Abstract: Objective: To systematically evaluate whether interventions targeted at improving sleep in intensive care units (ICUs) were associated with reduction in ICU delirium. Methods: We searched the Cochrane Library, PubMed, Embase, Web of Science, VIP, CNKI and Wanfang Data for randomized controlled trials on sleep interventions and delirium both at home and abroad till December 31, 2018. Besides, all the references referred in those literatures including primary ones and literature reviews were also searched. Two researchers were responsible to screen literatures in accordance with the inclusive and exclusive criteria, extract data and assess the methodology quality of those included literatures separately. The meta-analysis and sensitivity analysis were conducted using RevMan 5.3 software. Results: A total of fourteen studies involving 2548 patients were included, of which four trials took dexmedetomidine as the intervention, four light therapy and the rest ramelteon, earplug-using, simulating human circadian rhythm to produce sedation and nondrug bundle of care. Dexmedetomidine, according to the meta-analysis, could reduce the risk of delirium (OR=0.36, P<0.00001), but not the length of stay in ICUs (SMD= -0.25, P=0.25). Simulating human circadian rhythm to produce sedation showed a beneficial effect (OR=0.25, P=0.005). Light therapy and earplug-using were not associated with risk of delirium in ICUs, respectively, and light therapy wasn't associated with shorter ICU length of stay either. Conclusion: Dexmedetomidine and simulating humam circadian rhythm to produce sedation may have beneficial effects on reducing the incidence of delirium in critically ill patients. However, in consideration of the number and quality of the included studies, the conclusion needs to be further validated and supported by multi-centered randomized controlled trials with a large sample size.

Key words: delirium; sleep; critical illness; systematic review