Introduction
Patients admitted to the intensive care unit generally experience the least sleep of any hospital inpatient, despite being the most ill. This comes with various deleterious outcomes. Currently, the gold-standard of assessing sleep is polysomnography, which describes the recording of multiple simultaneous parameters related to sleep. Polysomnography, while gold-standard, is burdensome in terms of resources and labour. This is especially a challenge in under-equipped rural intensive care units. This study investigates actigraphy, a watch-like device that senses movement, and Bispectral Index, a processed electroencephalogram, in their ability to correctly identify sleep from wakefulness, as compared to polysomnography.
Aim
The study is particularly interested in answering two research questions:
- How sensitive to detecting sleep and specific to detecting wakefulness are actigraphy and BIS guided measures of sleep compared to the gold-standard polysomnography when measuring sleep in the ICU?
- Does actigraphy or BIS present a reduced resource burden compared to polysomnography when assessing sleep in the ICU?
Method/Description
This study used a cross-sectional observational study design. Polysomnography, actigraphy, and Bispectral Index were applied simultaneously on participants recruited from the intensive care unit. The following data was collected by all three devices: total sleep time (TST), sleep efficiency (SE%), and number of awakenings. TST includes any NREM or REM stage of sleep and is expressed in number of hours, SE is the TST expressed as a percentage of total data collection time.
Results
We recruited 3 participants, with 27.43 hours of sleep data. Participants’ recorded data varied from 4.32 to 12 hours. On average, participants slept for a total of 5.2 ± 2.76 hours (M ± SD), SE% was 54.73% ± 8.11% (M ± SD), and a number of awakenings was 70.33 ± 32.58 (M ± SD), as measured via PSG. Sleep was characterised by frequent awakenings and frequent short bouts of sleep. Comparing each 30-second interval of actigraphy and Bispectral Index to polysomnographic output, actigraphy was found to have high sensitivity (92.1%), but a low specificity (44.0%) in distinguishing sleep and wakeful states. Bispectral Index was found to have a comparatively poorer sensitivity (76.3%), and a similarly low specificity (45.4%).
Discussion
There was a large degree of variation in the total sleep time, sleep efficiency, and number of awakenings as measured by actigraphy and Bispectral Index. In general, both actigraphy and Bispectral Index overestimated sleep, though actigraphy was more accurate in correctly identifying sleep state. Clinically, the findings from this study can inform easier ways for intensivists to analyse sleep. Ideally, sleep would be assessed by simple, cost-effective measures, and sleep statistics such as TST, SE%, and awakenings could be presented alongside other patient observations such as blood pressure or oxygen saturation. This way, sleep defects could be detected early, disturbances could be minimised, and teams could respond appropriately. This research provides some insight to the feasibility of BIS and actigraphy as such a device.
Consclusions
In general, both actigraphy and BIS overestimated sleep, though actigraphy was more accurate in correctly identifying sleep from wakefulness. Both actigraphy and BIS were found to be technically feasible in terms of resource and labour burden. Viewed as a pilot study, this research demonstrates the feasibility of investigating low-tech sleep-measuring techniques in a small, rural ICU. Our results will allow intensivists in similar settings without access to polysomnography to apply the relevant techniques discussed in this study.