Relationships of mobile phone duration and unlock counts with sleep: double machine learning and traditional statistical methods

Background The association between mobile phone use and sleep is still contested, especially for youth. While most existing studies predominantly rely on self-reported smartphone duration using traditional statistical methods, the potential association between mobile phone unlock counts and sleep has been neglected in prior studies primarily due to the difficulty of estimating the data. This relationship has been largely overlooked due to objective measurement challenges. Aims To examine the associations of mobile phone duration and unlock counts with sleep quality and sleep time. Methods This multi-center investigation included 16,668 participants from six Chinese universities. Objective mobile phone duration and unlock counts were assessed by mobile phone use record screenshots. The Pittsburgh Sleep Quality Index was used to assess sleep quality and time. Double Machine Learning (DML), linear regression, and restricted cubic splines (RCS) models were applied. Results DML revealed that a 7-h/week increase in mobile phone duration was associated with higher PSQI scores ( ATE = 0.03, 95% CI : 0.02 to 0.03) and shorter sleep time ( ATE = −4.76, 95%CI : −6.13 to −3.65). Similarly, a 50-time/week increase in unlock counts was linked to increased PSQI scores ( ATE = 0.42, 95%CI : 0.21 to 0.63) and reduced sleep time ( ATE = −0.89, 95%CI : −1.32 to −0.46). RCS models identified non-linear associations of mobile phone duration with poor sleep and sleep time, notably an inverted U-shaped relationship with sleep time ( P for non-linearity < 0.001). Mobile phone unlock counts showed a significant non-linear association with odds of poor sleep ( P for non-linearity < 0.01), but a predominantly monotonic decline with sleep time ( P for non-linearity = 0.50). Conclusion Both mobile phone duration and unlock counts are associated with sleep quality and sleep time. Mobile phone duration exhibited non-linear associations with sleep time and poor sleep, showing an inverted U-shaped relationship with sleep time. Mobile phone unlock counts exhibited a non-linear relationship with sleep time, but a predominantly monotonic inverse relationship with odds of poor sleep.