Assessment of Changes in Child and Adolescent Screen Time During the COVID-19 Pandemic

Key Points

Question

To what extent has the COVID-19 pandemic been associated with changes in the duration, content, and context of daily screen time among children and adolescents globally?

Findings

In this systematic review and meta-analysis of 46 studies including 29 017 youths (≤18 years), pooled estimates comparing estimates taken before and during the COVID-19 pandemic revealed an increase in screen time of 84 min/d, or 52%. Screen time increases were highest for individuals aged 12 to 18 years and for handheld devices and personal computers.

Meaning

This study shows an association between the COVID-19 pandemic and increases in screen time; practitioners and pandemic recovery initiatives should focus on fostering healthy device habits, including moderating use, monitoring content, prioritizing device-free time, and using screens for creativity or connection.

Abstract

Importance

To limit the spread of COVID-19, numerous restrictions were imposed on youths, including school closures, isolation requirements, social distancing, and cancelation of extracurricular activities, which independently or collectively may have shifted screen time patterns.

Objective

To estimate changes in the duration, content, and context of screen time of children and adolescents by comparing estimates taken before the pandemic with those taken during the pandemic and to determine when and for whom screen time has increased the most.

Data Sources

Electronic databases were searched between January 1, 2020, and March 5, 2022, including MEDLINE, Embase, PsycINFO, and the Cochrane Central Register of Controlled Trials. A total of 2474 nonduplicate records were retrieved.

Study Selection

Study inclusion criteria were reported changes in the duration (minutes per day) of screen time before and during the pandemic; children, adolescents, and young adults (≤18 years); longitudinal or retrospective estimates; peer reviewed; and published in English.

Data Extraction and Synthesis

A total of 136 articles underwent full-text review. Data were analyzed from April 6, 2022, to May 5, 2022, with a random-effects meta-analysis.

Main Outcomes and Measures

Change in daily screen time comparing estimates taken before vs during the COVID-19 pandemic.

Results

The meta-analysis included 46 studies (146 effect sizes; 29 017 children; 57% male; and mean [SD] age, 9 [4.1] years) revealed that, from a baseline prepandemic value of 162 min/d (2.7 h/d), during the pandemic there was an increase in screen time of 84 min/d (1.4 h/d), representing a 52% increase. Increases were particularly marked for individuals aged 12 to 18 years (k [number of sample estimates] = 26; 110 min/d) and for device type (handheld devices [k = 20; 44 min/d] and personal computers [k = 13; 46 min/d]). Moderator analyses showed that increases were possibly larger in retrospective (k = 36; 116 min/d) vs longitudinal (k = 51; 65 min/d) studies. Mean increases were observed in samples examining both recreational screen time alone (k = 54; 84 min/d) and total daily screen time combining recreational and educational use (k = 33; 68 min/d).

Conclusions and Relevance

The COVID-19 pandemic has led to considerable disruptions in the lives and routines of children, adolescents, and families, which is likely associated with increased levels of screen time. Findings suggest that when interacting with children and caregivers, practitioners should place a critical focus on promoting healthy device habits, which can include moderating daily use; choosing age-appropriate programs; promoting device-free time, sleep, and physical activity; and encouraging children to use screens as a creative outlet or a means to meaningfully connect with others.

This systematic review and meta-analysis assesses changes in the duration, content, and context of daily digital device use among children and adolescents by comparing estimates made before and during the COVID-19 pandemic.

Introduction

To limit the spread of the COVID-19 virus, numerous restrictions were imposed on the daily lives of children and adolescents globally, including repeated school closures, cancellation of extracurricular activities, social and physical distancing from peers and other sources of interpersonal support (eg, teachers and coaches), and mandated home quarantining due to COVID-19 exposure. Parents, in parallel, also experienced substantial challenges, including financial instability, job insecurity, loss of child care, and increased home-schooling responsibilities, which individually and collectively resulted in increased family stress and mental distress.^1,2,3 To cope with such unparalleled disruptions to normal living conditions, many children and families likely used digital devices to occupy their time during the pandemic. Population-level increases in child and adolescent screen time have therefore been expected.^4,5 Trajectories of screen use demonstrate that children with high screen use often remain high users throughout preschool and middle childhood.^6,7 Meta-analyses have also documented significant associations of child screen time with poor sleep,⁸ physical activity,⁹ language and communication skills,¹⁰ mental health,¹¹ and academic¹² outcomes. Up to 80% of apps for children are also purposely built with manipulative design features (eg, fabricated time pressure, gifts, and attractive lures to encourage longer gameplay),¹³ which can be persuasive in maintaining children’s attention. Therefore, a critical time-sensitive research focus should be to determine the degree to which child and adolescent screen time increased during the COVID-19 pandemic in terms of the duration of use as well as the content and context of use.

Although most empirical studies suggest that screen time increased during the pandemic, there is considerable variability in the direction and magnitude of change between studies. For example, Welling et al¹⁴ reported no significant changes, Morrison et al¹⁵ reported a decrease of 15 min/d, and McArthur et al⁴ and Pietrobelli et al¹⁶ reported increases of 102 min/d and 292 min/d, respectively, before vs during the pandemic. Thus, there is a need to explain between-study variability in COVID-19–associated changes in screen time. The variation in design affordances across devices and platforms, such as their mobility and intended use, may yield variations in the patterns of change across device type. With more than 1.5 billion children worldwide moving to online school at the outset of the pandemic,¹⁷ context of use should also be examined because screen time could have increased for educational use.

One expected, developmentally relevant moderator of changes in screen time is child age because screen time increases across childhood.^18,19 Variability could also be sex specific, with studies showing that screen time is higher for boys than for girls,^19,20,21 and informant dependent because youths (vs parents) may be more reliable estimators of their own behavior.^11,22 Between-study variability may also be associated with the populations under investigation, such as children and adolescents with medical (eg, obesity) or clinical (eg, autism spectrum disorder) diagnoses who may have been prone to receiving or requesting more screen time.^23,24,25,26 Another source of heterogeneity could be study design, with some studies providing longitudinal change in cohorts of children by comparing pandemic data with historical prepandemic data, whereas other studies were cross-sectional and asked participants to retrospectively recall prepandemic screen time (an approach prone to recall bias).²⁷ Finally, government-mandated restrictions and their seasonal timing varied across countries, which could have affected estimates across studies.

The objectives of this study were to conduct a systematic review and meta-analysis of global changes in child and adolescent screen time before vs during the COVID-19 pandemic and to determine the degree to which these changes differed across devices, context of use, age groups, sexes, devices, population types, methods, and region and season (ie, geographic latitude). Together, these objectives can inform practitioners, programs, and policies seeking to put child and adolescent sedentary behaviors at the forefront of global pandemic recovery efforts.

Methods

Search Strategy

In this meta-analysis, 4 electronic databases (MEDLINE, Embase, PsycINFO, and the Cochrane Central Register of Controlled Trials) were searched for studies published between January 1, 2020, and March 5, 2022. Search strategy terms included screen time, sedentary behavior, and COVID-19 (eTable 1 in the Supplement). Retrieved studies were imported into Covidence,²⁸ where duplicates were automatically removed. Reference lists of included studies and relevant systematic reviews were also hand searched. This review was registered as a protocol with PROSPERO (CRD42022320709).

Selection Criteria

Study inclusion criteria were reported changes in the duration (minutes per day) of screen time before and during the COVID-19 pandemic within the same group of children; children, adolescents, and young adults (≤18 years); longitudinal or retrospective study; peer reviewed; and published in English. Exclusion criteria were case studies, reports, and qualitative analyses. Study inclusion was determined by 2 independent coders (S.M. and P.P.), who coded all titles or abstracts in Covidence (mean random agreement probability, 93%). Independent coders (S.M. and P.P.) reviewed all full-text articles against the inclusion criteria. Discrepancies were resolved via consensus.

Data Extraction

Changes in the duration of daily screen time before vs during the pandemic were extracted from each study. Inferential statistics (P value, z score, t value, and CI) were extracted to calculate the SE of these changes. When studies included male and female individuals, separate subsample data were extracted to account for heterogeneity arising from real differences in screen use between sexes. Data extraction was conducted by 2 coders (P.P. and R.D.N.). Intercoder agreement was 94%.

Moderators

Continuous moderators were baseline (prepandemic) screen time (minutes per day), number of months between assessments of screen time, sample geographic latitude, and study quality. Categorical moderators were device type or content (handheld device use, personal computers, television, videogaming, and social media), content (recreational and recreational plus educational [ie, total]), age group (preschool [≤5 years] and primary school [>5 to ≤12 years], and secondary school [>12 to ≤18 years]), sex (percentage of female individuals), study design (longitudinal or retrospective), informant (parent or youth), and population (clinical [autism spectrum disorder and psychiatric patients, k = 4] vs nonclinical; medical [obesity and diabetes, k = 16] vs nonmedical samples, where k is the number of sample estimates).

Study Quality

Study quality was assessed with items from the National Institutes of Health Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies.²⁹ Each study received a score of 0 (criterion unmet) or 1 (criterion met) for 11 quality indicators, which were tallied to give a quality score from 0 to 11 (eTable 2 and eTable 3 in the Supplement). The study followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline. Data for this study were freely available through published studies.

Data Analysis

Random-effects meta-analyses³⁰ were conducted in SAS, version 9.4 (SAS Institute Inc) from April 6 through May 5, 2022. The inverse square method was used to weight sample estimates.³¹ Between-sample heterogeneity was summarized with the τ statistic, representing the typical differences in the meta-analyzed mean between samples. Effect sizes were calculated by following the Cohen³² principle of standardization (ie, by dividing outcomes by their respective between-person SD of pre–COVID-19 screen time). Standardized thresholds for small, moderate, large, and very large effect sizes were 0.2, 0.6, 1.2, and 2 SDs, respectively.³³ Sampling uncertainty is represented as 90% CIs. Precision of estimation³³ was deemed inadequate or unclear when the 90% CI included substantial positive and negative values (ie, −0.2 and 0.2 SDs, respectively). When the 90% CI included both trivial and substantial (positive or negative) values, the outcome was interpreted as “possibly” substantial. Publication bias and potential outliers were evaluated with the random-effects output (ie, the random-effect solutions for each sample estimate) from the meta-analytic model described earlier. Publication bias was evaluated with a scatterplot of the random-effect solutions and the SEs for each sample estimate. Potential outliers were detected when the P value for the random-effect solution was less than a threshold given by P < .05 divided by the degrees of freedom for the sample estimate random-effect solution in question.

Results

Our search strategy produced 2474 nonduplicate records, and 136 underwent full-text review (Figure 1). Forty-six studies met the full inclusion criteria, with 146 available estimates. Of the 146 estimates, 87 represented changes for all devices combined, 20 for handheld devices, 13 for personal computers, 11 for television, 9 for video gaming, and 6 for social media.

Figure 1. PRISMA Flow Diagram Detailing Search Strategy.

Study Characteristics

Across the 46 studies (Table 1),^{4,14,15,16,24,25,26,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72} 29 017 children and young adults aged 18 years or younger were represented (57% male and 43% female). The mean (SD) age was 9 (4.1) years. Only 9 (20%) of the studies included in this meta-analysis reported data on the race or ethnicity of their sample, and the data on racial and ethnic categories among these 9 studies were inconsistently reported. Studies used parent-reported (29 studies [63%]) or child-reported (17 studies [37%]) data. In terms of context of use, 29 studies reported changes in recreational screen use (17 studies for recreational plus education use). Most studies (28 [61%]) reported longitudinal estimates of change in screen time; the remaining 18 studies (39%) were retrospective estimates of prepandemic data. Of the 46 included studies, 14 were from Asia (30%), 12 from Europe (26%), 12 from North America (26%), 3 from Australia or New Zealand (7%), 2 from South America (4%), and 2 from the Middle East (4%), and 1 study (2%) had pooled data from multiple countries. The mean study quality score was 6.8 (range, 3-9) (eTable 3 in the Supplement).

Table 1. Characteristics of the Included Studies.

Source	Country	No.a	SES	Male, %	Female, %	Age, y	Study design	Screen type	Screen purpose	Clinical condition	Medical condition
Aguilar-Farias et al,34 2021	Chile	3157	Low	51	49	3.1	Retrospective	AD	R	NOCLIN	NOMED
Beck et al,35 2021	US	145	NR	45	55	8.0	Retrospective	AD	R	NOCLIN	MED
Brzęk et al,36 2021	Poland	1316	NR	45	55	3.0	Longitudinal	AD	R	NOCLIN	NOMED
Burkart et al,37 2022	US	127	Diverse	53	47	9.8	Longitudinal	AD	R	NOCLIN	NOMED
Cardy et al,38 2021	Canada	414	Middle-upper	31	69	11.7	Retrospective	AD	R + ED	CLIN	NOMED
Chen et al,39 2021	China	535	NR	49	51	10.3	Longitudinal	SM, VG, HD	R	NOCLIN	NOMED
Cheng et al,40 2021	Malaysia	123	NR	53	47	11.1	Retrospective	AD	R + ED	NOCLIN	MED
Eales et al,41 2021	US	129	Middle-upper	50	50	6.1	Longitudinal	AD	R + ED	NOCLIN	NOMED
Garcia et al,26 2021	US	9	NR	89	11	16.9	Longitudinal	AD	R	CLIN	NOMED
Ghanamah and Eghbaria-Ghanamah,42 2021	Israel	382	NR	51	49	8.0	Retrospective	AD	R + ED	NOCLIN	NOMED
Hossain et al,43 2021	Bangladesh	35	NR	56	44	4.5	Longitudinal	AD	R	NOCLIN	NOMED
Hu et al,44 2021	US	129	Diverse	40	60	10.9	Retrospective	AD	R	NOCLIN	MED
Jáuregui et al,45 2021	Mexico	631	Middle-upper	53	47	3.0	Retrospective	AD	R + ED	NOCLIN	NOMED
Jia et al,46 2021	China	2146	Diverse	0	100	17.5	Longitudinal	AD	R + ED	NOCLIN	NOMED
Kim et al,47 2021	Japan	290	Diverse	52	48	4.8	Longitudinal	T, HD	R + ED	NOCLIN	NOMED
Kim et al,48 2021	Japan	171	NR	57	43	9.7	Longitudinal	T, HD, AD	R	NOCLIN	NOMED
López Gil et al,49 2021	Spain	1099	NR	50	50	4.2	Retrospective	AD	R	NOCLIN	NOMED
López-Bueno et al,50 2020	Spain	860	NR	51	49	4.0	Retrospective	AD	R + ED	NOCLIN	NOMED
Ma et al,51 2021	China	208	NR	53	47	8.9	Longitudinal	AD	R + ED	NOCLIN	NOMED
Maheux et al,52 2021	US	704	Diverse	48	52	15.1	Longitudinal	SM	R	NOCLIN	NOMED
Maltoni et al,53 2021	Italy	51	NR	100	0	14.7	Longitudinal	AD	R	NOCLIN	NOMED
McArthur et al,4 2021	Canada	1333	Middle-upper	53	47	9.5	Longitudinal	AD	R	NOCLIN	NOMED
Medrano et al,54 2021	Spain	106	Diverse	49	51	12.0	Longitudinal	AD	R	NOCLIN	NOMED
Mirhajianmoghadam et al,55 2021	US	38	NR	50	50	8.1	Retrospective	AD	R + ED	NOCLIN	NOMED
Mohan et al,56 2021	India	217	Diverse	47	53	13.5	Retrospective	AD	R + ED	NOCLIN	NOMED
Moore et al,57 2021	Canada	1526	Middle-upper	48	52	8.1	Longitudinal	AD	R	NOCLIN	NOMED
Morrison et al,15 2021	Slovenia	62	NR	50	50	12.0	Longitudinal	AD	R	NOCLIN	NOMED
Nathan et al,58 2021	Australia	121	NR	54	46	7.0	Retrospective	AD	R	NOCLIN	NOMED
Ng et al,59 2021	Hong Kong	64	NR	NR	NR	4.4	Longitudinal	AD	R	NOCLIN	NOMED
Delisle Nyström et al,60 2020	Sweden	100	NR	58	42	4.0	Longitudinal	AD	R	NOCLIN	NOMED
Okely et al,61 2021	Global	852	NR	39	61	4.0	Longitudinal	AD	R	NOCLIN	NOMED
Ostermeier et al,62 2021	Canada	48	Middle-upper	51	49	11.0	Longitudinal	AD	R	NOCLIN	NOMED
Peddie et al,63 2021	New Zealand	35	Diverse	100	0	16.6	Longitudinal	AD	R + ED	NOCLIN	NOMED
Pietrobelli et al,16 2021	Italy	32	NR	50	50	12.8	Longitudinal	AD	R	NOCLIN	NOMED
Rebelo et al,25 2021	Portugal	47	NR	91	9	5.5	Retrospective	AD	R + ED	NOCLIN	NOMED
Ribner et al,64 2021	Global	2516	NR	52	48	5.8	Retrospective	HD, SM, AD	R	NOCLIN	NOMED
Saxena et al,65 2021	India	1237	NR	59	41	11.9	Longitudinal	T, PC, HD	R	NOCLIN	NOMED
Schnaiderman et al,66 2021	Argentina	267	NR	47	53	11.1	Retrospective	AD	R + ED	NOCLIN	NOMED
Schmidt et al,67 2020	Germany	1711	NR	50	50	4.5	Longitudinal	AD	R	NOCLIN	NOMED
Seo et al,24 2021	South Korea	147	NR	83	17	9.5	Retrospective	SM, VG, AD	R	CLIN	NOMED
Shoshani and Kor,68 2021	Israel	1537	NR	48	52	13.9	Longitudinal	SM, VG, AD	R	NOCLIN	NOMED
Ten Velde et al,69 2021	Netherlands	131	NR	33	67	10.1	Retrospective	AD	R + ED	NOCLIN	NOMED
Welling et al,14 2022	Netherlands	54	NR	48	52	11.2	Longitudinal	AD	R	NOCLIN	MED
Xiang et al,70 2020	China	2426	NR	NR	NR	12.0	Longitudinal	AD	R + ED	NOCLIN	NOMED
Yum et al,71 2021	South Korea	103	NR	48	52	10.1	Longitudinal	HD, PC	R	NOCLIN	NOMED
Zhang et al,72 2021	Hong Kong	1793	NR	47	53	7.3	Longitudinal	T, PC, VG, HD, AD	R + ED	NOCLIN	NOMED

Abbreviations: AD, all devices; CLIN, diagnosed clinical condition; ED, educational; HD, handheld devices; MED, known medical condition; NOCLIN, no known clinical condition; NOMED, no known medical condition; NR, not reported; PC, personal computer; R, recreational; R + ED, R and ED combined; SES, socioeconomic status; SM, social media; T, television; VG, videogames.

^a Sample sizes for data extracted from studies (and not always the number reported for the study as a whole).

Meta-analysis

From a baseline value of 162 min/d (2.7 h/d), total daily screen time across all children increased during the COVID-19 pandemic by 84 min/d (90% CI, 51-116 min/d), corresponding to a moderate effect size when standardized (Figure 2). Between-study heterogeneity was small as summarized by a τ statistic of 0.3 SDs (90% CI, 0.2-0.5 SDs).

Figure 2. Contributing Studies for Change in Screen Time Before and During the COVID-19 Pandemic.

The studies are presented in order of smallest to largest change in screen time. The square data markers indicate the degree of change, with the lines through the markers indicating 90% CIs. The diamond data marker indicates the overall pooled effect based on the included studies.

Moderator analyses (Table 2)⁷³ revealed that increases in screen time were particularly marked for individuals 12 to 18 years of age, whose total daily screen time increased by 110 min/d (k = 26; 90% CI, 72-149 min/d), corresponding to a moderate to large effect size. The increase in total daily screen time for preschoolers and primary school children was smaller—approximately 65 min/d—corresponding to a moderate effect size (preschool k = 12 [mean, 66 min/d; 90% CI, 27-106 min/d]; primary school k = 49 [mean, 65 min/d; 90% CI, 36-95 min/d]). Time spent on both handheld devices and personal computers increased by approximately 45 min/d on both types of devices, corresponding to a moderate to large effect size (handheld device k = 20 [mean, 44 min/d; 90% CI, 11-77 min/d]; personal computer k = 13 [mean, 46 min/d; 90% CI, 12-81 min/d]). Moderator analyses also revealed that changes in total daily screen time were larger for sample estimates in which the data were reported retrospectively (116 min/d; 90% CI, 95-137 min/d; k = 36) rather than longitudinally (65 min/d; 90% CI, 50-80 min/d; k = 51). Both estimates were in the range of moderate effect sizes.

Table 2. Moderator Results of the Changes in Daily Screen Time Comparing Before vs During COVID-19.

Characteristic	k a	Mean (90% CI), min/db	β (90% CI)c
Categorical moderators
Device type or content
Handheld devices	20	44 (11 to 77)d	0.83 (0.21 to 1.45)d
Personal computer	13	46 (12 to 81)d	0.90 (0.24 to 1.59)d
Television	11	55 (10 to 118)	0.49 (0.09 to 1.05)
Videogaming	9	39 (4 to 73)	0.64 (0.07 to 1.20)
Social media	6	36 (−2 to 75)	0.35 (−0.02 to 0.72)
Age group
Preschool	12	66 (27 to 106)	0.64 (0.26 to 1.02)d
Primary school	49	65 (36 to 95)	0.63 (0.34 to 0.91)d
Secondary school	26	110 (72 to 149)	1.06 (0.69 to 1.43)d
Sex
Female	15	68 (33 to 102)	0.66 (0.32 to 0.99)d
Male	17	74 (39 to 109)	0.72 (0.38 to 1.05)d
Differencee	96f	−6 (−55 to 42)	−0.06 (−0.53 to 0.41)
Context
Educational and recreational combined	33	68 (36 to 100)	0.66 (0.35 to 0.97)
Recreational	54	84 (50 to 119)	0.82 (0.48 to 1.15)
Differencee	51f	−16 (−37 to 4)	−0.16 (−0.36 to 0.04)
Design
Longitudinal	51	65 (50 to 80)	0.63 (0.49 to 0.78)d
Retrospective	36	116 (95 to 137)	1.12 (0.91 to 1.32)d
Differencee	38f	51 (26 to 75)	0.49 (0.25 to 0.72)d
Clinical condition
No	83	76 (44 to 108)	0.73 (0.42 to 1.04)
Yes	4	84 (28 to 139)	0.81 (0.27 to 1.35)
Differencee	40f	8 (−39 to 55)	0.08 (−0.38 to 0.53)
Other medical condition
No	71	76 (44 to 107)	0.73 (0.42 to 1.04)
Yes	16	112 (67 to 157)	1.08 (0.64 to 1.51)
Differencee	60f	36 (−1 to 74)	0.35 (−0.01 to 0.71)
Continuous moderatorsg
Baseline	46	26 (6 to 46)	0.25 (0.06 to 0.44)
Quality	46	9 (−15 to 32)	0.08 (−0.15 to 0.31)
Region	46	−5 (−42 to 32)	−0.05 (−0.40 to 0.30)
Duration	46	−4 (−32 to 24)	−0.04 (−0.31 to 0.23)

^a The number of independent samples used for deriving the estimated mean value.

^b Mean changes in total daily screen time and individual categories of screen time reported in minutes per day. Mean values were calculated with the remaining moderators held constant at their mean values.

^c Standardized effect sizes were calculated by dividing mean changes by the corresponding SD for the category of screen time.

^d Effect sizes with adequate precision at the 90% CI (ie, when the chance of an outcome including both substantial negative and positive values [ie, values >0.2 and <−0.2 SDs, respectively] was less than 5%).

^e Differences in estimated mean changes in total daily screen time were calculated with sex = female, design = longitudinal, and clinical status = no as the reference values.

^f The df is reported instead of the k value.

^g Continuous moderators were analyzed by estimating the difference in mean changes between studies with lower (mean, −1 SD) and higher (mean, 1 SD) values for total daily screen time.⁷³

Moderator analyses (Table 2) signaled possible increases in television viewing, video gaming, and social media use. Changes in daily screen time were also possibly larger for sample estimates with higher baseline (pre–COVID-19) screen time levels, sample estimates of recreational screen time, sample estimates representing children and adolescents with weight-related medical diagnoses, and sample estimates based on parental reports. However, sampling uncertainty in each of these outcomes was too large to be definitive (ie, 90% CIs included a wide range of trivial values). Sampling uncertainty for the remaining moderators shown in Table 2 (ie, sex, regional and seasonal characteristics, studies of samples with clinical diagnoses, and studies conducted over different durations) should be interpreted as unclear.

Publication Bias and Outliers

The standardized slope of the regression line representing publication bias was a trivial effect size (β = 0.09; 90% CI, −0.06 to 0.25) (eFigure in the Supplement). A single outlier was identified against the weighted threshold of P < .001. The direction or effect sizes of study outcomes were not sensitive to the removal of this outlier.

Discussion

This meta-analysis of 46 studies (146 effect sizes) from 29 017 children and adolescents revealed that, on average, screen time increased by 52%, or 84 min/d (1.4 h/d), during the pandemic. Compared with a prepandemic baseline value of 162 min/d (2.7 h/d), this increase corresponds to a daily mean of 246 minutes of screen time per day (4.1 h/d) across all children and adolescents during the pandemic. This substantial change in screen time is more than what can be expected according to developmental changes^19,20 and time trends.²¹ Substantial mean increases were observed in samples examining changes in recreational screen time alone (increase of 84 min/d) as well as combined estimates of recreational plus educational (increase of 68 min/d) screen time from prior to during the pandemic. As such, changes in screen time estimated in this study can very likely be associated with the unprecedented disruptions of the COVID-19 pandemic. These findings should be considered along with another meta-analysis suggesting a 32% decrease in children’s engagement in moderate to vigorous physical activity during the pandemic.⁷⁴ Policy-relevant pandemic recovery planning and resource allocation should therefore consider how to help children, adolescents, and families to “sit less and play more” to meet the 24-hour movement guidelines.⁷⁵

In this meta-analysis, we identified several moderators that explained existing heterogeneity across studies examining changes in screen time before vs during the pandemic. Changes were larger for individuals 12 to 18 years of age (110 min/d) compared with preschoolers (66 min/d) and middle school children (65 min/d). Adolescents were more likely than their younger counterparts to own and access digital devices.⁷⁶ This finding could also be explained by the fact that adolescence is marked by an increased emphasis on both a wider interpersonal and virtual peer network as well as the development of romantic relationships.⁷⁷ In most circumstances, the social distancing restrictions implemented during the pandemic prohibited face-to-face social interactions between children and adolescents from different households, especially early in the pandemic. Therefore, it is likely that they resorted to and relied on digital devices to stay connected. This finding aligns with a recent census of screen use among children and adolescents, in which 83% of respondents reported using screens to stay connected with family and friends.⁷⁸ Adolescents were also more likely than younger children during the pandemic to seek new outlets for creative expression, learning new skills and building on existing skills in a remote context, much of which took place on digital devices.⁷⁸

The estimated mean changes in screen time spent on handheld devices (44 min/d) and personal computers (46 min/d) were particularly marked, whereas changes in television, gaming, and social media were similar. This finding aligns with the observation that, as devices became a central component of daily living and interactions during the pandemic—for work, schooling, learning, socialization, and recreation alike—1 in 5 parents reportedly purchased new devices for their children, primarily computers and handheld devices.⁷⁹ Handheld devices and personal computers also provide access to text messaging, instant messaging, video chatting and sharing, etc, which children and adolescents are more likely to engage in to connect with peers.

Although the observed mean values were both moderate effect sizes, there was a larger range of increases in screen time estimated when prepandemic screen time data were collected in studies retrospectively (90% CI, 95-116 min/d) rather than longitudinally (90% CI, 50-80 min/d). Given the unprecedented nature of the pandemic as well as the time-sensitive need to study pandemic-related associations in real time, some scholars collected pandemic data in a largely pragmatic manner, including the use of retrospective recall of prepandemic experiences and behaviors. However, retrospective study designs are vulnerable to recall bias.²⁷ For example, parents may have become more acutely aware of their children’s screen time during lockdowns, which may have biased their perception of and ability to accurately recall their children’s prepandemic screen time. Comparatively speaking, longitudinal designs are often more methodologically rigorous. As such, within-person studies of child and adolescent screen time should be more heavily relied on to inform decision-making regarding policy and practice given their scope for enhanced precision of estimation.

Although we examined duration, content, and context of use in this meta-analysis, we could not examine how children and adolescents were using screens (eg, solitary viewing, gaming with others, or video chatting). It is possible, for example, that some youths used screens as a supportive tool for connecting with peers and other supports during physical distancing, which could explain their increased use. Children and adolescents who used screens to coview or connect with others during the pandemic had half as much screen time as their peers who viewed screens in a solitary manner.⁸⁰ Thus, future research should examine duration of screen time and its association with whatever devices or platforms children and adolescents are using, examine how they are engaging with screens, and determine when and for whom problematic screen use may develop.⁸¹

Studies have found small associations between increased screen use among children and poor mental health both before (see Eirich et al¹¹ for a meta-analysis) and during the pandemic^82,83,84,85; however, the association may be nonlinear. That is, there is support for an inverted U-shaped association between screen time and well-being—the “Goldilocks hypothesis”—in which children who receive less than 1 hour of screen time per day and those who receive high doses of screen time have been shown to have the poorest psychosocial functioning compared with children with moderate screen use.⁸⁶ Thus, restricting screens altogether is likely not a feasible or optimal solution to managing children’s and adolescents’ screen use during the pandemic or afterward. Understanding how screens have been used during the COVID-19 pandemic, for better and for worse,⁸⁷ and determining who is at greatest risk for sustained problematic outcomes require priority in future studies. Cohort study designs with repeated measures that can account for changes in screen use and mental health before, during, and after the COVID-19 pandemic will be particularly important for this endeavor.

Implications

The observed increase in screen time during the COVID-19 pandemic may be temporary and context dependent for some youths (eg, those isolated during school closures). However, for others, sustained problematic screen use habits may be formed. Practitioners working with children, adolescents, and families should focus on promoting healthy device habits among youths, which can include moderating and monitoring daily use, choosing age-appropriate programs, and prioritizing device-free time with family and friends. Youths should be prompted to think about how they use screens and whether they can focus their time on screens to meaningfully connect with others or as a creative outlet. It is also critical to discuss balancing screen use with other important daily functions, such as sleep and physical activity. Last, given that screen use is often interconnected among family members, that parents’ level of screen use is strongly associated with children’s screen use,⁸⁸ and that parents’ stress during the pandemic was associated with children’s increased duration of screen use,⁴ it is important for practitioners to speak jointly with youths and their caregivers to effect change in familywide screen use.⁸⁹

Limitations

This study had several limitations. First, although there was representative coverage of various continents in this meta-analysis, there were no samples from South Africa and limited samples from South America and the Middle East. Thus, findings may be relevant only to specific geographic regions of the world. Second, no reports of screen time were validated against passive sensing apps.⁹⁰ Third, only 1 study explicitly reported that all participants were engaging in virtual learning, and included samples were homogeneous in terms of socioeconomic status, precluding consideration of these variables as potential moderators. Greater diversity in sampling for future research studies on child and adolescent screen use is urgently needed.

Conclusions

The COVID-19 pandemic led to substantial changes in daily routines of children and adolescents. This systematic review and meta-analysis revealed that their screen time during the pandemic increased by 52% compared with prepandemic baseline estimates, which is greater than what would be expected based on age changes and time trends. Recovery initiatives should focus on promoting healthy device habits among children and adolescents, including moderating daily use, monitoring content, and promoting the use of screens as a creative outlet and to meaningfully connect with others. Cohort study designs with repeated measurement of screen time that can account for developmental change, as well as preexisting risks and stable contextual factors or vulnerabilities, are needed to disentangle the associations of the COVID-19 pandemic with the screen time and mental health outcomes of children and adolescents.

Supplement.

eTable 1. Search Strategy From Ovid MEDLINE

eTable 2. Quality Assessment Criteria

eTable 3. Quality Assessment of Included Studies

eFigure. Assessment of Publication Bias Using a Scatterplot of the Random-Effect Solution and the SE for Each Sample Estimate

eReferences.