Sleeping habits have been reported to be associated with diabetes mellitus. This study aimed to explore the relationship of sleep efficiency with diabetes mellitus in individuals with or without sleep-disordered breathing based on polysomnography records.We enrolled participants from the Sleep Heart Health Study. Objective indicators of sleep characteristics including sleep efficiency, sleep latency, slow wave sleep, wake after sleep onset, and total arousal index were monitored via in-home polysomnography. Sleep efficiency was divided into grade 1 (>85%), grade 2 (80%-84.9%), and grade 3 (<80%). Multivariate logistic regression models were utilized to investigate the association between sleep quality and diabetes mellitus.The present study comprised 4737 participants with a mean age of 63.6±11.0 years. The prevalence of diabetes mellitus was higher in those with grade 3 sleep efficiency than that in those with grade 1 and grade 2 sleep efficiency in participants with (10.9% vs. 8.5% vs. 8.3%, respectively; P=0.134) or without (9.5% vs. 5.6% vs. 3.5%, respectively; P<0.001) sleep-disordered breathing. After adjusting for potential confounding factors, sleep efficiency <80% was associated with the prevalence of diabetes mellitus only in participants without sleep-disordered breathing (odds ratio, 1.894; 95% confidence interval, 1.187-3.022, P=0.007).Poor sleep efficiency was associated with diabetes mellitus in those without sleep-disordered breathing. Therefore, the relationship between sleep efficiency and diabetes mellitus was worth further investigation. This article is protected by copyright. All rights reserved.