.

Circadian pattern disturbances

About 40–50% of the total sleep time in ACUs has been found to occur during the day [2, 18]. For example, using a 24-h PSG monitoring, approximately half the sleep of ventilated patients in the ICU occurred during daytime (06:00–22:00), whereas sleep in healthy controls tended to be more nocturnal [25]. Sleep in humans follows a circadian pattern, in which people usually sleep at night and are awake during the day. This circadian pattern, which is coupled to physiologic cycles, is maintained by internal and external cues called Zeitgebers (makers or cues of time). Among these cues, the most important is exposure to light. Disturbances of the light-dark cycle and frequent environmental noises and interventions in the ACUs may disturb the normal circadian rhythm. Measurable markers for this circadian pattern include body temperature and the levels of certain hormones, such as melatonin and cortisol. Body temperature normally follows a circadian pattern, rising during the day and falling at night. In general, a falling body temperature induces sleep, whereas a rising temperature provokes wakefulness. As shown by recording the rectal temperature of 15 patients who spent greater than 1 week in the ICU, there was a marked disruption of circadian rhythms [53]. Similar findings were reported in otherwise healthy patients following cardiac surgery [54].

Melatonin is considered to be the best marker of circadian rhythm [55, 56]. Individual melatonin profiles are highly reproducible and are less subject to masking factors than are other rhythm markers such as core temperature and cortisol level. Melatonin secretion is normally increased at bedtime and remains high until early morning. Several studies have shown that the circadian release of melatonin is disturbed in critically ill patients [57–61]. For example, when melatonin levels in the blood and urine were measured over three consecutive days in eight critically ill patients during deep sedation and mechanical ventilation, the circadian rhythm of melatonin secretion was abolished in all but one patient, and there was no correlation between melatonin levels and levels of sedation [57]. This suggested that the impairment of the circadian release of melatonin may play a role in sleep disruption and delirium in the ICU. When the urine concentrations of the melatonin metabolite 6-sulphatoxy melatonin (6-SMT) and free cortisol were monitored in 16 patients for their entire stay in the ICU, the diurnal rhythms of both hormones were consistently or periodically disturbed in 65–75% of the patients, despite controlling for illumination [58]. Treatment with adrenergic drugs and benzodiazepines increased 6-SMT excretion [58]. In addition, there was a significant correlation between ICU psychosis and an irregular melatonin circadian rhythm [59].

Hyposecretion of melatonin may have detrimental effects on critically ill patients. In animal models, hyposecretion of melatonin has been shown to impair mitochondrial oxidative phosphorylation [62] and the capacity to survive endotoxemia [63, 64]. Moreover, the antioxidant effects of melatonin have been observed to prevent ischemia-induced renal damage [65, 66]. Although some investigators have recommended that melatonin be given to patients in the ICU to induce sleep and to resynchronize their biological clocks [67], there is little data to support that. Treatment with melatonin may also regulate the secretion of growth hormones and prolactin [68]. Future studies should assess the effect of exogenous melatonin administration on sleep and circadian pattern in acutely ill patients.

Medications

Many drugs are used frequently in ACU settings, including hypnotics, narcotics, and ionotropic drugs. However, heavy sedation does not imply good sleep. For example, although the benzodiazepines decrease sleep latency and awakenings and increase sleep duration and efficiency (sleep duration/time in bed), these drugs also significantly reduce slow-wave and REM sleep, increase spindles, increase cortical activity at low doses, and decrease EEG amplitude at high doses [69–71]. Narcotics also suppress deep and REM sleep and increase arousals and stage 1 sleep [72]. However, the effects of benzodiazepines and narcotics on sleep in patients in the ACUs have not been properly investigated.

Many critically ill patients receive intravenous ionotropic drugs, such as dopamine and norepinephrine. Both of these drugs are important central neurotransmitters and are associated with cortical activation, thus possibly increasing arousals. Fortunately, both drugs do not appear to cross the blood-brain barrier under normal conditions [73]. Nevertheless, the administration of epinephrine to patients sedated with propofol increased their sedation score and BIS values [74]. In rats, blood-brain barrier permeability is increased by tumor necrosis factor alpha (TNF-α), which is secreted as part of the sepsis cascade [75]. Additional studies are needed to examine the effects of these drugs on sleep architecture in critically ill patients.

Mechanical ventilation

Mechanical ventilation is an invasive treatment modality that may affect sleep quality in critically ill patients. Among the causes of sleep interruption in ventilated patients are the severity of the illness, nursing interventions for suctioning and other reasons [49], sedation and other medications, mode of ventilation, asynchrony between the patient and the machine [76], and failure of patients to trigger ventilators due to the use of improper settings, which may result in unloading of the respiratory muscles or intrinsic positive end-expiratory pressure [77]. As factors other than mechanical ventilation may affect sleep in this group of patients, it is difficult to assess the effects of mechanical ventilation on sleep. To date, therefore, there have been only a few studies subjectively or objectively evaluating sleep during mechanical ventilation.

In interviews with 158 survivors of critical illness 2 months after discharge, 30% indicated having experienced agony/panic during mechanical ventilation, which was associated with asynchrony between the patient’s breathing and the ventilator [78]. Using PSG, ventilated critically ill patients were found to have as many as 20–63 arousals and awakenings per hour [22, 23, 25, 79]. In general, PSG monitoring of sleep during mechanical ventilation revealed an increase in stage 1 sleep and reductions in slow-wave and REM sleep [22, 25, 26]. As stage 1 occurs throughout the night as a transitional stage of sleep, increases in stage 1 are considered indicators of sleep disruption [11].

Pressure support has been shown to induce central apneas in healthy subjects during sleep, which in turn may cause sleep disruption [80]. In assessing the influence of the mode of ventilation on sleep quality, 11 critically ill patients were randomized to receive at least 2 h each of three ventilator modes: assist-control ventilation, pressure support alone, and pressure support with dead space [81]. Sleep fragmentation was greater during pressure support (79 vs 54 arousals and awakenings per hour) and sleep efficacy was greater during assist-control (75 vs 63%). Six of the 11 patients developed central apneas while on pressure support but not during assist-control ventilation. Central apnea was more common among patients with heart failure. Among patients with central apneas, adding dead space decreased sleep fragmentation (arousals and awakenings per hour). The most important determinant of central apneas was the difference between a patient’s apnea threshold and arterial partial pressure of carbon dioxide (PaCO2) while breathing at rest [81]. Central apneas were more likely to occur in patients with resting PaCO2 values close to their apnea thresholds. Future studies should systematically explore the effect of other modes of ventilation, such as proportional assist ventilation (PAV) [80], on sleep architecture and the development of central apneas.

When the circadian pattern of melatonin, as 6-SMT, was measured in 16 critically ill patients, the diurnal fluctuation in 6-SMT levels was more likely to be disrupted in patients receiving mechanical ventilation [58]. That is, 6-SMT excretion was significantly lower during periods with mechanical ventilation compared to periods without ventilation. This difference was independent of the mode of ventilation (CPAP vs assist-control ventilation) [58].

Underlying acute and chronic illnesses

Underlying medical problems and complaints may also disturb sleep in acutely ill patients. Among these underlying problems are chronic respiratory disorders, neurological disorders, sleep disorders, heart failure, renal impairment, liver impairment, infection, fever, and pain. For example, there was a significant correlation between sleep fragmentation, measured as the sum of arousals and awakenings, and acuity of illness, measured as APACHE II scores [82, 83].

Sleep disruption has been demonstrated in certain groups of acutely ill patients, such as post-myocardial infarction patients [3, 23] (Fig. 1a), postoperative patients [4, 18, 84, 85], and patients with sepsis [61]. In a nocturnal PSG study of the sleep patterns of 12 patients with acute myocardial infarction, both during their acute illness and after discharge to the general ward for 9 days, the initial PSG in the ICU showed significant disruptions of sleep architecture, with a significant increase in wakefulness and a significant reduction in REM sleep [3]. In these patients, however, sleep quality improved gradually over time. In another study, we evaluated nocturnal sleep quality of 20 patients with acute myocardial infarction who were off sedatives and ionotropes for 48 h and under controlled dark-light exposure and daytime nap, within 3 days of the acute event and 6 months later. To avoid the environmental effects of the acute care areas, both studies were performed in the sleep disorders center and a protocol was implemented to minimize staff-patients interaction (unpublished data). The comparison revealed a significant increase in arousal index, spontaneous arousals, stage shifts, REM latency, and wake time and a significant reduction in total sleep time, sleep efficiency, and REM sleep during the acute event, which indicate that factors other than the ACU environment affect sleep quality in patients with acute myocardial infarction. In a group of patients recovering from elective abdominal surgery, nocturnal PSG demonstrated initially increased awakenings and significant reductions in REM and deep sleep, whereas REM rebound was observed on post-operative night 3 [85]. Using 24- to 48-h PSG monitoring in five critically ill patients with sepsis, Freedman et al. [26] demonstrated that those patients showed no evidence of clearly definable sleep or wake states. This finding is consistent with previous studies showing EEG changes in patients with severe sepsis [86, 87]. Measurements of 6-SMT levels every 4 h over a 24-h period in septic sedated ICU patients, nonseptic ICU patients, and healthy controls showed that the amplitude of the circadian fluctuation of 6-SMT was markedly lower in the septic ICU patients compared with the two other groups [61].

It is worth mentioning here that patients with sleep-related breathing disorders such as obstructive sleep apnea and sleep hypoventilation may present with acute respiratory failure, necessitating ICU admission [24, 88–90]. These patients are usually obese and present with hypercapnic respiratory failure that can be misdiagnosed as chronic obstructive lung disease or congestive heart failure. Early diagnosis and institution of proper treatment may improve outcome [24].

Potential implications of sleep deprivation and disruption

The impact of sleep deprivation and arousals in critically ill patients has not been systematically studied, although the effect of sleep deprivation on healthy adults and the effect of arousals on the cardiovascular system in patients with obstructive sleep apnea or periodic breathing have been extrapolated to critically ill patients. Theoretically, the deleterious effects of sleep deprivation and arousals may be more pronounced in critically ill patients, but more data are needed. The immune system has long been regarded as a vulnerable target for sleep deprivation. Cytokines synthesized by the immune system may play a role in normal sleep regulation, by increasing NREM sleep and decreasing REM sleep, and during inflammatory events, an increase in cytokine levels may intensify their effects on sleep regulation [91]. Current evidence suggests that acute and chronic sleep deprivation is associated with decreased proportions of natural killer cells [92], lower antibody titers following influenza virus immunization [93], reduced lymphokine-activated killer activity [94], and reduced interleukin-2 production [94]. Moreover, sleep deprivation may alter endocrine and metabolic functions, altering the normal pattern of cortisol release and contributing to alterations of “glucocorticoid feedback regulation” [95], glucose tolerance, and insulin resistance [96]. In healthy subjects, sleep deprivation is associated with a negative energy balance [97]. Although the effect of sleep deprivation on ventilatory responses in critically ill patients is not known, sleep deprivation per se in ten healthy subjects did not reduce the sensitivity of central chemoreceptors or change resting ventilation or metabolism [98]. In a recent study of cardiovascular autonomic modulation during 36 h of total sleep deprivation in 18 normal subjects, acute sleep deprivation was associated with increased sympathetic and decreased parasympathetic cardiovascular modulation and decreased baroreflex sensitivity [99]. In addition, frequent arousals were associated with elevated catecholamine release and increased blood pressure [100].

Delirium, formerly known as ICU psychosis or the ICU syndrome, is a serious problem facing the staff in ACUs. The incidence of delirium in a cohort of mechanically ventilated patients in medical and coronary ICUs was found to be as high as 81% [101]. In addition, delirium was an independent predictor of higher 6-month mortality and longer hospital stay, even after adjusting for relevant covariates [102]. Sleep deprivation has been suggested as a risk factor for delirium in critically ill patients. Using staff observations, without objective methods, there was a higher prevalence of delirium among sleep-deprived patients [103, 104]. Delirious patients were reported to have irregular patterns of melatonin release [59] and disrupted circadian rhythms, resulting in fragmented sleep-wake cycles and nighttime awakenings [105]. These findings suggest a strong association between delirium and sleep deprivation but do not establish causality. It must therefore be determined whether sleep deprivation is a cause or a symptom of delirium.

Conclusions and future perspectives

Critically ill patients are prone to sleep deprivation, poor sleep quality, and disturbed circadian patterns. The causes of sleep disruption in these patients seem to be multifactorial and include the ACU environment and staff, diagnostic and therapeutic interventions, medications, mechanical ventilation, and the underlying illness. The role and contribution of each of these factors to sleep disruption in critically ill patients need to be clearly delineated. This will assist in the development of evidence-based guidelines regarding sleep-promoting interventions in ACUs. The effects of newer modes of ventilation on sleep architecture should also be explored. Objective methods that are practical and easy to interpret are needed to assess sleep in the ACUs. The use of BIS may prove to be an easy and practical way to assess sleep in critically ill patients, but more research is needed. Staff working in ACUs should be educated about the importance of sleep, the factors that may disturb sleep, and the role of the ACU staff in promoting good sleep in critically ill patients. Finally, the impact of sleep deprivation and disturbances on patients’ stay in ACUs and on outcome should be determined.