This was a cross-sectional study using survey data from the National Survey of Children's Health (NSCH) 2016–2017 (14) and 2018–2019 (14). The NSCH aims to provide national and state-level data on the physical and emotional health of children in the United States. In this study, a total of 818 children who had been or were currently diagnosed with epilepsy were extracted from the NSCH 2016–2017 and 2018–2019. Children who were underweight and had missing study variables were excluded.

A total of 131,774 surveys were completed from 2016 to 2019 combined, 50,212 surveys were completed in 2016, 21,599 in 2017, 30,530 in 2018, and 29,433 in 2019. Survey data were weighted and adjusted to represent the national and state populations of non-institutionalized children aged 0–17 living in housing units in each state and nationally. The Overall Weighted Response Rate was 40.7% for 2016, 37.4% for 2017, 43.1% for 2018, and 42.4% for 2019 (15, 16).

Since we analyzed the de-identified public database released by the Data Resource Center for Child and Adolescent Health (DRC), our study was deemed exempt from the institutional review board at the researchers' institution. For more details on NSCH data availability and ethics, visit (www.childhealthdata.org).

Since 2016, the NSCH has been administered online and by mail. Families with one or more children under 18 years old were randomly selected, and one child from each family was randomly selected as the respondent through the main topical questionnaire. The adult in the family who was familiar with children's physical condition and care was invited to complete a short screening questionnaire and a more detailed online or written 10–17 aged questionnaire (17).

The diagnosis of epilepsy was reported by respondents (primary caregiver) in the NSCH survey as “a doctor or other health care provider ever told you that this child has epilepsy or seizure disorder?” then “if yes, does this child currently have the condition?” then, “if yes, is it mild, moderate, or severe?” The severity of the seizures was reported by the primary caregiver.

The outcome variable was overweight and obesity in children 10–17 years. Parents or guardians were asked, “What is this child's current height?” And “How much does this child currently weigh?” Data on height and weight will not be published separately, but they were combined into body mass index (BMI), which was released. The weight status of children and youths was defined by the BMI percentile of age and gender. “Normal weight” status was determined by BMI between 5th and 84th percentile, “Overweight” status was at or greater than 85th to less than 95th percentile, and “Obesity” status was at or greater than 95th percentile (18).

Social determinants included children's behavioral, and socioeconomic characteristics. According to the federal poverty level (FPL), the household poverty level was divided into four levels, below 100%, 100%–200%, 200%–400%, and above 400%. Parents' education was chosen by themselves from less than high school, high school or General Educational Development (GED), some college or technical school, college degree or higher.

Children's behavioral data included physical activity, afterschool activities, volunteering, screen time, and physical condition. (1) Physical activity was defined as “How many days did this child exercise, play a sport, or participate in physical activity for at least 60 min during the past week” (0 days, 1–3 days, 4–6 days, every day). (2) Afterschool activities were defined as “Did this child participant in a sports team or did he or she take sports lessons after school or on weekends?” (yes/no). (3) Volunteer was defined as “Any type of community service or volunteer work at school, church, or in the community” (yes/no). (4) Screen time: in the 2016–2017 questionnaire, there were two questions about screen time for children. “On an average weekday, about how much time does this child usually spend in front of a TV watching TV programs, videos, or playing video games?” And “spend with computers, cell phones, handheld video games, and other electronic devices, doing things other than schoolwork?” In the 2018–2019 questionnaire, NSCH merged the two problems and the question was “on most weekdays, about how much time did this child spend in front of a TV, computer, cellphone or other electronic device watching programs, playing games, accessing the internet or using social media? Do not include time spent doing schoolwork.” Response options are none, less than 1 h, 2–3 h, 4 or more hours (19). (5) Physical condition: In general, the parent would be asked, “how would you describe this child's health?” Parents choose from excellent, very good, good, fair, and poor and filled in the questionnaire as the basis for describing the physical conditions of the child.

Demographic data included sex (male/female), age [preadolescence (less than 13 years)/adolescence (13–17 years)], and race (Hispanic/non-Hispanic white/non-Hispanic black/non-Hispanic Others). Babies born before 37 weeks gestation were the major predictor of premature birth (20). So, the question “Was the child born more than 3 weeks before his or her due date?” (yes/no) was in the survey, which was used to assess premature birth. And, “how much did he or she weigh when born?” Answer in pounds and ounces or kilograms and grams to judge whether the birth weight was low. The NSCH dataset included parent reports of birth weight which was categorically recorded as not born with low weight (greater than 2,500 g), born with low weight (less than 2,500 g), and born with very low weight (less than 1,500 g) (Bureau).

Sleep duration would be examined “During the past week, how many hours of sleep did this child get on an average day?” (less than 6 h, 6 h, 7 h, 8 h, 9 h, 10 h, 11 or more hours). Children under 18 years of age should sleep at least 8 h per 24 h on a regular basis to promote optimal health according to the study, and younger children need more sleep (21).

Questions like “Lived with anyone who had a problem with alcohol or drugs (yes/no)” and “Does anyone living in your household use cigarettes, cigars, or pipe tobacco? if yes, does anyone smoke inside your home? (yes/no)” were questioned about the family environment.

STATE and SAMPLE variables as strata, IDNUMR as the cluster, and NSCHWT as the weight were used in the analyses. The study population was divided into three groups (normal, overweight, and obesity) according to BMI and descriptive analysis was utilized. Counting data were described by cases and the weighted constituent ratio [n (weighted%)], and differences between groups were evaluated by the Chi-square test. Variables with significant differences between groups were included in the multivariate ordinal logistic regression model to explore the correlation between social determinants and overweight/obesity in children with epilepsy. Model 1 was the crude model, Model 2 adjusted for sex, age, and race, and Model 3 adjusted for sex, age, race, use of cigarets, cigars, or pipe tobacco inside, afterschool activity, and physical activity. Children with epilepsy were divided into three subgroups: those who were currently asymptomatic, mild, and moderate/severe symptoms to explore the correlation between socioeconomic factors and overweight/obesity. All the results presented are weighted. Statistical tests were conducted by bilateral tests, and P < 0.05 was statistically significant. Statistical analyses were performed by SAS v. 9.4 (SAS Institute, Cary, North Carolina) software. The proc power module in SAS v. 9.4 was used for power calculation.

There were 818 children with previous or current epilepsy in NSCH 2016–2017, 2018–2019, we excluded: (1) 84 underweight children (BMI < 5%); (2) 79 children with missing one or more data that we need for the following research. Finally, Data were analyzed on 655 children and youths with epilepsy aged 10–17, which was weighted to represent 247,823 children (Figure 1). We adopt the simple deletion method to deal with them due to these missing values in the data, then sensitivity analysis was conducted to compare the original data with the deleted data between groups. All the P-values were >0.05 in Supplementary Table S1, it showed no difference in observations that were excluded because of missing or invalid information.

Of the children included in the analysis, 423 were categorized into the normal weight group (BMI 5th–84th), 108 were categorized into the overweight group (BMI 85th–94th), and 124 were categorized into the obesity group (BMI ≥ 95th). There were 331 (52.64%) males and 324 (47.36%) females, 186 (34.15%) people under the age of 13, and 469 (65.85%) people between the ages of 13–17. Children and youths who do not currently have conditions [295 (42.84%)] accounted for the majority, followed by current conditions rated mild [191 (31.53%)] and current conditions rated moderate or severe [169 (25.63%)]. Characteristics of children with epilepsy in this study were shown in Table 1.

Results indicated statistically significant differences (P < 0.05) across several selected variables, physical conditions (P = 0.012), poverty status (P < 0.001), parents' education (P = 0.004), seizure severity (P = 0.033), and volunteer (P = 0.031) in Table 1. Furthermore, there was no difference in physical activity, birth data, parental alcohol or drug use, and other factors among normal, overweight, and obese children with epilepsy, P > 0.05.

In Figure 2, parents' education was significantly correlated with obesity of the surveyed children in Model 3 (adjusted for sex, age, race, use of cigarets, cigars, or pipe tobacco inside, afterschool activity, and physical activity), odds ratio (OR) of high school or GED was 0.365 [95% confidence interval (CI), 0.142–0.934, P = 0.036], OR of college education or higher was 0.382 [(95% CI, 0.165–0.888), P = 0.026], less than high school as the reference. When the household poverty level was less than 100% FPL as the reference, greater than 400% FPL was negatively related to childhood obesity, the adjusted OR = 0.274 [(95% CI, 0.111–0.677), P = 0.005]. Children whose parents thought they were in poor or fair health were significantly correlated with obesity compared with those in good health, with the adjusted OR = 1.573 [(95% CI, 1.164–2.125), P = 0.003].

Because our study population was children with epilepsy, we stratified them according to seizure severity, Figures 3–5 described results from the logistic regression for the three subgroups. Children with poor physical conditions reported by their parents were correlated with obesity in Model 3, OR of three subgroups (do not currently have the condition, current condition rated mild, current condition rated moderate or severe) were 1.612 [(95% CI, 1.004–2.588), P = 0.048], 2.221 [(95% CI, 1.239–3.981), P = 0.008], and 1.704 [(95% CI, 1.037–2.802), P = 0.036], respectively.

Among do not currently have condition children, their parents who graduated from high school or GED [OR = 0.101 (95% CI, 0.024–0.423), P = 0.002], some college or technical school [OR = 0.141 (95% CI, 0.035–0.572), P = 0.006], and college or higher [OR = 0.108 (95% CI, 0.025–0.473), P = 0.003] were all negatively correlated to childhood obesity with less than high school as the reference (Figure 3).

In the study population with current condition rated mild, poverty level less than 100% FPL as a reference, poverty level between 200% and 400% FPL [OR = 0.156 (95% CI, 0.028–0.876), P = 0.035] or above 400% FPL [OR = 0.121 (95% CI, 0.023–0.641), P = 0.013] were related to childhood obesity (Figure 4). Figure 4 demonstrated that not volunteering was also associated with obesity compared to children who volunteered [OR = 3.214, (95% CI, 1.089–9.486), P < 0.001].

For children with current conditions rated moderate or severe, poverty level less than 100% FPL was taken as the reference, as shown in Figure 5, there was a negative association between obesity and poverty level above 400% FPL, OR = 0.103 (95% CI, 0.023–0.460), P = 0.003.