We conducted a study to examine the association between objectively-measured smartphone usage, sleep quality, and physical activity level among Hong Kong Chinese adolescents and young adults. There were several advantages to using the objective measurement of smartphone usage. First, we could identify different purposes for which smartphone was being used, and showed the effect of smartphone use on different aspects of life. For instance, we found that only the use of social network apps and games and comics apps was associated with total sleeping time. Second, together with sleeping parameters objectively measured with accelerometer, we could define sleeping time and bedtime smartphone usage objectively, and exclude the possibility of reverse causation, whereby people with sleep problems would use their smartphone during bedtime as a way of spending their energy [24]. Third, while the validity of self-reported smartphone usage was questionable, the source of measurement error could be eliminated through objective measurement of smartphone usage.
In studies on the association between smartphone usage and sleep quality, most used a self-reported questionnaire to measure sleep quality and only three studies measured sleep quality with a wrist-worn accelerometer [25–27]. Surprisingly, all three studies found no association. On the one hand, the finding of a null association might be a true finding; on the other hand, it is possible that these studies committed a type II error because the sample sizes ranged from 55 to 110 only. With a repeated measure of 814 nights in a study involving 187 subjects, ours has the largest in sample size and showed that both daytime and bedtime smartphone usage were associated with total sleeping time but not with sleep efficiency and WASO.
Several researchers attempted to differentiate between the physiological effects of daytime and bedtime smartphone usage. Compared to other devices with electronic screens such as televisions and video game consoles, the smartphone is more commonly used in bed as they are portable and are usually placed very close to the user [28, 29]. To the best of our knowledge, all existing studies have relied on self-reported data to determine bedtime smartphone usage. As the validity of self-reports of bedtime smartphone usage requires the accurate recalling of both bedtime and smartphone usage, we questioned the validity of self-report and used objective measure of sleep onset and smartphone usage to determine bedtime smartphone usage. The invalidity of self-reported bedtime smartphone usage is shown by the disagreement of our results with other existing self-reported results; for example a Taiwanese study using self-reported data showed that bedtime smartphone usage was very common (95%) among junior college students [30], but the current study recorded such usage in less than half of the nights (42.4%). Our results also highlighted the importance of distinguishing between daytime and bedtime smartphone usage, as total sleeping time was negatively related to daytime usage but not related to bedtime usage. We found no association between smartphone usage, both daytime and bedtime, and sleeping efficiency, and these results did not agree with some previous findings [31–34]. We believe that the effects of smartphone usage and sleep outcomes are still controversial (some other studies found a null association [35]) and warrant further investigations, especially through employing well-designed controlled trials. We suspect that, in previous studies, participants might mixed up bedtime smartphone usage and usage during sleep, as we found that smartphone usage during sleep did improve sleep efficiency. To the best of our knowledge, very few randomized controlled trials have been conducted in this direction, and some found a null association between smartphone usage and sleep outcomes due to small sample size [36].
Besides the time of use, the purpose for which a smartphone is used may also have an impact on sleep outcomes. This study found that only using a smartphone for social network apps and games and comics apps (with marginally-significant p-value of 0.053 among secondary school students), but not for other types of apps, was associated with total sleeping time. It was hypothesized that interactive smartphone use should have stronger effects on sleep than the passive viewing of a smartphone screen [37]. Our results only partially agreed with this hypothesis. Social network app usage can be regarded as passive viewing as its use does not involve frequent touching of the phone. However, the contents of social network are updated extremely quickly and new updates can be expected to occur every minute. Therefore, users may be anxious about missing out on new content, and such anxiety may cause users to stay awake during the night [38]. Yet, some studies found that social network use was, in fact, positively related to sleep quality [39], partially because social network sites can help users to connect, make new friends, and share their stress [39]. In addition to social network apps, instant messaging apps were also commonly used among our subjects, and we found no association between its use and all sleep outcomes. It was hypothesized that it is difficult to top oneself from messaging, which thereby delays sleep onset, and this was supported by studies showed that texting was negatively associated with sleep quality [40] and its effect was the strongest on sleep duration among other electronic screen usages [41]. However, we questioned the validity of these results as they relied on self-reported time spent on instant messaging and such data, being recalled data, is inherently unreliable. Our data showed that instant messaging apps were used frequently, at an average of 363.7 times per day and that each session only lasted for an average of 1.7 minutes. We believe that the frequent and short usage nature of instant messaging apps made the recalling of time spent on them challenging.
It was generally expected that the use of a smartphone would lead to a sedentary and inactive lifestyle due to its sedentary nature [7, 42]. Surprisingly, our data showed that social network, instant messaging, tools, and multimedia app usage at daytime were all positively associated with physical activity level. We believe that there was a bi-directional causation. Those who were more physically active might have spent more time on social network and instant messaging apps to share their photos and information. Similarly, some of the multimedia app usage might have been occurred concurrently with physical activity (e.g., playing a video during jogging on a treadmill). On the other hand, our data found no association between games and health app usage and physical activity level. Recently, some augmented reality games developed for smartphone have encouraged walking in the real world (e.g., Pokémon GO), and there is evidence to show that their use is promoting engagement in physical activity [43, 44]. In the current study, too few subjects played Pokémon GO (11 out of 187, 5.9%); therefore, the health-promoting effect of this kind of augmented reality game might have been diluted by the sedentary effect of other kinds of games. Similarly, users of health-related apps, for instance pedometer apps, were found to be physically more active than non-users [45, 46]. In the current study, too few subjects used health-related apps (31 out of 187, 16.6%); less than the previously reported percentage (34.1%) in a US sample [46]. Most importantly, no secondary school students in our sample used any kind of health-related apps during the measurement period, therefore, we were unable to detect the associations between the usage of health-related apps, sleep quality, and physical activity level.
The major limitation of this study lied in the area of sampling and data collection. Only Android users were recruited in this study and cautions should be taken when our results are generalized to users of other smartphones. This also explains the small number of adolescent participants in our sample; most of the students in our recruited secondary schools were using an iPhone that is not supported by our tracking app. In our sample, about 10% of the participants owned more than one smartphone. Since the smartphone usage tracking app was only installed in one smartphone of the subjects, we might have underestimated the total smartphone usage. Another limitation was that our smartphone tracking app could not record the usage of some Android default system apps, for example Google Chrome, the default web browser of Android system, in some versions of Android operating system. A minor limitation was that different recruitment approach was used in secondary school and universities that might affect the generalizability of our results.