Identifying a feasible objective physiological assessment of sleep is necessary to understand the impact of sleep disturbance on intensive care patients. This study undertook a comparison of sleep monitoring techniques to assess sleep and wakeful states using the accepted gold standard of sleep monitoring; polysomnography and motion accelerometer monitoring via actigraphy. Findings suggest that ACTG has a moderate level of sensitivity and specificity in identifying binary states of sleep and wakefulness, compared to PSG. Ventilation status (mechanical ventilation: non-mechanically ventilated) of patients was found to impact on the accuracy of the ACTG in distinguishing sleep and wake states. However, this was not found to be attributed to the administration of opioid or benzodiazepine infusions, rather higher APACHEII scores were linked to the need for mechanical ventilation. ACTG was found to have moderate to high levels of correlation with PSG in identifying sleep duration and wakefulness in non-ventilated patients.
Sleep is increasingly recognised as an important factor in patients’ overall well-being and recovery, resulting in its integration into the Clinical Practice Guidelines for the Preventions and Management of Pain, Agitation/Sedation and Delirium developed by the Society of Critical Care Medicine . Despite its inclusion, the ability for clinicians to feasibly monitor sleep and evaluate the effectiveness of sleep protocols and interventions has been problematic. However, our findings suggest ACTG may provide a surrogate method that can be clinically applicable amongst select ICU patients. This study is one of the first studies to report the feasibility of ACTG as a sleep monitoring technique within ICU, to assess sleep duration and disruption compared to PSG. Findings indicate that ACTG provides a clinically viable method for non-ventilated patients that is minimally invasive, easy to establish unattended monitoring and interpret findings .
Previous research in non-critical care contexts has validated ACTG against PSG (level of agreement >90%) in distinguishing sleep and wake states [33-34]. However, its application in critically ill patients is limited, with previous research conducted by Beecroft et al  and van der Kooi et al  reporting low levels of agreement between PSG and ACTG. The findings of this study contest these and report an overall moderate level of agreement (69.4%), and accuracy to identify sleep (65.5%) and wakeful states (76.1%). The non-ventilated patient subset were found to have a improved level of agreement, which may be attributed to lower APACHEII scores (<20), a reduction in the administration of opioids and benzodiazepines and RASS scores within the cohort. Sensitivity and specificity were found to vary considerably based on ventilation status with an increased specificity to identify wakeful states in non-ventilated patients and an increased sensitivity towards identifying sleep in mechanically ventilated patients. ACTG accuracy appears to decline when physiological activity is reduced which is consistent with mechanically ventilated patients. The findings of this study suggest that higher APACHEII scores may be an impacting factor on the accuracy of ACTG compared to PSG.
ACTG monitoring of non-ventilated patients reported a higher level of correlation to PSG in detecting wakeful and sleep states. This may be related to this patient demographic having more distinct physiological activity consistent with increased movement during wakefulness, with distinct reductions during sleep. In contrast, ACTG was found to correlate to PSG in reporting sleep states amongst mechanically ventilated patients, with a poor ability to detect wakefulness. This patient cohort are the greatest challenge in monitoring as spontaneous movement is frequently reduced due to pharmacological agents, higher clinical acuity, and atypical EEG activity which may be present in patients with higher APACHII scores. As a result, ACTG’s inability to distinguish between sleep and diminished physiological activity during wakeful states contributes to an erroneous interpretation. Mechanically ventilated patients within this study were assessed has having prolonged total sleep times, with limited arousals and spontaneous movements detected by both PSG and ACTG. The findings suggest that there is considerable variability between the two measures with ACTG having a modest level of agreement with PSG. The clinical application of ACTG for sleep monitoring may be more appropriate for non-ventilated patients cared for in the Intensive Care, whose wake states are observable.
Actigraphy was found to overestimate sleep and underestimate wakefulness which reiterates previously published findings [5,19,35]. The sleep architecture of patients in this study was found to be atypical despite limited administration of benzodiazepines and less than 50% of the recruited patients requiring mechanical ventilation. Sleep was comprised of primarily stage N2 sleep and highly fragmented, with limited restorative components identified (SWS and REM sleep), which has been previously reported amongst ICU patients. Actigraphy’s overestimation of sleep has been purported to be related to its lack of sensitivity in distinguishing between sleep, motionless resting activity and movement occurring during sleep or the provision of clinical care . Specifically, how the commencement of sleep is recognised may be a critical factor impacting on accuracy. With ACTG denoting immobility as the beginning phase of a sleep period, whereas PSG identifies changes in neuroelectrophysiological activity patterns marking the onset of sleep. As these changes commence well after a period of wrist immobility, actigraphy is frequently reported to overestimate sleep time. This finding has been reported to be more pronounced in those patients with abundant wakefulness through the night . The presence of sedating medication often effects EEG characteristics, which limited the accuracy of distinguishing between traditional sleep stages. When this occurred stage N2 was used as a default stage to indicate sleep.
The findings of this study affirm the challenges encountered in monitoring sleep via PSG within this cohort. As sleep traverses across the 24-hour spectrum maintaining electrode placement over this duration is challenging as identified in this study. Further, PSG tethers patients to additional invasive monitoring, and increases the logistical complexities of providing patient care and increases patient discomfort. Technical issues included both poor signal quality due to electrode instability, which was exacerbated by the provision of clinical care in manoeuvring patients on and off the bed, as well as data loss due to the implications of the complexity of recording PSG. These factors frequently resulted in studies with usable data of less than 14 hours, and highlight that PSG is suitable for widespread clinical implementation. As a result, it is imperative that assessment methods are identified that are both feasible and capable of prolong monitoring and, imposes a limited burden to the patient and does not further increase the complexity of providing patient care is required. Recording tolerance was identified as an issue amongst conscious patients with requests to terminate monitoring upon waking from the night of sleep. This was consistent with findings reported by Knauret et al  who reported 31% of medical ICU patients experience issues with monitoring tolerance related to discomfort. The combination of technical issues and patient acceptability of PSG monitoring reiterates that it is an unfavourable approach for widespread sleep monitoring in the ICU and may exacerbate sleep disturbance as a result.
Acknowledging the complexities of PSG monitoring and the limitations of ACTG in terms of patients with limited mobility and higher sedation scores are important considerations when implementing current available sleep monitoring techniques. Judicious assessment of patients should be considered when implementing ACTG for ICU sleep monitoring. In this study the devised exclusion criteria were developed based on previously reported sleep research, and the identified factors that can confound electroencephalographic activity and interpretation. The sedation level primarily reported in this study was a RASS score of zero (calm and cooperative), limited benzodiazepine administration, reduced requirement for invasive mechanical ventilation support, and lower APACHII scores may be critical factors that aided in ACTG’s ability to track patients sleep-wake patterns more accurately. As a result, ACTG monitoring was not implemented in clinical cases where it would have been deemed futile secondary to immobility and high sedation levels.
Sleep disturbance is recognised as a clinical issue that can negatively affect the rehabilitation of ICU patients and their engagement in activities aimed at expediating their recovery . Furthermore, functional, and cognitive recovery has been reported to be impeded amongst patients who are subjected to sleep disturbance and fragmentation [36-37]. The important role sleep has in functional recovery is supported by clinical research which indicates that addressing sleep disturbance amongst critically ill patients leads to reduced disability at discharge [38-39]. Patients capable of spontaneous movement appear to be vulnerable to high levels of sleep disturbance and are potentially a target population within the ICU for sleep promoting interventions. Although, ACTG will not provide detailed information on sleep architecture, its potential application for monitoring sleep characteristics for those with protracted ICU admission may be of value, in that it can track sleep-wake cycles over extended durations, along with assessing the potential response to sleep supporting interventions.
The findings reported in this study need to be considered before extrapolating them more widely, acknowledging that the research involved a single study centre, and as a result the findings may not be transferrable to other Intensive Care Units who have their own unique features and patient demographics. The research site employs a minimal sedation and early mobilisation practice in the provision of care, and as a result, patients may have a lower RASS scores making patients more aware and interactive with their environment.
The implemented exclusion criteria for the study prohibited the recruitment of the most critically ill patients from participating in the study, and the findings cannot be transferred to this patient cohort. The ability to effectively monitor sleep in heavily sedated patients remains clinically challenging, with PSG likely to be the most appropriate assessment method currently available. It may be contended that in most critically ill patients within the ICU the focus of care is primarily on preserving life, and the sleep quality and quantity may not be a critical factor at this point of care. Rather, sleep may be a critical factor in the transitioning of care to recovery, weaning from mechanical ventilation and rehabilitation from critical illness whereby, sleep disturbance and deprivation may contribute to the potential onset of delirium [24, 40-43]. It is amongst this cohort, that ACTG may be most appropriate to implement. ACTG monitoring is not infallible and its limitations need to be acknowledged in terms of its potential to distinguish between voluntary and involuntary movements. Whereby, patients’ wakeful inactivity may be identified as rest or sleep, and activity associated with care interventions such as repositioning may be inaccurately identified as a wakeful state .