Analytic Sample and Data Source
Our sample children aged ≥2 and £ 18 years (n = 13, 363) in 2019 and 2020, representing 130,582,543 children nationwide. The survey year 2019 was the year immediately before the COVID-19 pandemic. Our data measures the impact of COVID-19 on children’s mental health and behaviors after the nation-wide shut down and documents the life changes in children including school closures, virtual school classes and the closure of business dedicated to children’s recreation. Testing for COVID-19 was not available in the early months of 2020, and numbers of COVID-19 reported positive cases likely underrepresents the numbers of those who were actually infected in 2020 (Liu, Liu, Liu, Luo, & Xia, 2020). Nonetheless, the data represents an opportunity to measure children’s mental health during the period of national quarantine.
The NHIS is the principal data source for the health of the US civilian noninstitutionalized population (National Health Interview Survey [NHIS], 2021). The NHIS is conducted through home-based interviews using clustered sampling techniques to select dwelling units. A knowledgeable adult 18 or older responds for him/herself and all other family members (Family Respondent). One adult 18 or older (Sample Adult) and one child under age 18 (Sample Child) are randomly selected for more extensive set of questions. The sample adult answers for his/herself, while a knowledgeable adult answered for the child (NHIS, 2021). The NHIS is a multistage probability sample. Sampling weights were used to produce representative estimates allowing each person’s participation to be inflated to represent the total US population (IPUMS NHIS, n.d.). We report both unweighted and weighted statistics
Due to the COVID-19 pandemic the 2020 NHIS data collection was modified to consist of four separate designs: primarily in-person (Q1); telephone only (Q2); telephone attempts first with some in-person where feasible (Q3 and Q4) (NHIS, 2021). The NHIS employed a multistage probability sample that incorporates stratification and oversampling of some subpopulations. Because the sampling methodology was the same in 2019 and 2020, and was designed to represent the US population, and the study years were sequential, changes in the composition of the underlying population from 2019 to 2020 is not expected (NHIS, 2019; Integrated Public Use Microdata Series [IPUMS] NHIS, n.d.).
Measures
Sociodemographic Characteristics
Children were categorized as male or female, by race and ethnicity (Hispanic, non-Hispanic White [White], non-Hispanic Black [Black], and non-Hispanic Asian [Asian]), age (2-4, 5-11, 12-14, 15-18), family structure of the child, region of residence at the time of interview (Northeast, Midwest, South, West), and education level of a family respondent (<high school graduate, high school graduate or General Educational Development [GED] equivalent, some college or an associate’s degree, bachelor’s or master’s degree, doctoral or professional degree). Marital status and region of residence are associated with health (McCarthy et al., 2012; Weissman & Russel, 2016).
Health-Related Characteristics
COVID-19 Positivity Status
COVID-19 status was based on a “yes” response to “has a health professional told you that you had COVID-19?” The question is only in the 2020 NHIS survey.
Place of Health Care
Respondents were asked “what kind of place do you receive your health care- a doctor's office or health center; an urgent care center, a clinic in a drug store or grocery store; a hospital emergency room; a VA Medical Center or outpatient clinic; or some other place?” A positive response for any one option was recorded.
Healthcare Access, Coverage and Utilization Indicators
Private insurance was defined as coverage through employer(s), union(s), or purchase. Public insurance was defined as Medicaid and Medicare. Persons without private or public insurance were considered uninsured. Some adults had both private and Medicare coverage. A variable called Health Coverage Type grouped insurance coverage as Medicaid, Private Insurance, and No Coverage. Additional access and utilization indicators were defined as follows: did not get medical care because of the coronavirus pandemic (did not get medical care due to the COVID-19 pandemic), delayed getting medical care because of the coronavirus pandemic (delay in medical care due to the COVID-19 pandemic), and had a virtual appointment in the last 12 months due to COVID-19.
Preventive Medical Care Indicators
Family respondents were asked: “About how long has it been since you last saw a health professional for a wellness visit?” Responses to having had a visit less than 12 months ago were coded as “1”. Responses that included “never”, “less than 3 years ago”, “3 years but less than 5 years ago, 5 years but less than 10 years ago, and 10 years ago or more were coded as “0”. Respondents were also asked “during the past 12 months, did you receive physical therapy, speech therapy, rehabilitative therapy, or occupational therapy? A response of “yes” was coded as “1” and “no” was coded as “0”.
Sleep Disturbances and Social Behavior
Frequency with which the child wakes up well rested
Family respondents were asked “How often does the child wake up well-rested? never, some days, most days, or every day?” A response of “some days”, “most days” or every day” was coded as “1” and “never” was recorded as “0”.
Frequency with which Child Woke up Tired
Family respondents were asked “How often does the child complain about being tired during the day? never, some days, most days, or every day?” A response of “some days”, “most days” or every day” was coded as “1” and “never” was recorded as “0”.
Frequency with which Child has Difficulty Getting Out of Bed
Family respondents were asked “how often does the child have difficulty getting out of bed in the morning? never, some days, most days, or every day?” A response of “some days”, “most days” or every day” was coded as “1” and “never” was recorded as “0”.
Social Behaviors
Respondents were asked “does your child have difficulty controlling their behavior?” Responses included “no difficulty” “some difficulty” “A lot of difficulty” or “cannot do at all”. The variable was also made binary with “no Difficulty” coded as “0” and the remaining responses coded as “1”. The child must be between 2 and 17 years of age.
Childhood Recreation
Respondents with children aged 2 to 4 years of age were asked “does your child have difficulty playing? Would you say your child has: no difficulty, some difficulty, a lot of difficulty, or cannot do this at all? This variable was also made binary with “No Difficulty” coded as “0” and the remaining responses coded as “1”.
Child’s Mental Health
Anxiety and Depression
Family respondents were asked “how often does the child seem very anxious, nervous, or worried? Would you say: daily, weekly, monthly, a few times a year, or never?” A response of “some days”, “most days” or every day” was coded as “1” and “never” was recorded as “0”. Children from ages 2 to 4 years were not assessed for anxiety and depression and were removed from the multivariable models but remained in the sample.
Statistical Analysis
We used SAS-callable SUDAAN11.0.3 (RTI International) to calculate point estimates and 95% confidence intervals. Rao–Scott χ2 statistics for weighted surveys evaluated categorical variables at α = .05 alpha level (2-sided) (Weissman et al., 2020). We compared proportions by COVID-19 positivity status, access and utilization indicators, sociodemographic characteristics including gender, race and ethnicity, income, family marital status, and region of residence, and health coverage type. For Table 2, we compared place of health care, health coverage types, preventive medical care indicators, sleep disturbances and mental health by survey years 2019 and 2020 (Sun, Shook, & Kay, 1996; Harrell, 2001).
The first multivariable logistic regression model predicted the likelihood of anxiety in 2020 compared to 2019. Independent variables included gender, race/ethnicity, age group, family structure of the child, highest family educational level, region of residence, and family income relative to the federal poverty line (FPL) poverty index ratio (PIR). In separate models, depression and difficulty with social behaviors were dependent variables and the independent variables were the same as Model 1.