Stress and infant media exposure during COVID-19: A study among Latino families

Abstract

Objective:

COVID-19 disproportionately harmed Latino families; however, its effects on their stress and media routines remain understudied. We examined economic and parenting stress patterns during COVID-19 and estimated associations between these forms of stress and non-adherence to American Academy of Pediatrics (AAP) infant media exposure recommendations among Latino families. We also explored how non-adherence with AAP recommendations varied with COVID-19 cases.

Methods:

We analyzed baseline data from an ongoing clinical trial recruiting low-income Latino parent-infant dyads. Non-adherence with AAP media exposure recommendations (ScreenQ) and economic and parent stress were measured via parent report. Additional variables included epidemiological data on COVID-19 cases. Linear models examined associations between the pandemic and both stress variables as well as between stress and ScreenQ. Using LOESS curve fitting, the rise and fall of NJ COVID-19 cases were overlayed with ScreenQ scores over time to visualize and explore trends.

Results:

All parents identified as Latino (62.6% unemployed, 91.5% limited English proficiency). Mean infant age was 8.2 months. Parent stress increased over time during COVID-19 (r=0.13, p=0.0369). After covariate adjustment, economic and parent stress were associated with increased non-adherence with AAP recommendations (standardized beta=0.16, 95%CI: 0.03-0.29; standardized beta=0.18, 95% CI: 0.04-0.31, respectively). Non-adherence to media exposure recommendations appeared to track with rises in COVID-19 with a lag of 7 days.

Conclusion:

Parent and economic stress were associated with non-adherence to infant media exposure recommendations among Latino families. These findings highlight the need for practitioners to support families from under-resourced communities and to promote healthy media routines.

INTRODUCTION

The COVID-19 pandemic disproportionately affected Latino communities in the U.S. in comparison to White communities.¹ Prior to the pandemic, Latino communities had the lowest rates of health insurance coverage, a smaller proportion of workers that could work remotely, and a higher burden of multiple chronic conditions; COVID-19 exacerbated these disparities.^{1, 2} During the early months of the pandemic, in addition to the higher infection and age-adjusted mortality rates among Latino communities, in comparison to White Americans, Latino children were more likely to experience poverty.^{1, 3, 4} For example, Latino families experienced food insecurity more commonly than White families.⁴ These disproportionate economic and health effects have several known repercussions and previous work documents the increased stress among Latino communities during early stages of the pandemic.^{5, 6} Previous work highlights increased media use among older children during the COVID-19 pandemic in a predominantly White study sample.⁷ However, to date, little work has examined how stress during the pandemic affected Latino infants and their families, and family routines, in particular media use, which changed dramatically during the pandemic.

Infancy is a critical developmental stage when family routines can promote optimal development and serve as an important buffer for stress.⁸ The American Academy of Pediatrics (AAP) recommends that families avoid media exposure in this age group except for video chat, given the potential harm and limited evidence for benefit.⁹ The COVID-19 pandemic had dramatic effects on family routines around media use and exposure in general. Children, including preschoolers, were introduced to more screen time during the COVID-19 pandemic lockdown.^{10, 11} However, there is a paucity of work that focused on infants specifically, where screen exposure is at odds with AAP recommendations. Moreover, to our knowledge, no studies have explored how screen exposure tracks with COVID-19 case volumes, a marker of pandemic severity.^{12, 13} One study using data from 12 countries found that screen exposure among young children was exacerbated for countries with longer initial lockdowns.¹⁰ Examining how screen exposure varied with pandemic severity may offer additional insight given the shifts in precautions and behaviors that occurred in response to pandemic conditions. Therefore, exploring how infant media exposure may have varied in accordance with the rise and fall of COVID-19 case volumes could provide useful insight into how the pandemic affected family media routines.

Previous work suggests that families’ exposure to stress may play a role in infant media exposure during the pandemic, but additional work is needed since stress can exist in different forms. In one study, parents who reported higher levels of parenting stress also reported higher screen time for their children, but this study focused on older children from higher-income households with highly-educated caregivers.¹¹ Additionally this study did not examine the effect of economic stress, which is known to have increased during the pandemic.¹⁴ Distinguishing the effects of economic stress from parenting stress on family media routines could yield useful knowledge that could help more effectively tailor support for families. In summary, to date, little is known about screen exposure among infants, particularly in Latino families who were at elevated risk for both economic and parent stress during the pandemic, and how pandemic severity and different forms of stress affected infant screen exposure.

To address these knowledge gaps, we (1) examined economic stress and parent stress patterns over time during COVID-19, (2) examined the extent to which economic stress and parent stress were associated with screen exposure among infants, which coincides with non-adherence to AAP guidelines on screen media exposure in this age group, and (3) explored how media exposure varied during different stages of the pandemic using epidemiological data on COVID-19 cases as a marker of severity. To achieve these objectives, we leveraged a unique sample of Latino parent-infant dyads from low-income backgrounds, a group that is often underrepresented in research. We hypothesized that increased levels of both forms of stress would be associated with infant media exposure.

METHODS

Study design

We analyzed baseline data from an ongoing study (Clinicaltrials.gov identifier: NCT04609553) occurring at three community health centers (CHCs) in urban areas in central New Jersey that serve a low-income Latino population. The overall study is an effectiveness implementation hybrid type I three-arm randomized trial that tests distinct approaches to literacy promotion in primary care. The Rutgers Biomedical Health Sciences Institutional Review Board approved this study and all participants provided informed consent.

Sampling

Data were collected during the enrollment study visit between November of 2020 and February of 2022. We focused on parent-infant dyads enrolled before the end of the COVID-19 omicron BA 1.1 winter wave of the pandemic, as policies and safety precautions dramatically relaxed after this point.^{15, 16} Parents were eligible to participate if they were ≥18 years, identified as Latino, spoke English or Spanish as their primary language, had an infant between the ages of 6 and 12 months old, and received regular pediatric care at one of the CHCs. Parents were ineligible if their infant had been diagnosed with multiple congenital anomalies, genetic conditions or developmental disabilities, or if they planned to discontinue care at the CHC.

Measures

Dependent Variable

Screen exposure:

The dependent variable was screen exposure measured by the ScreenQ. The ScreenQ is a parent-reported measure of screen-based media exposure among young children that includes substantial information on the quality of media content and time with the screen. The items were selected based on a cohesive conceptual framework that reflects AAP screen media guidelines.^{17, 18, 19} The total ScreenQ score ranges from 0 to 26. A higher score corresponds with more media exposure and non-adherence with AAP guidelines. Responses to each question are ordinal and scored from 0 to 2. The ScreenQ has four sub-dimensions: Access, Frequency, Content, and Interactivity/Co-Viewing. Access measures the child’s access to screens via different devices and in different locations; scores range from 0 to 10. The Frequency subdomain measures the child’s abundance of screen exposure with scores ranging from 0 to 8. Content score ranges from 0 to 6 and measures the type of screen exposure in which the child engages. Interactivity/Co-Viewing measures whether the infant uses screens with an adult and engages in discussion during screen exposure; scores range from 0 to 4. A Spanish translation of the ScreenQ was unavailable so we translated the scale from English to Spanish using forward back translation led by two native Spanish speakers, one of whom is a linguist specializing in bilingual (Spanish/English) development. The translation was then reviewed by an additional native speaker on our team.

Independent Variables

We measured two different forms of stress experienced by parents during the COVID-19 pandemic: economic stress and parent stress.

Economic stress

The economic stress variable is a composite measure of responses to 5-point Likert items expressing parents’ perceived stress regarding the following areas: finding or keeping safe and affordable housing, finding or keeping reliable transportation, getting enough healthy food throughout the month, and paying bills or managing finances. Responses were adapted from a larger survey of parent-reported needs and experiences.²⁰ Scores ranged from 4 to 20.

Parent stress

The Parental Stress Scale (PSS) is an eighteen-item measure designed to quantify parental stress. For each item, participants expressed agreement on a 5-point Likert scale with statements related to parents’ satisfaction in their role as a parent, their bond with their child, stressors pertaining to their parental role, including their child’s behavior and responsibility as a parent, as well as their views on life as a result of having children.²¹ Scores ranged from 18 to 90 where higher scores corresponded with higher parent stress. The PSS has shown consistency against varying parental characteristics as well as high validity and internal reliability.²¹

Covariates

We determined covariates a priori for use in adjusted linear models. Covariates included the date when the baseline survey was completed, the parents’ highest level of completed education, employment status, self-reported English proficiency, the infant's age at baseline, the number of children in the household, and infant initiative. We measured infant initiative using the Devereux Early Childhood Assessment (DECA) Infant and Toddler Program, which is a reliable and valid tool standardized nationally in both English and Spanish.²² We summed parents’ responses to a series of 5-point Likert items about their child’s social and emotional resilience to create scores ranging from below 20 to above 70. Higher scores corresponded with an infant’s greater ability to independently satisfy his or her needs.²² We categorized the composite scores dichotomously based on scoring guidelines with typical scores above the 16^th percentile, and scores ≤ the 16^th percentile below the cutoff.

Other Variables

COVID-19 data - We obtained New Jersey COVID-19 cases data from John Hopkins Tracking²³ for each day of the study period between November of 2020 and February of 2022 given that parent-infant dyads were residents of New Jersey.

Statistical Analysis

We used descriptive statistics to summarize participants’ characteristics. We summarized the relationship between time and parental stress using Pearson’s correlation and both the relationship between time and economic stress and between economic and parent stress using Spearman’s correlation. Using locally weighted scatterplot smoothing (LOESS) curve fitting, a nonparametric method to fit a curved line to a scatterplot and to explore patterns and trends in nonlinear data,²⁴ we overlayed plots of the rise and fall of incident NJ COVID-19 cases and total ScreenQ scores over time as an exploratory analysis of common trends. We chose a one-week lag after exploring different overlays of the two curves and based on a commonly reported pattern of symptoms to allow time to capture changes in individual behaviors. We found the one-week lag to correspond best and given the exploratory nature of this analysis, we did not perform any formal analyses to compare lags. We explored different LOESS smoothing factors to achieve the best fit of lines to the scatterplots.

We examined the association between stress experienced by parents, both economic and parent, and infant media exposure using linear regression. In multivariable analyses, we adjusted for the following covariates, which we determined a priori as described above: the date when the baseline survey was completed, the parents’ highest level of completed education, employment status, self-reported English proficiency, the infant's age at baseline, the number of children in the household, and infant initiative. We used standardized parameter estimates to allow comparison across models along with 95% confidence intervals. To further assess the association between stress and media exposure, we used logistic regression models with a dichotomous outcome measure of whether the infant uses screens to calm down.

We performed all statistical analyses using a significance level of 0.05 and generated graphical outputs in SAS desktop version 9.4.

RESULTS

Study Sample

Our analytic sample included 247 parent-infant dyads. Most parents received at least a high school diploma (64.0%), were unemployed (62.6%), and reported limited English proficiency (91.5%). On average infants were 8.2 months old and households had an average number of 2.4 children. Demographic characteristics are presented in Table 1.

Table 1.

Demographic characteristics among analytic sample (N=247)

Characteristic	N	Mean (SD) or %
Parent Age	247	31.1 (6.3)
Country of Origin
U.S. States and U.S. Territories	20	8.1%
Mexico	76	30.8%
Dominican Republic	60	24.3%
Honduras	42	17.0%
Other Latin American country	49	19.8%
Parent’s Highest Level of Completed Education
≤ Eighth grade	45	18.2%
High School, no diploma	44	17.8%
High School diploma or above	158	64.0%
Parent’s Employment Status
Employed	92	37.4%
Unemployed	154	62.6%
Parent’s English-Speaking Proficiency
Not very well	226	91.5%
Very well	21	8.5%
Infant’s Age at baseline (in months)	247	8.2 (1.6)
Infant’s Temperament
Typical	174	70.4%
Below cutoff	73	29.6%
Number of Children at Home	247	2.4 (1.2)
Economic Stress Score*	236	10.2 (4.5)
Parent Stress Score**	247	32.9 (6.8)
ScreenQ Total Score***	247	6.7 (3.9)
ScreenQ Access Subdomain	247	1.8 (1.8)
ScreenQ Frequency Subdomain	247	3.0 (1.8)
ScreenQ Content Subdomain	247	0.6 (0.9)
ScreenQ Interactivity/Co-Viewing Subdomain	247	1.2 (0.9)

^*Continuous measure of economic stress ranging from 4 to 20 where higher scores corresponded to higher economic stress.

^**Continuous measure of stress related to parental role and responsibilities ranging from 18 to 90. Higher numbers corresponded to increased parent stress.²⁰

^***Parent-reported measure of infant’s media exposure with four sub-dimensions: Access, Frequency, Content, and Interactivity/Co-Viewing.^{16, 17, 18} A higher score entailed more media exposure and non-adherence with American Academy of Pediatrics guidelines.

We found a significant positive correlation between parent stress score and time (r=0.133, p=0.0369) from November 2020 to February 2022. While in the positive direction, the correlation between economic stress score and time did not achieve statistical significance (r=0.114, p=0.0803). Figure 1 illustrates these patterns over time. Parent stress and economic stress variables were strongly and positively correlated (r=0.346, p<0.001). A one-week lag in the increase and decrease in the total ScreenQ score appeared to correspond with the rise and fall of incident NJ COVID-19 cases over time (Figure 2). In January 2022, the omicron variant resulted in a substantial peak in NJ COVID-19 cases. A corresponding sharp rise in total ScreenQ score is visible in this same time frame.

Figure 1-. Stress and COVID-19.

Trend in stress scores over time during the COVID-19 pandemic (a) Economic stress score (r=0.114, p=0.0803) (b) Parent stress score (r=0.133, p=0.0369). * indicates a significant association at α=0.05.

Figure 2-. ScreenQ Total Score during COVID-19.

Smooth curves of ScreenQ total score and NJ COVID-19 cases. The date corresponding to reported ScreenQ total score was shifted back 7 calendar days to explore a one-week lag in ScreenQ total score in response to the number of NJ COVID-19 cases. Smoothing factors for COVID-19 cases and ScreenQ total score were 0.05 and 0.09, respectively, where a higher smoothing factor indicates greater smoothing of the data.

Regression Analyses

Economic stress

In unadjusted linear models, economic stress was significantly associated with total ScreenQ score, ScreenQ frequency, and ScreenQ interactivity/co-viewing. For every 1 standard deviation increase in economic stress score, there was a 0.14 standard deviation increase in total ScreenQ score (95% CI: 0.01-0.26, R²=0.02), a 0.16 standard deviation increase in the frequency subdomain (95% CI: 0.03-0.29, R²=0.03), and a 0.18 standard deviation increase in the interactivity/co-viewing subdomain (95% CI: 0.05-0.31, R²=0.03). These results persisted with minimal differences in adjusted models (Table 2). In an adjusted logistic regression model, economic stress was not significantly associated with parents’ use of screens to help infants calm down (AOR=1.10, 95% CI: 0.93-1.29).

Table 2.

Association Between Economic Stress and Infant Media Exposure

	Total ScreenQ Score				ScreenQ Access				ScreenQ Frequency				ScreenQ Content				ScreenQ Interactivity/Co- Viewing
	Unadjusted		Adjusteda		Unadjusted		Adjusteda		Unadjusted		Adjusteda		Unadjusted		Adjusteda		Unadjusted		Adjusteda
	Std. ß	95% CI	Std. ß	95% CI	Std. ß	95% CI	Std. ß	95% CI	Std. ß	95% CI	Std. ß	95% CI	Std. ß	95% CI	Std. ß	95% CI	Std. ß	95% CI	Std. ß	95% CI
Economic Stress Score	0.14*	0.01 to 0.26	0.16*	0.03 to 0.29	0.003	−0.13 to 0.13	0.03	−0.10 to 0.17	0.16*	0.03 to 0.29	0.18**	0.05 to 0.31	0.09	−0.04 to 0.21	0.09	−0.04 to 0.23	0.18*	0.05 to 0.31	0.17*	0.04 to 0.30
Parent Stress Score	0.15*	0.02 to 0.27	0.18**	0.04 to 0.31	0.07	−0.05 to 0.20	0.11	−0.02 to 0.24	0.15*	0.03 to 0.27	0.18**	0.05 to 0.31	0.10	−0.02 to 0.23	0.13	−0.002 to 0.26	0.09	−0.04 to 0.21	0.05	−0.08 to 0.18

Notes: * indicates a statistically significant association where p<0.05 ** indicates a statistically significant association where p<0.01

^aAdjusted for parent’s education, employment status, and English proficiency; infant initiative and age; number of children at home, and survey date

Parent stress

Unadjusted linear models revealed a significant association between parent stress score and total ScreenQ score (standardized ß=0.15, 95% CI: 0.02-0.27, R²=0.02) and the frequency subdomain (standardized ß=0.15, 95% CI: 0.03-0.27, R²=0.02). In adjusted models, parent stress remained significantly associated with both total ScreenQ score (standardized ß=0.18, 95% CI: 0.04-0.31, adjusted R²=0.01) and the frequency subdomain (standardized ß=0.18, 95% CI: 0.05-0.31, adjusted R²=0.03) (Table 2). Further, a one standard deviation increase in parent stress score was significantly associated with a 25% standard deviation increase in the parents’ odds of using screens to help infants calm down (AOR=1.25, 95% CI: 1.06-1.47).

Economic and Parent stress

When including both economic stress and parent stress scores in the model with total ScreenQ score, there was little change in the magnitude of the regression coefficient for parent stress (standardized ß=0.15, 95% CI: 0.01-0.29, adjusted R²=0.02). For economic stress, the magnitude of the coefficient decreased from 0.14 to 0.11 (~ 20% reduction) and was no longer statistically significant. In the ScreenQ subdomain models economic stress score was significantly associated with the interactivity/co-viewing subdomain only (standardized ß=0.17, 95% CI: 0.03-0.31, adjusted R²=0.04), while parent stress was associated with the frequency subdomain (standardized ß=0.15, 95% CI: 0.01-0.29, adjusted R²=0.04).

DISCUSSION

In a sample of Latino parent-infant dyads, we found that parenting stress increased for families during the COVID-19 pandemic. This linear pattern suggests a chronic cumulative effect that could result in wear and tear and negative long-term consequences, as chronic stress has been linked to adverse physical and behavioral health outcomes.^{25, 26} We also found that both parenting and economic stress were associated with non-adherence to AAP guidelines on media exposure among infants. Our findings underscore the imperative for clinicians to look for long-term consequences of the pandemic, including the consequences of such cumulative stress and media routines.

Previous studies have documented an increase in stress among Latino parents and in familial economic stressors such as unemployment during early stages of the COVID-19 pandemic.^{5, 6} Our findings build on this work by documenting an increase in parenting stress over time through the 2022 omicron BA 1.1 winter wave of the pandemic. As the pandemic moved through different phases, these rising stress levels, in an especially vulnerable community, are indicative of lingering effects of public health concern that pediatric clinicians must be prepared to address. Social determinants of health screening that include assessments of parental behavioral health and linkage to community resources represent one action that could identify and support families in need.²⁷

In this sample of Latino families, we found significant associations between stress experienced by parents, both economic and parent stress, and non-adherence with AAP media exposure guidelines for infants. This study builds on prior work that established an association between high parent stress and high screen use among older children,¹¹ and adds insight and practical knowledge that can be useful for clinicians. First, the current study focused on Latino infants from under-resourced communities, who face pervasive inequities and have been under-represented in research to date. Second, the availability of data on economic and parent stress provides more nuanced insight into the association between stress and infant media exposure, which may in turn require different strategies to support families effectively (e.g., parenting support to address parenting stress versus material supports to address economic stress). Third, the use of the ScreenQ to assess infant media exposure provides practical information, given its correspondence with non-adherence AAP media exposure guidelines and subdomains that address both the quantity and quality of screen exposure. Past work has linked material hardship with longer duration of digital media exposure among school-age children, ⁷ our findings build on this work by also documenting less interactivity among infants and parents during screen exposure. Past work in a similar population suggests that even relatively modest differences in the amount of media exposure during infancy are associated with reduced language development in the second year of life. ²⁸ Additionally, verbal interactions during media use may moderate adverse impacts on language development with adverse associations found in the absence of interactivity, but not in its presence.²⁹ Given the limited benefit of screen time in this age group and its potential harm,^{30, 31, 32} these findings highlight the need for practitioners to specifically ask about infant screen exposure and how it is being used. In this way, guidance on screen exposure can go beyond simply discouraging screen time to understanding how screens are being used and offering practical alternatives, especially for families at risk of elevated stress.

The current study has notable strengths. First, as noted above, the sample consisted of Latino parent-infant dyads, which adds much needed diversity to the extant literature. We also merged detailed data on stress and media exposure with epidemiological data, which added insight into these patterns and to our knowledge is novel. However, there are also certain limitations. We relied on self-reported data, which potentially introduces recall and social desirability bias. Our analyses of relations between stress and screen use are based on cross-sectional analyses of data collected concurrently, and therefore causality regarding relationships between economic or parent stress and infant media exposure cannot be inferred. Although we controlled for important factors, we cannot exclude that there could be other unmeasured confounders. Furthermore, the interrelationships between stress, COVID-19 and ScreenQ included variables we could not control for and therefore we could not do a mediational analysis. Although data on the reliability and validity of the ScreenQ subscales have not been published, the items were selected based on a cohesive conceptual framework from the AAP media use guidelines. The subdomains provide important insight into the quality of media use, which has been a gap in the existing literature that has largely focused on the amount of screen exposure only. Additionally, while the focus on Latino families is needed and important, our findings may not generalize to all populations. Furthermore, there is substantial diversity within the Latino population. Our study population included three predominant Latin American countries of origin (Mexico, Dominican Republic, Honduras) plus the US, but our sample size did not allow for subgroup analyses. Future work with larger sample sizes and subgroup analyses would offer additional insight.

Despite these limitations, the current study provides unique insight into the experiences of Latino families during the pandemic. Specifically, we found increasing parent stress during COVID-19 and significant associations between both economic stress and parent stress and infant media exposure. These findings underscore the rippling effects of the pandemic on families from under-resourced communities and opportunities for practitioners to support them.

Data Sharing:

Deidentified quantitative data and associated documentation may be made available to interested readers conducting non-profit research consistent with National Institutes of Health policy through the establishment of data sharing agreements. The timeline for release of data will be after the study is completed and main study findings from the final data set are published in peer-reviewed journals. De-identified data will be made available to researchers from accredited institutions who provide a methodologically sound proposal for use in achieving the goals of the approved proposal.