SleepApp

Knowledge Discovery in Ubiquitous and Personal Sleep Tracking: Scoping Review

“…The Pittsburgh Sleep Quality Index was the most widely used questionnaire to measure the subjective perception of sleep duration, continuity, efficiency, and satisfaction [ 31 , 38 ]. The SleepAsAndroid app was used to collect sleep data based on user movement patterns in the study by Daskalova et al [ 20 ], and the SleepApp was used to collect users’ subjective sleep quality ratings in the study by Ravichandran et al [ 19 ]. Although sleep tracking apps such as SleepAsAndroid were widely downloaded and used by millions of users, their ability to distinguish between quiet awakenings, deep sleep, or empty beds was still limited [ 20 ].…”

“…Along with sleep quality metrics, researchers have considered a wide range of contextual factors in their studies. As presented in Table 3 , the contextual factors of interest include the sleep environment [ 16 , 20 , 42 ], daily activities [ 16 , 19 , 20 , 31 , 36 , 37 , 40 , 42 , 43 ], physiological states [ 16 , 35 , 36 , 43 , 44 ], and mental states [ 16 , 19 ]. It was a common practice to collect demographic information (eg, age, gender, and BMI) and medical history using questionnaires at the beginning of a sleep tracking study [ 16 , 35 , 36 , 43 , 44 ].…”

“…These studies developed their own tools to track users’ web-based behavior and linked them to sleep quality metrics. Using consumer wearable trackers such as Fitbit, MS Band, and Garmin, researchers were able to expand their list of factors to include, for example, bedtime, steps, distance, hours of exercise, and calories burned [ 16 , 19 , 20 , 31 , 36 , 37 , 40 , 42 , 43 ]. Biosignals such as heart rate during sleep and daytime activities were included in the studies by Faust et al [ 35 ], Farajtabar et al [ 43 ], and Choi et al [ 44 ].…”

“…Biosignals such as heart rate during sleep and daytime activities were included in the studies by Faust et al [ 35 ], Farajtabar et al [ 43 ], and Choi et al [ 44 ]. Several studies have also manually collected input features such as coffee, alcohol, mood, and stress [ 16 , 19 , 20 ]. These factors may have a significant effect on the circadian cycle of hormone secretion and thus may provide useful information for sleep quality prediction.…”

“…Although there are many methods for data cleaning in the literature, the methods adopted in the reviewed publications are extremely simple and straightforward. Most of the studies (5/14, 36%) simply excluded records that contain missing values (eg, sleep logs with missing fields or sleep duration of 0 minutes) [ 16 , 19 , 37 ] or excluded users who do not contribute sufficient data (eg, fewer than 30 sleep records) [ 35 , 36 ]. One study excluded a certain data source (eg, search engine interactions originating from mobile devices) to avoid causing distortion to the data distribution [ 34 ].…”