Cumulative Social Risk and Child Screen Use: The Role of Child Temperament

Abstract

Objectives

It is critical to understand what children, and in which context, are at risk for high levels of screen use. This study examines whether child temperament interacts with cumulative social risk to predict young children’s screen use and if the results are consistent with differential susceptibility or diathesis-stress models.

Methods

Data from 1,992 families in Calgary, Alberta (81% White; 47% female; 94% >$40,000 income) from the All Our Families cohort were included. Mothers reported on cumulative social risk (e.g., low income and education, maternal depression) at <25 weeks of gestation, child’s temperament at 36 months of age (surgency/extraversion, negative affectivity, effortful control), and child’s screen use (hours/day) at 60 months of age. Along with socio-demographic factors, baseline levels of screen use were included as covariates.

Results

Children high in surgency (i.e., high-intensity pleasure, impulsivity) had greater screen use than children low in surgency as social risk exposure increased. In line with differential susceptibility, children high in surgency also had less screen use than children low in surgency in contexts of low social risk. Children with heightened negative affectivity (i.e., frequent expressions of fear/frustration) had greater screen use as social risk increased, supporting a diathesis-stress model.

Conclusions

Young children predisposed to high-intensity pleasure seeking and negative affectivity in environments characterized as high in social risk may be prone to greater durations of screen use. Findings suggest that an understanding of social risks and individual characteristics of the child should be considered when promoting healthy digital health habits.

Introduction

Accessibility and use of screens are now pervasive in the lives of children, and many children are exceeding the recommended guideline of no more than 1 h of high-quality programming per day (American Academy of Pediatrics Council on Communications and Media, 2016; Madigan et al., 2020). Due to its habit-forming nature, there is growing concern that screen use may become maladaptive in early childhood (Christakis, 2019; McArthur et al., 2020). A variety of social risks (e.g., low income, caregiver mental illness) have been associated with children’s increased screen use (Hoyos Cillero & Jago, 2010; McArthur et al., 2020). However, when these social factors are examined across studies or within systematic reviews, results are mixed, and the direct relations often vary across periods of child development (Duch et al., 2013; Hoyos Cillero & Jago, 2010). One explanation for these mixed findings could be that one risk examined in isolation (e.g., low income) does not depict the environment as a whole (e.g., low income is often paired with single parenthood and poor housing).

The cumulative social risk perspective outlines that risks tend to co-aggregate and that individual risk factors convey little harm when examined in isolation but convey significant harm when present with other risks (Browne et al., 2015; Evans et al., 2013). Using cumulative social risk indices to examine child outcomes has been encouraged as a way to provide more parsimonious models of social risk, reduce multicollinearity between risk factors, and provide relevance for practice and policy implications as it more accurately depicts the environmental ecologies that children grow up in (Evans et al., 2013). That said, even in the context of multiple risk factors, not all children exhibit problematic outcomes. The bio-ecological model has emphasized that child temperament interacts with the broader social environment to influence children’s outcomes (Bronfenbrenner & Ceci, 1994). However, despite much scientific and public discourse regarding the impact of screen use on child health outcomes, very little is known about how social risks and child traits (e.g., temperament) interact to predict screen use behaviors (Browne et al., 2020; Christakis, 2019). The current study examines the interaction between cumulative social risk and temperament on children’s later screen use.

Child temperament, thought of as the behavioral expression of heritable characteristics, is considered to vary across three main dimensions: negative affectivity (i.e., expressions of fear, sadness, anger, or frustration), surgency (or extraversion; i.e., behavioral impulsivity, activity, and approach), and effortful control (i.e., self-regulation and attentional processes; Putnam & Rothbart, 2006; Rothbart et al., 2001). Two dominant theoretical models have been proposed to better understand the unique interplay between cumulative social risk and temperament when predicting child outcomes. The first model, the diathesis-stress model, suggests that some children possess vulnerability characteristics, which are underlying diatheses that are “triggered” by social-environmental stressors leading to negative outcomes (Ingram & Luxton, 2005). For example, children exposed to hostile parenting are more likely to have negative behavioral outcomes when they also carry vulnerability factors, such as high negative affectivity (Morris et al., 2002). The second model, the differential susceptibility model, posits that certain temperament characteristics are better seen as “plasticity factors,” where these characteristics can increase risk when social-environmental stressors are present but also increase positive outcomes when positive social-environmental experiences are present (Belsky & Pluess, 2009). In other words, certain temperament traits are proposed to engender greater sensitivity to both positive and negative environments in a “for better and for worse” manner (Belsky, 2016). Identifying the theoretical framework that best explains the relationship between cumulative social risk and child temperament is useful for the purpose of understanding for whom (i.e., under what conditions) children are most susceptible to high levels of screen use.

The first aim of this study was to examine if the effect of early social risk on screen use at 60 months of age varied according to child temperamental characteristics at 36 months of age. The second aim was to determine if significant interactions between cumulative social risk and temperamental characteristics are consistent with models of diathesis-stress or differential susceptibility. To further strengthen the design of this study, screen use at 36 months of age was controlled in analyses to mitigate the impact of individual predispositions and improve the longitudinal prediction.

Methods

Study Design and Population

Participants were from All Our Families, a pregnancy cohort of mothers and children from Calgary, Canada (McDonald et al., 2013; Tough et al., 2017). Women were recruited between August 2008 and December 2010 through primary health care offices, community advertising, and laboratories. Inclusion criteria were (a) ≥18 years of age, (b) fluent in English, (c) gestational age <25 weeks, and (d) receiving community-based prenatal care. The current study used data from the baseline (T1; <25 weeks of gestation), 36-month (T2; M_age = 36.35, standard deviation [SD] = 1.95) and 60-month (T3; M_age= 61.59, SD = 3.07) postnatal assessment periods. Participants who completed the 60-month assessment survey (N = 1,992) were included in the current study (Table I). Informed consent was provided by mothers at each timepoint, and all procedures were approved by the institutional ethics board.

Table I.

Sample Demographics and Study Characteristics

Characteristic	Value
Maternal race/ethnicity,a  N (%)
White	1,604 (81.5)
Black/African American	23 (1.2)
Indigenous	10 (0.5)
Asian	211 (10.7)
Latin American	31 (1.6)
Multiracial/other	79 (4.0)
Missing	10 (0.5)
Cumulative social risk,b  N (%)
0 risks	1,185 (60.2)
1 risk	494 (25.1)
2 risks	175 (8.9)
3 or more risks	114 (5.8)
Child sex, N (%)
Male	1,037 (52.7)
Female	931 (47.3)
Child age (months), mean (SD)
36	36.4 (2.0)
60	61.6 (3.1)
Child temperament,c mean (SD)
Surgency/extraversion	4.52 (0.79)
Negative affectivity	3.62 (0.79)
Effortful control	5.17 (0.69)
Screen use (h per week), mean (SD)
36 months	24.6 (12.43)
60 months	10.8 (5.29)
School attendance at 60 months, N (%)
Not in school	246 (12.5)
Half- or full-day school	1,722 (87.5)

Note. SD = standard deviation.

^aMaternal race and ethnicity was collected via maternal self-report based on ethnic/racial categories typically used by Statistics Canada at the time of data collection.

^bA cumulative social risk score was calculated by summing all risk experiences and scores above 3 were truncated due to low response rates (2.2%; n = 44), resulting in a cumulative risk score ranging from 0 to 3 (see Supplementary Table 1 for descriptive information).

^cChild Behavior Questionnaire-Very Short Form (Putnam & Rothbart, 2006), range = 1–7.

Measures

Cumulative Social Risk

Consistent with metrics of cumulative social risk (Browne et al., 2015; Wade et al., 2018), at baseline (<25 weeks of gestation; T1) mothers reported on their risk level (1 = risk; 0 = no risk) for the following items (Supplementary Table 1): (a) income (1 = <40,000 CAD$; 0 = ≥40,000 CAD$), (b) maternal education (1 = ≤high school education; 0 = some or completed post-secondary), (c) maternal depression (coded based on the at-risk cutoff for the Center for Epidemiological Studies Depression Scale; Lewinsohn et al. 1997; 1= ≥16; 0 = <16), (d) marital status (1 = single/divorced; 0 = married/common law), (e) housing instability (1 = ≥3 moves within past 2 years; 0 = <3 moves within past 2 years), (f) maternal unemployment (1 = experienced unemployment; 0 = have not experienced unemployment), and (g) maternal stress (based on the Perceived Stress Scale; Cohen et al., 1983; 1 = ≥+1 SD; 0 = <+1 SD); and (h) maternal history of childhood adversity prior to age 18 using an adapted version of the Adverse Childhood Experiences (ACEs; Anda et al., 2006; Felitti et al., 1998) checklist (1 = 4 or more ACEs; 0 = less than 4 ACEs). Risk factors that were not naturally categorical or did not have pre-established norms for critical values were dichotomized as close as possible to the top or bottom 15% of the distribution (i.e., approximately ±1 SD). This method is commonly used when calculating cumulative risk scores (Evans & De France, 2021; Sameroff et al., 2004). A cumulative social risk score was calculated by summing all risk experiences and scores above 3 were truncated due to low response rates (2.2%; n = 44), resulting in a cumulative risk score ranging from 0 to 3.

Child Temperament

At 36 months of age (T2), mothers reported on their child’s temperament using the Child Behavior Questionnaire-Very Short Form (CBQ-VSF; Putnam & Rothbart, 2006). The CBQ-VSF captures three temperament dimensions: surgency/extraversion (i.e., activity level, high-intensity pleasure seeking, impulsivity), negative affectivity (i.e., anger/frustration, sadness, discomfort), and effortful control (i.e., attention focusing, inhibitory control, perceptual sensitivity). Item responses ranged from 1 (extremely untrue) to 7 (extremely true), with mean scores calculated for each subscale. Higher scores indicate greater endorsement of temperament characteristics (range = 1–7). Internal consistency was adequate across subscales (α = .70 to α = .75).

Screen Use

At 60 months of age (T3) mothers reported the hours their child spent using electronic devices (i.e., watching television programs; movies; using a computer, gaming system, or other screen-based device) on a typical weekday and weekend day. A weighted average across week and weekend days was calculated to yield screen use in hours/week. Outliers greater than 4 SDs from the mean (n = 7) were winsorized (Berry et al., 2014).

Covariates

Analyses controlled for child sex (1 = female; 0 = male), school attendance (“Is your child in preschool or kindergarten [60 months; T3]?”; 0 = no, 1 = yes [half- or full-day]), and prior levels of screen use (hours/week at 36 months; T2; calculated exactly as described above for the 60-month timepoint; outliers greater than 4 SDs from the mean [n = 16] were winsorized; Berry et al., 2014).

Statistical Analysis

Three multiple linear regression analyses (one for each temperament dimension) were conducted to explore the interactions between cumulative social risk (T1) and child temperament (T2; surgency/extraversion, negative affectivity, and effortful control, respectively) to predict screen use at 60 months (T3). Each respective model included the covariates, the predictor (cumulative social risk), the moderator (temperament), and the interaction term (risk × temperament). Cumulative risk was treated as an ordinal variable and centered through dummy coding (0 risk = −1, 1 risk = 0, 2 risks = 1, 3+ risks = 2) and the moderator variables were mean-centered before computing the interaction terms and entering variables in the analysis. To probe significant interactions, an online application (https://www.yourpersonality.net/interaction/ros.pl; last accessed July 2021), and supplement to the Roisman et al. (2012) paper, was used to calculate the region of significance (ROS) results. The ROS method uses the Johnson–Neyman technique to determine the range of values of the independent variable for which the moderator and dependent variable are significantly associated (Roisman et al., 2012). The Proportion Affected (PA) index was also used to explore the interaction effects to distinguish between differential susceptibility and diathesis-stress models. Based on the recommendations outlined by Roisman et al. (2012), the PA index provides an estimate of the proportion of cases that are differentially affected by the crossover interaction, or the proportion of children that fall within the “for better” region. The PA index is not influenced by sample size or variations in the range of the data used to examine the interaction (i.e., 1 or 2 SD) and is considered a robust predictor of the form of the interaction. An a priori criterion of a PA value greater than .16 was used to support a differential susceptibility model, whereas a PA value closer to 0 would support a diathesis-stress model (Roisman et al., 2012). Lastly, we checked for nonlinearity as recommended by Roisman et al. (2012). However, the nonlinear model did not converge. All regression analyses were conducted in Mplus, version 8.1 (Muthén & Muthén, 2017). Effect sizes are based on the guidelines established by Funder and Ozer (2019) for interpreting effects in psychological research. Specifically, an effect-size (r; correlations) of .05 indicates a very small effect, .10 indicates a small effect, .20 indicates a medium effect, and .30 or greater indicates a large effect.

Missing Data

From the initial pregnancy cohort (T1; N = 3,388), 95% (N = 3,223) agreed to be contacted for follow-up research. Of those who agreed to follow-up and were eligible at the time of questionnaire completion, 69% completed the 36-month questionnaire (T2; N = 1,994) and 71% completed the 60-month questionnaire (T3; N = 1,992). Screen use data at 60 months was available for 98.8% of these participants (n = 1,968). Attrition rates observed in the current study are similar to other prospective birth cohorts (Browne et al., 2018; Sontag-Padilla et al., 2015). Predictors of dropout are reported elsewhere and include mothers who were younger and had a lower level of education (Madigan et al., 2019). To account for missing data, and skewness and kurtosis (Lai, 2018), models were run with the maximum likelihood robust estimator (Graham, 2009; Yuan & Bentler, 2000).

Results

Descriptive statistics are included in Table I and Supplementary Table 2. Results of the regression models can be found in Table II. The significant interactions are displayed in Figures 1 and 2.

Table II.

Linear Regressions Predicting Screen Use (hours/week) at 60 months With Covariates (N = 1,968)

Predictor	Standardized estimate ($\beta$), 95% (CI)	Standard error	p-Value	R 2
SR × CR model
SRa	0.01 (−0.05 to 0.06)	0.03	.846
CRb	0.06 (0.01–0.10)	0.02	.012
Interaction (SR × CR)	0.07 (0.009–0.12)	0.03	.023
Female child	−0.05 (−0.09 to −0.01)	0.02	.008
Child in school	−0.01 (−0.05 to 0.04)	0.02	.728
Screen use at 36 m	0.49 (0.45–0.55)	0.02	.000
				0.27 (0.23–0.31)
NA × CR model
NAa	0.11 (0.02–0.20)	0.05	.015
CR	0.05 (0.004–0.10)	0.02	.034
Interaction (NA × CR)	0.08 (0.0001–0.17)	0.04	.049
Female child	−0.05 (−0.09 to −0.01)	0.02	.008
Child in school	−0.01 (−0.05 to 0.04)	0.02	.725
Screen use at 36 m	0.49 (0.44–0.54)	0.02	.000
				0.27 (0.23–0.31)
EC × CR model
ECa	−0.08 (−0.14 to −0.03)	0.03	.002
CR	0.06 (0.02–0.11)	0.02	.006
Interaction (EC × CR)	−0.02 (−0.08 to 0.04)	0.03	.452
Female child	−0.03 (−0.05 to 0.04)	0.02	.788
Child in school	−0.01 (−0.07 to 0.01)	0.02	.097
Screen use at 36 m	0.49 (0.44–0.54)	0.03	.000
R2				0.27 (0.23–0.31)

Note. The temperament and CR variables were centered before computing interaction terms and entering variables in the regression analysis. CI = confidence interval; CR = cumulative risk; EC = effortful control; NA = negative affectivity; SR = Surgency.

^aTemperament characteristics measured at 36 months of age via the Child Behavior Questionnaire-Very Short Form (Putnam & Rothbart, 2006).

^bCumulative social risk composite measured at T1 (<25 weeks of gestation; see Supplementary Table 1 for descriptive information).

Figure 1.

Simple slope plot of the interaction between surgency/extraversion (at ±1 SD) and cumulative social risk (range 0–3+ risks) predicting screen time at 60 months of age. The gray shaded area represents the region of significance. SD = standard deviation.

Figure 2.

Simple slope plot of the interaction between negative affectivity (at ±1 SD) and cumulative social risk (range 0–3+ risks) predicting screen time at 60 months of age. The gray shaded area represents the region of significance. SD = standard deviation.

Surgency as a Moderator of Cumulative Social Risk

Results from the regression model show a significant interaction between surgency and cumulative social risk in predicting screen use at 60 months of age, β = 0.07; 95% confidence interval (CI), 0.01–0.12; r = .17, suggesting a small effect (Funder & Ozer, 2019). The findings, displayed in Figure 1, indicated that cumulative social risk (T1) was associated with higher levels of screen use at 60 months of age (T3), for children high in surgency (+1 SD) at 36 months of age (T2), t(1961) = 28.74, p < .001, and lower screen use for children low in surgency (−1 SD) at 36 months of age (T2), t(1961) = 5.35, p ≤ .001.

Next, we performed an ROS analysis, which determines the values of the exposure (i.e., cumulative social risk) at which those high and low on the moderator (i.e., surgency ±1 SD) are significantly different on the outcome (i.e., screen use; Dearing & Hamilton, 2006; Roisman et al., 2012). As shown by the gray shaded regions in Figure 1, children high in surgency (+1 SD) had higher reported screen use than those low in surgency (−1 SD) when one or more cumulative social risk(s) were reported (t(1961) = 17.54, p < .001), and showed significantly less screen use than those with low surgency (−1 SD) when no social risks were reported (t(1961) = 21.20, p < .001). The association between surgency and screen use was significant for all values of cumulative social risk below 0.75 and above 0.99.

Next, the PA index was calculated to determine the likelihood of differential susceptibility. The PA index was .84 and indicates that 84% of the children fell below the crossover point in the “for better” region. As per Roisman et al.’s (2012) recommendation, a PA value greater than .16 provides support for differential susceptibility. Given the results from the ROS and PA index, the significant interaction between cumulative social risk and surgency in predicting screen use is consistent with a differential susceptibility model.

Negative Affectivity as a Moderator of Cumulative Social Risk

Results from the regression model show a significant interaction between negative affectivity and cumulative social risk in predicting screen use at 60 months of age, β = 0.08; 95% CI, 0.0001–0.17; r = .19, suggesting a small effect (Funder & Ozer, 2019). The findings, displayed in Figure 2, indicated that cumulative social risk (T1) was associated with higher levels of screen use at 60 months of age (T3), for children high in negative affectivity (+1 SD) at 36 months of age (T2), t(1961) = 26.17, p < .001, and lower levels of screen use for children low in negative affectivity (−1 SD) at 36 months of age (T2), t(1961) = 2.14, p = .032.

As shown by the gray shaded regions in Figure 2, the ROS analysis revealed that children high in negative affectivity (+1 SD) had higher reported screen use than those low in negative affectivity (−1 SD) at high levels (3+ Risks) of cumulative social risk (t(1961) = 50.81, p < .001), but did not have significantly different screen use than those with low negativity in contexts of low cumulative social risk (0 risks; t(1961) = 0.75, p = .454). The association between negative affectivity and screen use was significant for all values of cumulative social risk below −0.39 and above 0.85, suggesting that only an upper bound of significance was within the observed range of cumulative social risk. More specifically, the slope between negative affectivity and screen use was significant only when one or more cumulative risk(s) were reported. There were no significant differences below this value.

As seen in Figure 2, at no point in the simple slopes plot did children high in negative affectivity display lower screen use than those low in negative affectivity. The PA index was 0.0, suggesting that 0% of the children fall within the “for better” region. Given the results from the ROS and PA index, the significant interaction between cumulative social risk and negative affectivity in predicting screen use is most consistent with a diathesis-stress model.

Effortful Control as a Moderator of Cumulative Social Risk

Results from the regression model show a significant main effect of effortful control negatively predicting screen use at 60 months of age for children with one risk factor, β = −0.08; 95% CI, −0.14 to −0.03; r = −.11, suggesting a small effect (Funder & Ozer, 2019). However, there was no interaction effect between effortful control and cumulative social risk (β = −0.02; 95% CI, −0.08 to 0.04). Consequently, the additional analyses to evaluate differential susceptibility versus diathesis-stress models were not conducted.

Discussion

This study shows that children’s early social contexts relate to children’s later screen use behaviors differentially in children with certain temperamental characteristics. Specifically, children high in surgency (i.e., high-intensity pleasure seekers) had greater screen use than children low in surgency as cumulative social risk exposure increased. However, in line with the differential susceptibly model, children high in surgency also had less screen use than children low in surgency in contexts of low cumulative social risk. Additionally, compared to children with lower levels of negative affectivity, children with heightened negative affectivity had greater screen use as cumulative social risk increased, supporting a diathesis-stress model.

Surgency is a temperament characteristic related to greater engagement in the environment, sensation seeking, impulsivity, and reward sensitivity. While surgency has been associated with behavioral difficulties, like higher aggression and risk taking (Hentges et al., 2018), it has also been linked with some positive developmental outcomes, including greater sociability and verbal expressiveness (Rimm-Kaufman & Kagan, 2005; Slomkowski et al., 1992). Thus, one possibility is that the developmental advantages and disadvantages of surgency are altered based on exposure to varying levels of risk. In low adversity environments, surgency may be disproportionately associated with positive child outcomes (e.g., lower screen use). Conversely, in high adversity environments, surgency may be disproportionately associated with more negative child outcomes (e.g., higher screen use). In considering the specific mechanisms that might account for this finding, it is possible that children high in surgency may be more successfully redirected to other activities (e.g., outdoor play, organized sports) that occupy the child’s need for high levels of activity and sensation seeking when in low-risk contexts, thereby reducing overall screen use. However, in high-risk households children high in surgency may be more likely to engage in problematic screen use due to resource dilution in these families or ability to provide alternative activities (Repetti et al., 2002). For example, families with higher social risks likely have fewer financial resources to direct children to more adaptive activities and less emotional resources to manage high sensation seeking/impulsive children, so they may resort to screens as an instrumental tool or parenting method to manage difficult behaviors.

The current findings also support a diathesis-stress model in that children with heightened negative affectivity appear to be “triggered” by poor social-environmental conditions resulting in more screen use. Negative affectivity has been associated with a range of poor child outcomes, including externalizing and internalizing problems and poor social competence, particularly in the context of social or environmental stress (Rothbart & Bates, 2006). Children high in negative affectivity often have a hard time regulating their behavioral and emotional responses to stress (Morris et al., 2007), and one potential outlet of this inability to regulate may come in the form of heightened screen use. Indeed, studies with adult samples find that people experiencing heightened distress (e.g., depression) use television viewing and video gaming as ways to escape emotional distress (Anderson et al., 1996; Maroney et al., 2019). Children high in negative affectivity who are also experiencing social risks may have an even harder time regulating their propensity to experience negative emotions (Morris et al., 2007; Rothbart & Bates, 2006). In addition, parents experiencing these same social risks may be less able to model appropriate emotion regulation strategies and may be more likely to use screens as a form of dealing with their children’s negative emotional displays.

The results show that increased levels of effortful control (i.e., attention focusing, inhibitory control) longitudinally relate to decreased levels of screen use. However, there was no significant interaction between effortful control and cumulative risk in predicting screen use. Perhaps, children with high levels of effortful control have better internal self-regulation abilities and are less impacted by external social risks. There is a need for research that explores effortful control as a potential protective factor mitigating screen use, and a more nuanced understanding of the potential mechanisms that contribute to this relation (e.g., improved ability to disengage with media, better regulation of emotional and physiological responses to screen use).

A number of practice implications arise from this study. Encouraging and supporting parents to manage and monitor how, when, and where screens are used in the home may be particularly helpful for families with high social risks and children with high levels of surgency or negative affectivity. Our findings highlight that child temperament can impact who does “for better” or “for worse” in low- and high-risk households and that prevention and intervention efforts need to be specific and tailored to the individual needs and circumstances of each family and child. To account for what children, and in which context, are at risk for high levels of screen use, one potential future effort would be to screen children for temperamental characteristics and use this information in combination with cumulative risk factors to provide caregivers with individualized recommendations for screen use practices. Thus, pediatricians need to tailor recommendations and provide additional time to counsel families proactively whose children may be at increased risk for excessive screen use in early childhood.

Limitations

Our findings must be interpreted with the following limitations in mind. First, while parents are arguably the best informants of child characteristics and outcomes, single informant measurement introduces the potential for bias (Radesky et al., 2020; Yuan et al., 2019). Second, parental report of screen use can be inaccurate leading to misclassification (Radesky et al., 2020), and findings herein would be strengthened if replicated with passive sensing measures of screen use. Also, technology has progressed rapidly over the past decade (Rideout, 2017) and screen use exposure may have changed since the data for this study were collected (2014–2016). Third, our sample was predominantly White and middle-income, representative of the region of data collection; however, limiting generalizability of the findings. This sample is generally representative of those families living in an urban Canadian setting (i.e., predominantly White, educated and relatively high income), and consequently, study findings for populations more vulnerable to poor outcomes may differ. Further research is needed to better understand the relation between cumulative risk and temperament when predicting screen use for children from diverse and marginalized populations. Fourth, opinions on whether a moderator must occur concurrently to the variable that is being moderated vary with some statisticians arguing that temporal precedence is relevant, whereas others do not establish temporal criteria for moderators and predictors (Kraemer et al., 2008). Despite the inconclusive evidence for the temporal sequencing of predictors, the current study would be improved by repeated measurement of cumulative social risk and child temperament across development. Lastly, we examined duration, but not the quality (e.g., media type, educational content) or context (e.g., in isolation or co-viewing) of screen use. Collecting more nuanced information about the quality (e.g., video gaming vs. movies) and context (e.g., with others or solitary) of screen use could illuminate additional information about the nature of screen use for at-risk children and families.

Conclusions

Given the increased consumption of digital media and the unchartered territory of the digital family ecology (Barr, 2019), there is a critical need to understand what children, and in which context, are at risk for high levels of screen use. This study provides evidence that for young children, social risk interacts with individual child characteristics to predict later screen use behaviors. Our findings suggest that when clinicians encourage appropriate engagement with technology, an understanding of the social risks, and individual needs and characteristics of children and their families, should be considered.

Supplementary Data

Supplementary data can be found at: https://academic.oup.com/jpepsy.