Early childhood screen use and symptoms of problematic media use

Abstract

Objective

To assess associations between early childhood screen time trajectories and problematic media use scores by age 5.5.

Methods

The present study is based on a prospective, community-based convenience sample of 315 parents of preschoolers, from Canada studied at the ages of 3.5 (2020), 4.5 (2021), and 5.5 (2022) during the Covid-19 pandemic. Parent-reported screen use at the ages of 3.5, 4.5, and 5.5 was used to estimate preschooler screen use trajectories. Using latent growth modeling, we identified low (mean = 0.9 h/day, 23%), average (mean = 3.0 h/day, 56%), and high (mean = 6.38 h/day, 21%) screen time trajectories. Parents reported child problematic media using the Problematic Media Use Measure – Short Form (PMUM-SF).

Results

A multiple regression, adjusted for child sex, effortful control and parent education and stress revealed that compared to children in the low screen time trajectory, children in the high screen time trajectory had higher problematic media use scores at age 5.5 (β = 0.378, p < 0.001). In addition, children in the average screen time trajectory scored higher than children in the low screen time trajectory (β = 0.229, p ≤ 0.001).

Conclusion

Our findings suggest that higher screen use in early childhood is associated with an increased risk for the development of dysregulated media use, which can interfere with family functioning. As such, parents should be encouraged to follow screen time recommendations of ≤1 h/day for children between the ages of 2 and 5.

Preschooler screen time trajectories and their association to problematic media use

Screen-based media consumption continues to increase among young children (Rideout & Robb, 2020). Increasing access to screen-based contents and devices by very young children has led to concerns over their potential for fostering behavioral addiction to screen media (Felt & Robb, 2016; World Health Organization, 2015). Childhood problematic media use represents a constellation of behaviors indicative of maladaptive use that interferes with functioning, including withdrawal from others because of screen use, and the loss of interest in non-screen-based activities (Domoff, Borgen, & Radesky, 2020). These behaviors have been shown to forecast worse psychosocial functioning in children aged 4 to 11 (Domoff et al., 2019). Adding to the potential burden of childhood problematic media use, research also indicates that specific types of problematic use of media, such as gaming disorder or internet addiction that persist into adolescence, are associated with interpersonal, mental health, and academic difficulties (Domoff, Foley, & Ferkel, 2020; Gentile, 2009; Milani, Osualdella, & Di Blasio, 2009). As such, understanding risk factors associated with the development of problematic media use symptoms is critical for prevention.

There remains limited research on early risk factors for the development of problematic media use symptoms in young children. Problematic media use increased during the pandemic in children between the ages of 2 and 12 (Eales, Gillespie, Alstat, Ferguson, & Carlson, 2021). During the same time, child screen time also markedly increased (Madigan, Eirich, Pador, McArthur, & Neville, 2022). Behavioral addictions are likely to emerge through repeated exposure (McCrory & Mayes, 2015). Consequently, child screen time, or the duration of time they accumulate in front of screens, may be associated with the emergence of problematic media use symptoms. Indeed, high levels of screen media exposure may lead children to develop a dependency on the type of stimulation provided by screen media and a corresponding loss of interest in activities that require cognitive effort or less immediate reward (Christodoulou, Majmundar, Chou, & Pentz, 2020).

Many behavioral addictions take root in childhood and adolescence (Derevensky, 2019). Therefore, these behaviors in early childhood provides a window of opportunity for preventive intervention. The objective of the present study is to estimate how trajectories of preschooler screen time across the ages of 3.5 and 5.5 during the COVID-19 pandemic prospectively contributes to problematic media use by age 5.5. Previous research has shown that individual child characteristics, including child sex and having lower levels of effortful control, may contribute to their engagement and reactivity to screen media (Domoff, Borgen, & Radesky, 2020). In addition, social determinants of health inequities, such as lower socioeconomic status and parenting stress also contribute to higher levels of child screen use (Hartshorne et al., 2021; Kroshus et al., 2022). For this reason, in the present study, we consider these variables as covariates. We hypothesize that higher levels of screen time during the preschool years will be positively associated with problematic media use scores by age 5.5.

Methods

Sample

Participants in our study (N = 315) were followed repeatedly across the ages of 3.5 (M_age = 3.45 [SD = 0.86], range: 2.0–5.42), 4.5 (M_age = 4.3 [0.86], range: 2.75–6.33), and 5.5 (M_age = 5.79 [0.84], range: 4.24–7.73) in the context of a larger, longitudinal 3-wave study on child and family screen media use (Nova Scotia Media Use Study, NSMUS, 2020–2023). Families with preschool-aged children were recruited using posters and flyers distributed in preschool or pre-kindergarten classes and advertisements on social media and in radio stations. Most respondents reported being born in Canada (91.0%), married (82.0%), and white (90.5%). Mothers were the respondent (93.4%) in most cases. There were slightly more boys in our sample (54.0%) than girls and the majority of parents (74.3%) reported having a university degree.

Study design

At the first and second assessments (in 2020 and 2021) data collection occurred between April and August. During the third assessment data collection occurred between May 2022 and March 2023 (with 77% taking place between September and October 2022). At each data collection wave, parents completed the online Media Assessment Questionnaire (MAQ), a comprehensive measure of child and family media habits including child screen time (Barr et al., 2020). The MAQ contains demographic questions, including child sex and parent education. For our study, parents also completed measures of child effortful control using the Child Behavior Questionnaire and parenting stress using the Parenting Stress Index. Finally, parents completed the Problematic Media Use Measure – Short Form at Wave 3 only.

Although our analysis plan was not pre-registered, all research objectives were previously formulated for a Canadian Institutes of Health Research grant awarded in 2022 (application # 474993). Parents provided written informed consent and were compensated for the first two years in the amount of $50 (CAD) for time devoted to completing the online questionnaires. During the third year of our study, parents were compensated in the amount of $300 (CAD) for time devoted to the entire research protocol which included a home visit, wearing of an accelerometer for a week, and completion of questionnaires. The present cohort study was prepared using the criteria from the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE; Vandenbroucke et al., 2007).

Predictor measure: screen time

Parents reported child screen time using the MAQ. The MAQ was developed by an international consortium of researchers with the aim of providing an exhaustive portrait of multiple aspects of child and parent screen use habits. Details on the development of this assessment have been published elsewhere (Barr et al., 2020). For the purpose of this study, we used the MAQ's child screen time measure. Parents reported the average amount of time their child spent doing each of the following on weekdays and weekend days separately when children were 3.5, 4.5, and 5.5 years old: (1) watching TV or DVDs; (2) using a computer; (3) playing video games on a console; (4); using an iPad, tablet, LeapPad, iTouch, or similar mobile device (excluding smartphones); or (5) using a smartphone. Response options included: (1) Never; (2) Less than 30 min; (3) 30 min to 1 h; (4) 1–2 h; (5) 2–3 h; (6) 4–5 h; (7) more than 5 h. We then converted these categorical responses into variables reflecting the number of hours spent with each type of media device. Our approach involved using the midpoint for each response range, except for “Never” where a score of 0 was used, and “5 or more hours a day” where a more conservative score of 5 was used. Weighted daily averages of time spent with each type of media device were then created by multiplying weekday estimates by 5 and weekend day estimates by 2 and dividing the total by 7. At each survey year, we calculated an overall daily screen time estimate by summing the weighted daily average across media devices. This measure has been widely used with independent samples to study preschool-aged children's screen time and its consequences and has been shown to positively correlate with child behavior problems (Gueron-Sela, Shalev, Gordon-Hacker, Egotubov, & Barr, 2023) and negatively correlate with child vocabulary (Sundqvist, Koch, & Barr, 2025) and global development scores (Binet et al., 2024), providing evidence for its predictive validity.

Outcome measure: problematic media use

Parents completed the Problematic Media Use Measure – Short Form (PMUM-SF; Domoff et al., 2019) when their child was 5.5. Items were adapted from the DSM-5 criteria for internet gaming disorder and formulated to capture child problematic use of any digital media via parent report (Domoff et al., 2019). This 9-item scale aims to assess early signs of dependence-like symptoms to screen media in children aged 4 to 11 (i.e., My child's screen media use causes problems for the family, The amount of time my child wants to use screen media keeps increasing). These items measure problematic media appropriately for both girls and boys (Domoff et al., 2019). Items were rated on a 5-point Likert scale ranging from 1 (never) to 5 (always). A mean problematic media use score was computed by summing all items and then dividing them by 9. Higher mean scores indicate greater problematic media use. Internal consistency among items was good in our sample (α = 0.90) and in the validation study (α = 0.93) (Domoff et al., 2019). Furthermore, a 1-factor solution has been shown to provide good fit to the PMUM-SF items and has provided evidence for sex-invariance (Domoff et al., 2019). The PMUM-SF has been found to be correlated with parent-child conflict and difficulties in functioning, yielding evidence of criterion validity (Domoff et al., 2019).

Covariates

All covariates were measured when children were 3.5. Child factors include child sex (1 = male; 2 = female) and effortful control. Effortful control was measured using the Child Behavior Questionnaire – Short Form (CBQ-SF; Putnam & Rothbart, 2006),²² which is designed to assess distinct dimensions of child temperament, including effortful control. Effortful control was measured using two six-item subscales: attentional focusing (i.e., When drawing or coloring a book, shows strong concentration) and inhibitory control (i.e., Can easily stop and activity when s/he is asked to). Higher effortful control scores are indicative of better attention control abilities. Internal consistency was good in this sample (α = 0.79). The short form of the CBQ shows satisfactory internal consistency, criterion validity, longitudinal stability, and inter-rater agreement (Putnam & Rothbart, 2006; Rothbart, Ahadi, Hershey, & Fisher, 2001).

Parental factors include their level of educational attainment and parenting stress. Education reflects the highest school grade completed by the parent. Responses were dichotomized as: (0) High school or college vocational or (1) Undergraduate or Graduate degree. Parents also completed the parenting distress subscale of the Parenting Stress Index (Abidin, Flens, & Austin, 2006). In total, parents completed 12 items (e.g., I find myself giving up more of my life to meet my child's needs than I ever expected). Items were rated on a 5-point Likert scale as: 1 (strongly disagree); 2 (disagree); 3 (not sure); 4 (agree) or 5 (strongly agree) and were then summed to create a total score. This measure showed good internal consistency, (α = 0.85) and has been found to forecast dysfunctional parenting, showing evidence of predictive validity (Abidin et al., 2006).

Data analytic strategy

To capture the evolution and stability of child screen use behavior across three time points during the preschool years, we used growth mixture modeling (GMM). Unlike traditional linear models that assume homogeneity across participants, GMM employs a person-centered approach that accounts for sample heterogeneity by estimating distinct developmental patterns across subgroups, offering a more nuanced understanding of risk profiles than variable-centered approaches. This allowed us to identify group-based trajectories of screen time and model screen use across the ages of 3.5, 4.5, and 5.5. GMM is useful for the identification of meaningful, unobserved groups based on similarities in participant growth trajectories. We begin by estimating an unconditional mean trajectory for the entire sample to verify that there is significant variance around the intercept, mean, and mean slope. We then estimate GMMs with separate intercepts, slopes, and variances for each group with a unique pattern of screen time. We estimate and compare fit indices for solutions with 1–4 classes. Our selection of a final solution is informed by indices proposed elsewhere (Jung & Wickrama, 2008). Our GMM analyses were conducted in Mplus 8.10 (Muthén & Muthén, 2019).

After selecting the best screen time trajectories solution, we model a multiple linear regression in SPSS (version 27) to estimate the contribution of preschooler screen time trajectories to problematic media use scores at the age of 5.5. To better estimate the unique contribution of child screen time trajectories to our outcome, we adjust for child sex and effortful control and both parent education and stress at age 3.5. Based on a ratio of N = 20/estimated parameter recommended by others (Kline, 2023), this study is sufficiently powered to detect the hypothesized associations in regression models with 7 parameters (N = 315 > 20 × 7 parameters).

Missing data

A flow diagram of the participants is presented in Fig. 1. At age 3.5 (T1) 315 children had complete data on screen time. By age 4.5 (T2) 265 children had complete screen time data indicating a 15.9 % attrition rate. Finally, by age 5.5 (T3) 209 had complete data indicating a 21.1% attrition rate from the previous year. Overall, 34.3% of the 315 participants had missing data on the observed outcome variables. An attrition analysis was conducted to compare participants with missing outcome data to those with complete data. Participants with missing data on problematic media use did not differ from participants without on baseline effortful control, parent stress, and parent education. Furthermore, there were no differences in the screen time trajectories of participants with or without missing data on problematic media use. There was a sex difference indicating that girls were less likely to have missing data than boys (OR = 0.478, 95% CI: 0.295–774). We conducted our analyses on complete cases only and using multiple imputations. Both approaches yielded a similar pattern of results, indicating minimal bias due to non-random attrition. Following best practices in the handling of missing data, we present data based on multiple imputation treatment of missing data as our main analyses (Cumming, 2013). More specifically 10 imputed data sets were estimated using SPSS (version 27). We then conducted analyses on the pooled estimates of these imputed data sets. Analysis based on complete cases are presented in Table 4.

Fig. 1.

Flow diagram showing the number of participants lost to follow-up across study waves

Table 4.

Linear regression results showing child problematic media use at 5.5 predicted by screen time trajectories using non-imputed data

	Problematic media use (5.5)
	B	β	p value
Screen time trajectories
High vs Low	0.737	0.401	<0.001
Average vs Low	0.475	0.308	<0.001
Child characteristics
Sex
Girl	−0.049	−0.032	0.615
Effortful control	−0.248	−0.276	<0.001
Parent characteristics
Education
University degree	−0.262	−0.145	0.028
Parenting stress	0.022	0.148	0.031
R 2	0.249
Adjusted R2	0.226

Ethics

Our research protocol was approved by the IRB of Univesité Sainte-Anne (#0090.d) and Université de Sherbrooke (2022-3398). Parents provided written informed consent.

Results

Descriptive results and bivariate associations

Descriptive statistics and sample characteristics are presented in Table 1. On average, preschool-aged children spent 3.5 h of screen time at the age of 3.5, 3.18 h at age 4.5, and 2.81 h at age 5.5, indicating a slight decrease in screen time across the preschool years. Boys had higher mean problematic media use scores at the age of 5.5 than girls (mean = 2.47 vs 2.16, p = 0.015). Higher levels of parenting stress at the age of 3.5 was correlated with higher child problematic media use scores (r = 0.19, p = 0.007) and higher levels of child effortful control contributed to lower scores of child problematic media use (r = −0.31, p < 0.001). Parent educational attainment was not associated with child problematic media use scores.

Table 1.

Descriptive statistics for child screen time trajectories, covariates, and problematic media use

Variable	M (SD) or %	N
Predictor: Screen time (ages 3.5 through 5.5)
Low trajectory	23.20	73
Average trajectory	56.20	177
High trajectory	20.60	65
Outcome (age 5.5)
Problematic media use	2.20 (0.79)	216
Covariates (age 3.5)
Child sex
Boy	54.00	170
Girl	46.00	145
Effortful control	4.70 (0.85)	315
Parent education
University	74.30	234
Secondary or college	25.70	81
Parenting stress	18.19 (5.60)	315

Screen time trajectories

Using GMM we identified low (three-year mean = 0.9 h/day, 23%), average (three-year mean = 3.0 h/day, 56%), and high (three-year mean = 6.38 h/day, 21%) child screen time trajectories. Trajectories are shown in Fig. 2. Table 2 shows fit indices for 1, 2, 3, and 4 class solutions. We selected a three-group model because it has a significant Vuong-Lo-Mendell-Rubin test p-value, lower BIC value compared to a 2 or 4 class solution and also yielded group sizes of at least 5% of the sample. This decision was based on previously established criteria.²¹ Missing data on screen time across the ages of 3.5, 4.5 and 5.5 was handled using full-information maximum likelihood estimation in Mplus. Intercepts were significant for the high (b = 6.17, p < 0.0001), average (b = 1.04, p < 0.0001), and low (b = 3.18, p < 0.0001) screen time trajectories. In addition, the slopes were significant for the average (b = −0.34, p 0.007) and low (b = −0.09, p 0.047) screen time groups indicating a slight decrease in screen time across the ages of 3.5 and 5.5.

Fig. 2.

Trajectories of preschooler screen time between 3.5 and 5.5 estimated using GMM

Table 2.

Fit indices for trajectories with 1, 2, 3, and 4 group solutions

	1 Class	2 Class	3 Class	4 Class
AIC	3273	3096	3021	3004
BIC	3303	3148	3104	3113
AdjBIC	3278	3104	3034	3021
Sizes (Ns)	315	197; 118	65; 73; 177	10; 72; 176; 57
Entropy	N/A	0.692	0.722	0.782
VLMR-LRT (p value)	N/A	0.0138	0.0002	0.0246

Note. VLMR-LRT = Vuong-Lo-Mendell-Rubin Likelihood Ration Test.

Associations between screen time trajectories and problematic media use scores

We used multiple regression to estimate associations between screen time trajectories and child problematic media use scores. These are presented in Table 3. Screen time trajectories were treated as a 3-level categorical variable. All regression models were adjusted for child sex, effortful control, and parent education and stress. Results indicate that children in the high and average trajectories obtained significantly higher problematic media use scores at 5.5, when compared to children in the low screen time trajectory. More specifically, children following the high screen trajectory had problematic media use scores on average 0.733 point higher (β = 0.378, p < 0.001), than children in the low trajectory. Children in the average trajectory scored on average 0.478 points higher on problematic media use (β = 0.2297, p ≤ 0.001), than those in the low screen use trajectory. Analyses conducted with non-imputed data reached the same results (see Table 4).

Table 3.

Linear regression results showing child problematic media use at 5.5 predicted by screen time trajectories using imputed data

	Problematic media use (5.5)
	B	β	p value
Screen time trajectories
High vs Low	0.733	0.378	<0.001
Average vs Low	0.478	0.297	<0.001
Child characteristics
Sex
Girl	−0.045	−0.028	0.695
Effortful control	−0.237	−0.251	<0.001
Parent characteristics
Education
University degree	−0.280	−0.153	0.034
Parenting stress	0.025	0.177	0.011
R 2	0.260
Adjusted R2	0.246

Note. R² value and standardized betas were computed based on an average of 10 imputed data sets.

Discussion

This study, to our knowledge, the first to use growth mixture modeling to identify developmental trajectories of screen time in children under age 6 and examine their prospective association with problematic media use. In support of our hypothesis, we found that high levels of screen time during early childhood are associated with higher problematic media use scores at age 5.5 years. In the present study, most children in our sample (77%) had patterns of screen time that were associated with increased problematic media use scores. That is, children regularly exposed to averages of 3 and 6 h of screen time daily, had higher problematic media use scores by the age of 5.5. In contrast, children who had 1 h or less of screen time per day (i.e., the current recommended amount for children under age 5 years), had the lowest problematic media use scores. As such, our findings suggest a dose-response association between screen media intake and problematic screen media use scores.

Researchers are increasingly interested in dysregulated, addiction-like, media use in children and adolescents. Still, little is known about these behaviors and their determinants in very young children. As such, it remains important to examine longitudinal associations between patterns of screen use, over an extended period of early childhood development and symptoms of problematic media use. Indeed, most studies have studied child screen time exposure at a single moment in development which may not accurately capture stability and changes in exposure over time.

We observed a slight decrease in screen time for children in the low and average screen time groups. This could reflect the progressive removal of lockdown measures and return to pre-pandemic routines and activities. In addition, as observed by others (Madigan, Browne, Racine, Mori, & Tough, 2019), the slight decline in screen time around the age of 5.5 could be occasioned by school entry, which is likely to reduce the amount of time children had available for screen time, and other leisure activities.

In addition to higher levels of screen time in early childhood, other factors predicted problematic media use scores at age 5.5 years. Children who had higher levels of effortful control had lower problematic media use scores. Effortful control is a child dispositional characteristic that reflects the extent to which they can effectively exercise control over their own attentional processes and behavior (Putnam & Rothbart, 2006). As such, effortful control may help children better regulate how much and when they engage with screen media. Parents may also have more difficulty regulating the screen use habits of children who have more difficulty maintaining their concentration and interrupting an ongoing activity when asked to (Clifford, Doane, Breitenstein, Grimm, & Lemery-Chalfant, 2020). Previous work has found that poor effortful control is associated with more screen use (Shin, Choi, Resor, & Smith, 2021) and may also contribute to problematic media use (Domoff, Borgen, & Radesky, 2020). Child sex did not predict child problematic media use scores in our sample. This is consistent with previous findings showing no associations between child sex and child problematic media use or adherence to screen time guidelines (Madigan et al., 2022; Swift et al., 2023).

Lower parent education and higher parenting stress predicted higher problematic media use scores. These results are consistent with previous work showing that family distress can contribute to patterns of family screen media use (Emond et al., 2018; Hartshorne et al., 2021). This finding reinforces the importance of considering structural socioeconomic factors in the development of interventions aiming to prevent later maladaptive screen media use and its associated impairment. Intervention strategies should be sensitive to contextual challenges faced by families, including parental stress and socioeconomic factors.

In order to design effective preventive interventions and inform clinical practice, it remains important to identify the early risk factors associated with problematic media use in young children. Our findings suggest it could be beneficial for health care and education professionals to work with parents to help them limit early childhood screen time. In particular, working with parents to help them create limits on child screen time and favor activities that help build effortful control skills could also have a protective effect. This may be especially beneficial for parents experiencing higher stress and socioeconomic and social risk such as lower educational attainment.

The following limitations merit consideration. A first limitation is our use of a convenience sample observed in the context of the pandemic, a unique socio-historical period. Replications post pandemic with larger probabilistic samples are warranted to confirm the generalizability of the results. Second, screen media use and child problematic media use symptoms were not based on objective measures but rather parent reported. This could have led to shared measurement bias. Another potential limitation is that we only account for baseline parenting stress. It is possible that parenting stress fluctuated over the course of the pandemic and our study, potentially influencing child screen time habits and their development of problematic media use habits. We also did not assess aspects of screen time beyond hours, such as content. Prior research has found that educational media content is associated with less risk of child problematic media use behavior (Coyne et al., 2022). Finally, it was not possible to control baseline level of problematic media use in children. As such it is not possible to rule out reverse correlation. Strengths of our study include our ability to track screen use habits over an extend period of time during early childhood. Furthermore, our study design allowed us to capture prospective associations between early childhood screen time habits and problematic media use scores.

Future research could further examine how different types of screen use and contents may be associated with the development of early childhood problematic media use symptoms. Watching videos on YouTube and streaming services has become the most popular screen-based activity for young children.¹ Many streaming apps use persuasive design features such as auto-play which can present children with an endless supply of content that is tailored to their interests (Domoff et al., 2019). These activities are likely to be highly engaging for young children and may lead to repeated activation of reward circuitry. As such, heavy screen use, during a developmental period characterized by high levels of neuroplasticity to experiences and environments, may pave the way for dysregulated screen use behaviors.

Many parents may be required to handle expressions of problematic media use from their children, in the post-pandemic recovery. Our study suggests the importance of adhering to screen time limits of 1 h or less for children under age 5 years to potentially prevent the emergence of addiction-like screen use. Helping families reduce and maintain moderate screen time habits may help protect children from the longitudinal sequelae of problematic media use, such as mental health and psychosocial difficulties (Coyne et al., 2024).

Funding Statement

Funding sources: All phases of this study were supported by grants from the Canadian Institutes of Health Research (#474993-2022), the Social Sciences and Humanities Research Council (CRC-2021-00009), and Research Nova Scotia (#2061-2019). Funders had no role in the design and conduct of the study.