Direct Measurements of Smartphone Screen-Time: Relationships with Demographics and Sleep

Abstract

Background: Smartphones are increasingly integrated into everyday life, but frequency of use has not yet been objectively measured and compared to demographics, health information, and in particular, sleep quality.

Aims: The aim of this study was to characterize smartphone use by measuring screen-time directly, determine factors that are associated with increased screen-time, and to test the hypothesis that increased screen-time is associated with poor sleep.

Methods: We performed a cross-sectional analysis in a subset of 653 participants enrolled in the Health eHeart Study, an internet-based longitudinal cohort study open to any interested adult (≥ 18 years). Smartphone screen-time (the number of minutes in each hour the screen was on) was measured continuously via smartphone application. For each participant, total and average screen-time were computed over 30-day windows. Average screen-time specifically during self-reported bedtime hours and sleeping period was also computed. Demographics, medical information, and sleep habits (Pittsburgh Sleep Quality Index-PSQI) were obtained by survey. Linear regression was used to obtain effect estimates.

Results: Total screen-time over 30 days was a median 38.4 hours (IQR 21.4 to 61.3) and average screen-time over 30 days was a median 3.7 minutes per hour (IQR 2.2 to 5.5). Younger age, self-reported race/ethnicity of Black and "Other" were associated with longer average screen-time after adjustment for potential confounders. Longer average screen-time was associated with shorter sleep duration and worse sleep-efficiency. Longer average screen-times during bedtime and the sleeping period were associated with poor sleep quality, decreased sleep efficiency, and longer sleep onset latency.

Conclusions: These findings on actual smartphone screen-time build upon prior work based on self-report and confirm that adults spend a substantial amount of time using their smartphones. Screen-time differs across age and race, but is similar across socio-economic strata suggesting that cultural factors may drive smartphone use. Screen-time is associated with poor sleep. These findings cannot support conclusions on causation. Effect-cause remains a possibility: poor sleep may lead to increased screen-time. However, exposure to smartphone screens, particularly around bedtime, may negatively impact sleep.

Fig 1. Geographical Distribution of Participants in the United States.

Abbreviations: AK, Alaska; HI, Hawaii. Dots represent the number of participants that resided in the zip-code corresponding to the placement on the map. All 50 states were represented and 147 (23%) resided in California. Created with Tableau Software (www.tableau.com) and published with permission of the company (S1 File). The U.S. map was used under a CC BY-SA copyright from OpenStreetMap contributors (www.openstreetmap.org/copyright).

Fig 2. Distribution of Screen-Time Over the Day (Hourly Average Screen-Time).

(A) Hourly average screen-time scaled to the maximum within each participant: blue = minimum; red = maximum. Each horizontal line represents data for one participant across 24 hours in a day. (B) Box plots of population summary statistics of hourly average screen-time. Horizontal line within box = median, boxes = IQR, whiskers = 1.5 interquartile range (IQR), dots = outliers.

Fig 3. Associations Between Baseline Survey Data and Average Screen-Time (N = 653).

Abbreviations: BMI, body mass index; AF, atrial fibrillation; CAD, coronary artery disease; CHF, congestive heart failure; HTN, hypertension; OSA, obstructive sleep apnea. Boxes (bivariate) and circles (multivariate) represent point estimates for linear regression coefficients, which correspond to the increase in average screen-time for a unit change in the corresponding variable. Whiskers give 95% confidence intervals. For categorical covariates (race/ethnicity, smoking, activity level) p values for the overall effect of the variable are presented. ^a Factors significantly associated with average screen-time at the p < 0.10 level in bivariate linear models were included in a multivariate linear model. ^b Education and income were both ascertained with 9 levels and analyzed as continuous variables. ^c PHQ-9 score is scaled to a unit increase of 5, the width of each category of depression. ^d Data were available on 267 participants. ^e White circles are regression coefficients adjusted for all other variables in the model.

Fig 4. Associations between Baseline Sleep Quality and Average Screen-Time.

Abbreviations: PSQI, Pittsburg Sleep Quality Index; SD, standard deviation. Diamonds (unadjusted) and circles (adjusted) represent point estimates for linear regression coefficients, which correspond to the increase in average screen-time for a unit change in the corresponding variable. Whiskers give 95% confidence intervals. Each PSQI score was analyzed as a continuous variable. Coefficients for PSQI total score are reported per SD increase, coefficients for Poor sleep and other PSQI component scores are reported per unit increase. ^a PSQI sub-scores range from 0 (good) to 3 (poor) for each component of sleep. The total score is the sum of the sub-scores (0–21). PSQI total score > 5 is a standard dichotomous measure for overall poor sleep. Decreased sleep duration and decreased sleep efficiency correspond to higher component scores. ^b Adjusted for age, sex, race/ethnicity, and history of obstructive sleep apnea.

Fig 5. Associations between Baseline Sleep Quality and Average Screen-Time Within Sleep-Related Hours.

Abbreviations: PSQI, Pittsburg Sleep Quality Index; SD, standard deviation. Among participants with a sleep survey and full screen-time data (N = 56), self-reported bedtime and wakeup-time was used to compute average screen-time (over 30 days) during the hour before bedtime, the hour of bedtime, the hour after bedtime, and during the sleeping period (all hours from bedtime to wakeup-time). All markers represent point estimates for linear regression coefficients after adjustment for age, sex, race/ethnicity, and history of obstructive sleep apnea. Coefficients correspond to the increase in average screen-time, during the indicated period, for a unit change in the corresponding sleep measure. Whiskers give 95% confidence intervals.