COVID-19 and Parent-Child Psychological Well-being

Individuals were eligible if they worked in an hourly service industry position in a retail, food service, or hotel business in a large US city, had a child aged 2 to 7, and had a mobile phone that could send and receive basic SMS text messages (see Fig 1). The sample was originally recruited for a study examining parents’ work schedule unpredictability and family well-being. Recruitment occurred between August and November 2019 by using a venue-based sampling approach, a commonly used technique for producing generalizable samples of hard-to-reach, unrostered populations.¹⁷  The key to successfully using this technique is generating a complete list of venues, which in this case were food service, retail, and hospitality businesses in the city, provided by the Columbia University Earth Institute. We then constructed a sampling frame of venue (business) day-time units (VDTs), randomly selected VDTs, and systematically identified and recruited eligible individuals present in those VDTs.¹⁸  To do so, at the time that study staff visited each business, they aimed to identify all workers who met sampling criteria who were at work at that time by approaching workers at each business, determining their eligibility, and asking those workers to direct them to any other employee with a young child who was currently at the venue. This strategy differs from snowball sampling in that study staff only followed-up with potentially eligible workers who were present at that time to preserve random sampling.

When first recruited, for the original study purpose of estimating effects of work schedules on family well-being, all participants were asked to complete 30 days of daily surveys and a one-time survey about demographic and household characteristics. All aspects of this study received human subjects approval from the Duke University Institutional Review Board.

Beginning on February 20, 2020, each participant from the original sample was contacted about participating in an additional 14-day wave of daily data collection. To continue participating, individuals had to be still residing with a young child but could have experienced a job change or unemployment. Our analysis sample was limited to parents who had begun participation in this wave of data collection by April 25, 2020, by using daily survey responses through April 27 (N = 645 individuals; 8222 person-days for analysis).

At this wave’s outset, respondents were asked where they worked for up to 3 jobs. Respondents were then prompted to report on each day’s work experiences and their own and their child’s well-being via SMS text message for 14 consecutive days. All survey materials used for this study were available in both English and Spanish.

The daily text surveys were programmed and automated by a third-party vendor. On the day of enrollment in this wave, participants received a text welcoming them to the start of wave 2. The next day, the 14-day data collection period began. During that period, the first survey question was sent out each evening at 7:00 pm. As soon as respondents sent back their answer to the first survey question, the second question was sent. This sequence was repeated until all questions and answers had been sent and received. A “thank you” text sent at the end of the sequence let individuals know they had completed all the day’s survey questions. If a respondent failed to reply to the first survey question, a reminder text was sent at 8:00 pm. Additionally, if an individual started the survey but did not complete all questions, a reminder text was sent after 2 hours of inactivity (with the question on which the individual left off resent as part of the reminder) and then again after 14 hours of inactivity. Additional details about the text message survey protocol are available in Ananat and Gassman-Pines.¹⁹ 

This method of data collection has several benefits. First, all text messages were date and time stamped, allowing the research team to know for certain that questions were answered on a given day. This is an important issue in this of type of data collection and analysis in general²⁰ ; in the context of the current study, it is a crucial innovation because the date- and time-stamped data provide certainty about which questions were answered before the COVID-19 crisis and which were answered postcrisis. Using daily surveys also has other methodologic strengths, including reducing recall bias and lessening the need for mental aggregation.^21–23 

Beginning on March 23, all participants who had already received their 2 weeks of daily survey prompts were asked to complete a one-time survey that asked questions about how the family was faring during the COVID-19 crisis. All survey questions and answers were sent and received via SMS text message. In subsequent days, respondents who finished receiving 2 weeks of daily survey prompts were sent the one-time survey the next day through April 26, 2020. Response rates were immediate and high, with more than three-fourths of respondents completing the survey despite limited ability of the survey team to complete follow-up in real time.

Daily parental negative mood was measured with a single item asking, “How much of the time today did you feel fretful, angry, irritable, anxious, or depressed?” Answers were on a 3-point scale: none of the time, some of the time, and all of the time. This question was modified from a question with a 4-week recall period from the Health Utilities Index^24,25  (During the past 4 weeks how often did you feel fretful, angry, irritable, anxious or depressed?). The single item has been validated as a daily measure of negative mood because it is positively correlated with daily stressors, including daily food insecurity²⁶  and daily work schedule disruptions.¹⁹ 

Daily perceived negative sleep quality was measured with a single item also used in other daily survey studies,²⁷  asking, “How well did you sleep last night?” Answers were on a 10-point scale from really badly to really well. We treat self-reported sleep quality as a measure of daily well-being because perceived sleep quality is associated with daily affect.²⁸  The sleep quality measure was reverse coded so that higher numbers indicated worse perceived sleep quality. This measure has been validated because it is correlated in expected directions with negative and positive daily mood, daily self-esteem,²⁷  and daily work schedule disruptions, a daily stressor.¹⁹ 

Daily child uncooperative behavior was measured with a single item asking, “How much was your child uncooperative today?” Answers were on a 4-point scale: not at all, just a little, some, and a lot. This question was modified from an item in the Inattention/Overactivity with Aggression Conners Rating Scale,²⁹  which asks parents to rate how much the adjective describes their child “at this time.”

Daily child worry was measured with a single item asking, “How much did your child appear to be sad or worried today?” Answer choices were on a 4-point scale: not at all, just a little, some, and a lot. This question was modified from an item in the Preschool Behavior Questionnaire,³⁰  which asks parents to rate how much the child exhibits each behavior.

For both child behaviors, previous research has demonstrated the reliability and validity of multi-item scale versions adapted for measuring daily externalizing and internalizing behavior problems.³¹  In the current study, single items were used to reduce respondent burden and attrition.

In the daily survey, we asked questions about jobs (up to 3 per respondent) on the basis of the number of jobs reported at wave 2 enrollment. For each job, respondents were asked whether they worked that day and if so when they started and stopped working and whether their hours worked were their originally scheduled hours. In addition, if respondents did not work at a given job on a given day, they were asked if they were originally scheduled to work at that job that day. We used the answers to those questions to create 2 dichotomous indicator variables: the first was equal to 1 if respondents reported working at any job that day and 0 otherwise. The second indicated a work disruption that day and was equal to 1 if the parent reported that they (1) did not work at a job that day but had originally been scheduled to work or (2) worked different hours at a job than originally scheduled; the indicator was set to 0 otherwise.

Parents were asked the following questions about their own job loss: “Since the crisis hit, have you lost a job/been laid off permanently?”; and “Since the crisis hit, have you been furloughed/laid off temporarily?” Those with another adult in the household were also asked, “Has another adult in your household lost a job/been laid off permanently?”; and “Has another adult in your household been furloughed/laid off temporarily?” An indicator variable representing household layoff was set equal to 1 if the respondent indicated a temporary or permanent layoff for him- or herself or another adult and was set to 0 otherwise.

Respondents were asked, “How has your household’s income changed since the crisis hit?” Answers were as follows: stayed the same, fallen by less than half, fallen by more than half, and increased. An income loss indicator variable was set equal to 1 if income had fallen by less than half or by more than half and was set to 0 otherwise.

Respondents were asked whether they had quit a job or reduced work hours to handle increased child care demands. Those who had responsibility for a disabled or elderly person who usually also got care from someone else were asked whether that care was still available. A dichotomous variable for increased caregiver burden was set equal to 1 if the respondent quit or reduced work hours because of increased child care responsibilities or had lost the elder and/or disabled adult care they usually relied on; it was set to 0 otherwise.

Respondents were asked, “Has anyone in your household felt sick since the crisis started?” An indicator variable was set equal to 1 if they answered yes and 0 otherwise.

A cumulative hardship index ranging from 0 to 4 was created by summing the 4 indicator variables described above, and represents the number of COVID-19 hardships experienced.

Because the first major restrictions in response to COVID-19 were announced on March 13, we define dates from March 14 on as the postrestriction period. An indicator variable equals 1 for all dates from March 14 on and equals 0 for all dates before.

To answer research question 1, we ran regression models predicting daily parent and child psychological well-being on the basis of whether the day in question was in the pre- or postrestriction period, controlling for whether the day in question was a weekend day, and adjusted SEs to account for the fact that repeated daily observations are potentially correlated within families (as opposed to each observation being independent, the default assumption in regression analysis). We used ordinary least squares (OLS) regression when examining perceived negative sleep quality. Because parent mood, child uncooperative behavior, and child worry were reported by using ranked categories, we used ordered probit regressions for these outcomes. Ordered probits are similar to OLS regressions but model true underlying well-being as normally distributed, with the lowest-ranked reported category corresponding to values in the lower tail of the normal distribution, the next ranked category corresponding to values in the next portion of the distribution, etc.

To answer question 2, we used logistic regression, the standard technique used when an outcome is yes or no, to predict how the frequency of work and work disruptions changed postcrisis, using the same model described above. We also used basic descriptive statistics for the hardships measured in the single point-in-time survey.

Finally, to answer question 3, for each respondent, we averaged all well-being measures across all reports provided postcrisis and ran a series of OLS regressions predicting each outcome from either COVID-19–related hardships individually or a set of indicator variables representing the number of hardships. All regressions were weighted by the number of days’ surveys completed by each person postcrisis.

Analysis sample characteristics appear in Table 1. Our sample is majority female, consistent with working in the service industry and with having custody of a young child.¹⁹  Approximately half are African American and approximately one-fifth are Hispanic, consistent with being central-city hourly workers.³²  The mean age is 31, consistent with being the parent of a young child,¹⁹  and the modal education is 12 years, consistent with hourly service employment.³³  Approximately half of focal children are girls; focal children are, on average, 4.9 years of age. Approximately two-thirds live with another adult in the household. The mean income is $2239 per month.

Our results reveal that parent psychological well-being decreased during the post–COVID-19 restrictions period (Table 2). Daily parent negative mood was significantly more frequent postrestrictions than prerestrictions (P < .05). Regression-adjusted means highlight the change: prerestrictions, parents reported negative mood some of the time on 30% of days and all day on 7% of days. Postrestrictions, the share reporting negative mood some of the time rose by 10% to 33% of days, and the share reporting negative mood all day rose by 29% to 9% of days. Parents’ negative sleep quality did not change postrestrictions. Although the frequency of child uncooperative behavior and worry each day did not change significantly from pre- to postrestrictions, the pattern was consistent with increasing frequency of both uncooperative behavior and worry.

Our results using the daily survey data reveal that parental work dropped substantially at the onset of the COVID-19 crisis (Table 2). Additionally, work disruptions increased significantly in the postrestrictions period. The regression-adjusted percentage of respondents working each day fell from 67.9% prerestrictions to only 43.8% postrestrictions (P < .001), a 35% decrease. The regression-adjusted percentage of workers experiencing a work disruption on a given day rose from 9.3% prerestrictions to 19.9% postrestrictions (P < .001).

The point-in-time survey results reveal how frequently the COVID-19–related hardships occurred among sample families (Table 3). The majority of families (60%) experienced job losses. Most of our respondents (69%) reported household income declines. Forty-five percent of families reported increased caregiving burden and 12% of families had experienced illness.

Finally, Table 3 shows the accumulation of COVID-19–related hardships among sample families. Only 14% of families had no hardships during the crisis, whereas the majority experienced ≥2. A small share of families, 3%, had experienced all 4 hardships.

In general, those who had experienced hardships had worse psychological well-being (Table 4). These findings were consistent across both parents and children. Each individual hardship was related to significantly worse parental mood, and both caregiving burden and household illness were significantly related to children’s uncooperative behavior and worry.

As hypothesized, the number of COVID-19–related hardships was strongly associated with all psychological well-being measures (Table 4). Psychological well-being of those with only 1 hardship did not differ from those with 0 hardships. However, those with 2 and 3 hardships had significantly more negative mood, worse sleep quality, and more uncooperative child behavior than those with 0 hardships. Finally, both parent and child psychological well-being was worst among families with all 4 hardships, with large and statistically significant associations with each of the 4 family psychological well-being outcomes.

The outbreak of COVID-19 has profoundly affected American families, and hourly service workers with young children are one of the groups most vulnerable to the crisis. The speed with which the pandemic has changed society, however, has made it difficult to capture the impact of the shock in a timely manner. In this article, we are able to take advantage of real-time data on hourly service workers collected fortuitously both before and after the crisis started. Our results clearly show that the coronavirus crisis has substantially worsened adult and child psychological well-being. These percentage changes over time are substantial, especially given that they occurred over just a few weeks.

Parents reported deterioration in their own psychological well-being since the start of the crisis, in line with findings from studies of previous economic downturns, which reveal that adult mental health worsens when economic conditions deteriorate.^34–36  In that previous work, however, the authors were able to document that mental health worsens in response to economic downturns over longer periods, such as over many months. In contrast, our daily survey data enable us to demonstrate that psychological distress has increased in response to the current crisis in a matter of days. This is perhaps to be expected because our study is focused on a particularly vulnerable group of families and the current crisis is both severe and multifaceted. Tracing immediate impacts on psychological well-being would not be possible without this type of granular, daily data. These results should raise concern given the strong links between parental psychological well-being and the well-being of children.^37,38 

Our one-time survey results also reveal how the crisis has increased family hardships. More than half of the families we surveyed have experienced job loss, more than two-thirds have experienced income loss, 45% face increased caregiving burden, and 12% have had a sick family member since the crisis began. All of these burdens were predicted to increase because of the crisis and are putative mechanisms linking the crisis to psychological well-being.

As hypothesized, the cumulative number of COVID-19–related hardships was strongly related to both parent and child well-being, with those experiencing all 4 hardships experiencing substantially worse well-being than those without any hardships. These findings are aligned with a cumulative risk perspective,¹³  which suggests that risk factors additively predict worsened mental health.

Although our study demonstrates the impact of COVID-19 on a vulnerable population, the sample population was limited and targeted. Hourly workers in other industries and salaried workers may be affected differently by the pandemic. In addition, our findings are local to a particular major city. The experience of the pandemic may differ from city to city on the basis of infection prevalence and governmental and social response. In this study, we do not provide insight into what policies would best mitigate these effects and support impacted workers. Additional research on support systems and systemic response to unemployment, uncertainty, and their negative health impacts should be referenced by individuals or institutions looking to take action.

These results highlight the severe immediate impacts of the COVID-19 crisis on vulnerable families. This crisis has harmed the psychological well-being of these families, suggesting the need for immediate increases in social support and for additional interventions aimed at addressing the economic and mental health needs of families.

As the crisis continues to unfold, pediatricians should screen for mental health problems among the children in their practices, with particular attention to children whose families are vulnerable to economic as well as disease aspects of the crisis. Our results reinforce the importance of considering both internalizing and externalizing behaviors as manifestations of mental health in young children. When parents present with a complaint about uncooperative behavior, they may not be seeking mental health guidance. Pediatricians, however, should consider helping parents understand that uncooperativeness may be a sign that children are being stressed by the pandemic. All adults providing services to young children may be more effective in supporting children’s mental health during this pandemic if they focus on understanding and empathy.

We thank Kenneth Dodge for helpful comments on a previous version of this article. Excellent project support was provided by Jennifer Copeland.

COVID-19

coronavirus disease 2019

OLS

ordinary least squares

VDT

venue (business) day-time unit