Factors associated with attention deficit/hyperactivity disorder among US children: Results from a national survey

NSCH is a population-based, cross-sectional, random-digital-dialing survey using a complex, multi-cluster, probability sampling design [33]. This survey was sponsored by the Department of Health and Human Services (DHHS), Maternal and Child Health Bureau in partnership with the National Center for Health Statistics, which is a part of the Center for Disease Control and Prevention [34]. The data were collected by random-digital-dialing households with children under 18 years of age from each of the 50 states and the District of Columbia during the period of April 2007 to July 2008. A child was randomly selected from each of the sampled households to be the subject of the survey. The parent or guardian who knew the most about the selected child’s health and health care was interviewed with questions over the telephone. The interview was conducted in both English and Spanish, and consisted of questions regarding demographics, child’s health and functional status, health insurance coverage, health care access and utilization, medical home, family functioning, parental health, and neighborhood/community characteristics. A total of 91,642 children of age 17 years or under, with an overall weighted response rate of 55.3% were included in NSCH 2007 study. Population based estimates were obtained by assigning sampling weights to each sampled child for whom an interview was completed. Detailed information about the design of the NSCH is available at http://​www.​nschdata.​org.

According to the DSM-IV, the diagnosis of ADHD requires the child to experience ADHD characteristics such as lack of attention, impulsive behavior, and hyperactivity in at least two different settings, namely home and school; while previous studies also show that the earliest onset age for a child to be diagnosed and treated for ADHD can be below 7 years [35]. As one of the two settings required for the diagnosis of ADHD is school, we considered children between 5 to 17 years old to be our study sample. There were a total of 68,634 responses given by the parents or guardians of children aged 5 to 17 years. Since the NSCH is a population based survey, the selected sample is a representative sample of children aged 5 to 17 years in the US.

The primary dependent variable used in our study was ascertained through the following question to the parent or guardian: “Has a doctor or health professional ever told you that selected child (S.C.) has attention deficit disorder or attention deficit hyperactive disorder, that is, ADD or ADHD?”. We categorized the variable into: ‘ADHD’ if the response was ‘Yes’, and ‘No ADHD’ if the response was ‘No’. The psychological factors included in the study were depression (DEP; “Has a doctor or other health care provider ever told you that S.C. had depression?”) and anxiety (ANX; “Has a doctor or other health care provider ever told you that S.C. had anxiety problems?’). The NSCH has a sex- and age-specific derived BMI variable that is categorized into underweight (<5^th percentile), normal weight (≥5^th and<85^th percentile), at risk of overweight (≥85^th and<95^th percentile) and overweight (≥95^th percentile). To explicate, we refer to the last two categories as overweight and obese, respectively. Other independent variables included in the analysis were sex, race/ethnicity (Non-Hispanic White vs. Non- Hispanic Black, Hispanic, Other), highest level of education (EDUC) attained by anyone in the household (More than high school vs. High school graduate or less), family structure (FAMILY; Two parent including biological, step, or adopted vs. Other including single mother), dichotomized poverty level based on DHHS Federal Poverty guidelines (POVERTY; ≤ 200% vs.> 200%), healthcare coverage (INS; Having any health insurance such as Health Maintenance Organizations (HMOs) or Medicaid vs. Not having healthcare coverage), participation in sports (SPORTS; S.C. on a sports team or taking sports lessons after the school or on weekends vs. Not participating in sports), participation in clubs (CLUBS; S.C. on any clubs or organizations after school or on weekend vs. Not participating in clubs), daily average computer usage (COMP) for purposes other than school work (≥ 1 hour vs.< 1 hour), daily average television watching/playing video games (TV; ≥ 1 hour vs.< 1 hour), family member’s smoking status (SMOKE; Cigarettes, cigars, or pipe tobacco used by anyone in the household vs. No one in the household smokes) [17‐32].

We also considered the effect of medication on the association between ADHD and BMI, as most of the medications prescribed for ADHD may have anorectic effects [36]. To do this we utilized the question: “Is S.C. currently taking medication for ADD or ADHD?”. We combined responses to this question with the ones to the ADHD question mentioned earlier to create a variable with three categories: ‘ADHD and currently taking medication’ (ADHD-CM), ‘ADHD and not currently taking medication’ (ADHD-NCM), and ‘No ADHD’; ‘No ADHD’ was considered as the reference group. This trichotomized variable is used as a dependent variable in one of the models. We chose not to use this variable as our primary dependent variable due to several limitations with the medication use question and its discrepancy in sample size with the ADHD question (elaborated later in the Discussion section).

We computed descriptive statistics based on unweighted sample sizes and weighted percentages for children aged 5 to 17 years. Next, we used chi-square tests for testing the association of each variable with ADHD. Then we performed logistic regression analysis with ADHD as the response variable and the independent variables specified above, first with each of the variables individually (bivariate analysis) and then with all of them in the model (multivariate analysis) to obtain unadjusted and adjusted odds ratios (OR), respectively. For the trichotomized dependent variable that incorporates the medication effect, we used the multinomial logistic regression with independent variables same as in our previous model. All analyses were carried out in SAS version 9.2 to account for the complex survey design of the study [37, 38]. In particular, we performed domain analysis for the 5–17 years age group [39].