Sleep disturbance in ICU patients

The topic of sleep disturbance in the ICU has been of some considerable interest, considering how many articles pop up when you look up "sleep disturbance in the ICU". Much of the information I used to generate the discussion section for  Question 25 from the first paper of 2008 has been derived from this excellent article from the The Open Critical Care Medicine Journal. The LITFL page on sleep in the ICU also offers an excellent brief overview, perfect for pre-exam cramming.

Sleep in the ICU

Kamdar et al (2012) have many negative things to say about sleep in the ICU population

• Fragmented and brief: ICU patients end upw ith 20-40 episodes per day of something that resembles sleep, and lasts 10-15 minutes (Freedman et al, 2001).
• Much of it is in the N1 stage
• Much of it (50%) is during daytime hours
• Many microarousals
• Many actual arousals - Cooper et al (2000) found that the patients were woken 22 times every hour, on average
• Total sleep time is decreased

Causes of sleep disturbances in the ICU

The following factors have been found to act as negative influences on sleep:

• Noise: 10-20% of wakings. The EPA recommends no higher than 45 dB in the ICU; however, this is actually quite loud - it is "the sound level recognized internationally as an upper limit for human comfort in residential interior spaces".
• Constant harsh light; misalignment of circadian cycles - artificial light is of insufficient intensity to act as a zeitgeber
• Sunlight exposure is limited or nonexistent
• Erratic stimulus (eg. hourly neuro obs)
• Appropriately timed meals are replaced by tube feeding
• Regular nursing care (eg. turns) disturbs nocturnal sleep
• Sepsis decreased REM sleep by influencing melatonin secretion
• Sedatives impair normal REM sleep
• Mechanical ventilation impairs sleep
• Pain including the discomfort of tubes and drains
• Anxiety and stress

Consequences of sleep deprivation in ICU patients

The following consequences have been ascribed to sleep deprivation, though in truth there really is no way of testing that.  Several good resources exist for this, for example Kamdar et al, 2012 and  Jolanta Orzeł-Gryglewska's "Consequences of sleep deprivation",  2010.  The experts offer the following list of problems:

• Respiratory exhaustion: after 1 night without sleep, COPD patients get worse (poorer FEV1 and FVC). Even in normal male subjects, inspiratory muscle endurance is affected (Chen et al, 1989)
• Decreased ventilatory response to hypercapnea - leading to hypoventilation. "Ventilatory chemosensitivity may be substantially attenuated by even short-term sleep deprivation", claim White et al (1983)
• Increased total body oxygen consumption (on the basis of a study by Bonnet and Arand (1995) which compared VO₂ among insomniacs and good sleepers)
• Increased risk of ischaemic cardiac events - studied by Liu and Tanaka (2002) in a population of chronically overworked Japanese executives.
• Delirium - it is either increased in incidence by sleep deprivation, or at least the two conditions share sufficient number of clinical features - for example, inattention, hallucinations and decreased short term memory (Weinhouse et al, 2009)
• Lasting neurocognitive deficits following critical illness may be exacerbated by sleep deprivation (Jackson et al, 2009), but the link is far from well-established.
• A hypercatabolic state: some chronically sleep-deprived rats seem to have developed wasting weight loss and malnutrition in spite of increased caloric intake (Everson et al, 1989). In fact, sustained deprivation of sleep ultimately killed the rats. The terminal stage of sleep deprivation in these rats resembled septic shock, wih multiorgan system failure.
• Increased catecholamine and corticosteroid levels, similar to the stress response of critical illness (Johns et al, 1971)
• Hyperglycaemia: blunted insulin secretion, decreased sensitivity to insulin, and impaired glucose regulation (Spiegel et al, 2009)
• Impaired immunity: a release of proinflammatory cytokines and a decline in T-helper cells is observed in short-term sleep deprivation (Everson et al, 1993)

Monitoring of sleep in the ICU

LITFL recommend daily sleep diaries, visual analog scales (VAS), questionnaires, and symptom or quality of life questionnaires with sleep items (subject to recall bias and other problems), direct observation of arousals and motor activity, actigraphy (using movement detectors), BIS and Bispectral index and multichannel polysomnography (gold standard). An excellent recent article by Delaney et al (2015) offers a detailed overview of the issues involved.

Polysomnography,  even though viewed as the gold standard for normal patients, has many barriers to its application in the ICU, particularly in context of the ECG-befuddling effects of exotic sedatives and encephalopathy. To give an extreme example, a patient who has had a hemispherectomy and then went on to develop HSV encephalitis will offer a highly unusual pattern of EEG activity, which will be difficult to interpret within the framework of standard EEG definitions for sleep stages. Drouot et al (2012) found that about 28% of all polysomnography studies collected in the ICU could not be classified using conventional scoring rules.

Bispectral (BIS) monitoring is barely even validated for use in anaesthetised normal subjects, and its application to sleeping ICU patients is even more dodgy. It suffers from all the deficits of polysomnography in the ICU setting, and is confounded by all the same problems, but it does not offer much opportunity for specialist waveform interpretation, as it reduces the majestic electrochemical complexity of the human consciousness into a single numerical variable. Patel et al (2001) made an attempt to assess sleep in the ICU using BIS, working from the knowledge that BIS at least correlates with EEG during normal sleep. They used a value of over 85 to define "awake", 60-85 as "light sleep" (presumably, N1) and under 60 as "slow wave sleep". REM was detected using a combination of BIS waveform analysis and EMG recordings. The study was frustrated by the fact that the patients did not demonstrate any recognisable sleep stage patterns (eg. there were no rapid eye movements during the stages that were supposed to be REM sleep), and the authors could not arrive at anything solid about the use of BIS for sleep monitoring, concluding only that "traditional classifications of EEG sleep staging are deficient when used to describe sleep in intensive care unit patients". The major advantage of BIS is the fact that it does frequently correlate with the clinically observed state of arousal, and requires little intelligence to interpret.

Actigraphy is performed by the use of a wristwatch-like accelerometer, which tells you when the patient is moving. The obvious limitation of this is the lack of correlation between movement and arousal among ICU patients. The classical extreme example of this is the patient who is paralysed with muscle relaxants.  Beecroft et al (2008) compared it to polysomnography in a group of twelve intubated ICU patients, and concluded that it was "inaccurate and unreliable", as it consistently overestimated sleep time and sleep efficiency.

Subjective behavioural assessment  is basically asking of the bedside nurse whether he or she thinks the patient is asleep or awake. It sure is a cheap method, but also almost useless. For instance, Bourne et al (2007) found that ICU staff consistently mistook sedation for sleep (why wouldn't they, as it looks the same) and overestimated the duration and efficiency of sleep. The limitations are basically the same as those of actigraphy, as the bedside staff really only have movement to go off when determining whether the patient is awake. Not only are they collecting inaccurate data, they also frequently fail to record it (unlike the always-reliable actigraphy robot).

Hybrid systems are being developed. A recent article (Namba et al, 2015) added validity to the use of a wristwatch-like ambulatory sleep monitor (the Watch PAT 200 by Itamar), normally meant for outpatient sleep apnoea studies. That thing measures peripheral arterial tone as a means of estimating autonomic nervous system activity. The utility of these devices in ICU remains to be established; presently it seems like an expensive toy (imagine how many central lines you could buy instead).

Strategies to limit sleep disturbance in the ICU

The best reference for this seems to be the 2015 article by Kamdar et al, which details the implementation of a "multifaceted quality improvement intervention" to improve sleep quality among ICU patients at the Johns Hopkins Hospital Medical ICU. The following list of interventions is largely modelled on their program (see their Table 2); it sounds like a nice program,  even though they did not assess the sleep quality in any objective way.

Non-pharmacological measures to improve sleep in critical illness

• Noise minimisation
  • Turn television off at night
  • Silence unnecessary alarms
  • Chase the family away after hours
• Light level fluctuation to model the day-night rhythm (or, actual daylight!)
  • Raise curtains during the day, lower them at night
• Optimise room temperature
• Minimisation of mechanical ventilation, and the use of patient-triggered modes
• Earplugs, eye masks
• Relaxation techniques (eg, white noise or calming music, biofeedback,  massage)

Pharmacological measures to improve sleep in critical illness

• Minimise caffeine use (this Kamdar study was at an American ICU, which most likely resembles a regular HDU here in Australia; many of the patients are likely to be extubated and demanding coffee)
• Atypical antipsychotics for delirium (as opposed to benzodiazepines)
• Minimise benzodiazepine use
• Zolpidem, apparently
• Use of melatonin (the studies are too few, and too heterogeneous, to make recommendations at this stage)