Many studies have reported a higher prevalence of Attention-Deficit/Hyperactivity Disorder (ADHD) among disadvantaged populations, but few have considered how parental history of ADHD might modify that relationship. We evaluated whether the prevalence of ADHD varies by socioeconomic status (SES) and parental history of ADHD in a population-sample of elementary school children age 6–14 years. Methods: We screened all children in grades 1–5 in 17 schools in one North Carolina (U.S.) county for ADHD using teacher rating scales and 1,160 parent interviews, including an ADHD structured interview (DISC). We combined parent and teacher ratings to determine DSM-IV ADHD status. Data analysis was restricted to 967 children with information about parental history of ADHD. SES was measured by family income and respondent education. Results: We found an interaction between family income and parental history of ADHD diagnosis (p = .016). The SES gradient was stronger in families without a parental history and weaker among children with a parental history. Among children without a parental history of ADHD diagnosis, low income children had 6.2 times the odds of ADHD (95% CI 3.4–11.3) as high income children after adjusting for covariates. Among children with a parental history, all had over 10 times the odds of ADHD as high income children without a parental history but the SES gradient between high and low income children was less pronounced [odds ratio (OR) = 1.4, 95% CI 0.6–3.5]. Conclusions: Socioeconomic status and parental history of ADHD are each strong risk factors for ADHD that interact to determine prevalence. More research is needed to dissect the components of SES that contribute to risk of ADHD. Future ADHD research should evaluate whether the strength of other environmental risk factors vary by parental history. Early identification and interventions for children with low SES or parental histories of ADHD should be explored. Keywords: Attention- Deficit/Hyperactivity Disorder; epidemiology; family history; prevalence; social class.
Introduction For many years, Attention-Deficit/Hyperactivity Disorder (ADHD) was believed to equally affect chil- dren of all social classes; that is, risk was not believed to vary by socioeconomic status (SES) (Barkley, 1998; Polanczyk & Jensen, 2008). This perception has persisted despite strong evidence that ADHD is more common among the poor, which has been docu- mented in clinical samples (Biederman, Faraone, & Monuteaux, 2002;Counts,Nigg, Stawicki,Rappley,& von Eye, 2005), population-based epidemiologic studies (Lambert, Sandoval, &Sassone, 1978; Scahill et al., 1999; St Sauver et al., 2004; Szatmari, Offord, & Boyle, 1989b), meta-analyses (Russell, Ford, Wil- liams, & Russell, 2016), and large national surveys like the National Health Interview Study (Pastor & Rubin, 2008), and the National Survey of Children’s Health (Larson, Russ, Kahn, & Halfon, 2011).
Given the strength of the evidence, it is puzzling that controversy exists; but two factors may play a role.
First, low-incomechildrenwithouthealthcarecoverage are less likely to be treated for ADHD (Froehlich et al., 2007), which skews the pool of clinically identified patients toward more affluent children. The second factor is historical; the authors of two influential population-based studies of ADHD minimized SES in the etiology of ADHD when interpreting their results (Lambert et al., 1978; Szatmari, Offord, & Boyle, 1989a). Citing these two studies, a leading textbook concluded that the impact of SES on the prevalence of ADHD remains an unresolved issue (Barkley, 2015).
In one of those studies, researchers sampled 5,200 children in 191 classrooms (Lambert et al., 1978). They reported that hyperactive children (the term for ADHD at the time) were disproportionately from poor families when they identified cases using parent and teacher ratings, but this association weakened when they required an additional physician diagnosis. The authors concluded that the diagnosis depended on the source of the diagnostic information (Lambert et al., 1978). However, by requiring physician diagnosis, they may have obscured the SES effect because physician access is strongly influenced by ability to pay.Conflict of interest statement: S
The second study evaluated over 2,500 children and reported a 2.8-fold increase in the odds of ADHD for participants on welfare (Szatmari et al., 1989a). However, after adjusting for developmental delays and comorbidity, the poverty effect disappeared for boys but persisted for girls. This discrepancy was difficult to explain and the authors discounted the SES finding in their conclusions. However, develop- mental delays are a common precursor to ADHD and comorbidity is concurrent with ADHD in about two- thirds of cases (Larson et al., 2011).
By statistically controlling for these two risk fac- tors on the causal pathway, the researchers may have obscured some of the SES effect. In both the epidemiology and the psychology literature, method- ologists have cautioned against simple statistical adjustment without careful thought about the underlying causal model (Beauchaine, Hinshaw, & Pang, 2010; Weinberg, 1993).
Two recent studies, and an accompanying edito- rial, have challenged the perception that ADHD affects all children equally, and prompted a reeval- uation of the relationship between social class and ADHD risk (Larsson, Sariaslan, Langstrom, D’Ono- frio, & Lichtenstein, 2014; Nigg & Craver, 2014; Russell, Ford, Rosenberg, & Kelly, 2014). Larsson et al. (2014) found an inverse relationship between SES level and ADHD risk in a population-based cohort of over 800,000 Swedish children after accounting for co-occurring risk among siblings and cousins to assess ‘unmeasured familial con- founding’. Russell et al. (2014) reported even stron- ger SES effects among a cohort of 19,000 British children using multiple measures of poverty. They also ruled out reverse causation; families who had a child with ADHD were no more likely than families without a child with ADHD to drop in their socioe- conomic position over time. Finally, Russell et al. presented evidence that parental attachment or family conflict partially mediates the association between SES and ADHD risk. These two papers have refocused attention on the possible importance of SES in the etiology of ADHD, but not all observers consider the matter settled (Faraone et al., 2015).
Neither the Larsson nor the Russell paper accounted for parental history of ADHD, even though the estimated heritability of ADHD is as high as 71%–90% (Thapar, Cooper, Eyre, & Langley, 2013) and family history is an established risk factor for ADHD. A study using the Swedish National registry system reported that the incidence of externalizing and internalizing disorders was higher in high- deprivation neighborhoods but familial effects accounted for nearly eight times the total variation in externalizing disorders as neighborhood-level effects (Sundquist et al., 2015). One clinic-based study reported a two- to eightfold increased odds of ADHD in children whose parents had ADHD as children (Faraone et al., 2005), and another reported odds ratios (ORs) of between 2.5 and 4.6 for fathers
with childhood ADHD and between 15 and 23 for mothers with childhood ADHD (Chronis et al., 2003). More recently, Musser et al. (2014) used a population-based sample to evaluate the risk of ADHD among children of mothers with ADHD. (Data on fathers were not available.) They reported the odds of having a child diagnosed with ADHD were fivefold greater (OR = 5.0, 95% CI 3.7–6.8, p < .001) among mothers with a prior ADHD diagnosis.
A recent study of 132 children collected histories of low birthweight and prenatal smoking, adverse life events, deviant peer affiliation, parental history of ADHD, and socioeconomic status in three groups: ADHD, ADHD with oppositional behavior, and con- trols (Noordermeer et al., 2017). The researchers reported that adverse life events and parental history of ADHD were important risk factors for ADHD but low SES was not, after adjusting for the other cofactors. However, the inclusion in a single statis- tical model of low SES, adverse life events, perinatal adversities and deviant peer affiliation, which may all be on the same causal pathway, make the results difficult to interpret.
Relatively few studies have explored how family history of ADHD might modify risk factors for ADHD, an idea with new importance based on recent advances in understanding of gene-environment interactions and epigenetics (Faraone et al., 2015; Meaney, 2010). One study reported that perinatal risk factors for ADHD were strongest among women without a family history of ADHD (Sprich-Buckmin- ster, Biederman, Milberger, Faraone, & Lehman, 1993). A recent study reported that levels of cognitive impairment were higher among youth with ADHD who had multiple family members with ADHD (Oer- lemans et al., 2015). In one of the few studies to look at the joint effects of parental history of ADHD and child adversity on the risk of ADHD, the authors reported that the relationship between maternal ADHD and child ADHD was stronger in families with less adversity (Breaux, Brown, & Harvey, 2017). Each of these studies suggests that a child’s family history of ADHD may modify the impact of other risk factors. By exploring the joint effects of SES and family history, we hope to gain insight into how risk factors may combine to shape the complex etiology of ADHD.
Here we use a population-based sample of ele- mentary school children in Johnston County, North Carolina, to examine the joint effects of SES and family history of ADHD on the prevalence of ADHD. We focus on these risk factors because of the controversy about the role of SES, the strength of family history as a risk factor for ADHD, and the few studies suggesting that family history of ADHD may modify risk factors for ADHD. We hypothesized that ADHD would be more common among the poor and among those with a family history of ADHD but we did not have enough information to hypothesize the direction of the joint effects.
Sampling and screening procedure
The NIEHS IRB approved the original study protocol and subsequent data analysis was approved by the University of New Mexico IRB. Our overall sampling goal was to screen children for ADHD in two stages: first by identifying potential cases using teacher ratings and later by parent telephone interview. We then combined parent and teacher ratings to determine case status.
There were 7,847 children enrolled in grades 1–5. After excluding children with severe developmental disabilities in self-contained classrooms or severe health problems, 7,587 children were eligible. Parents or guardians of 6,139 (81%) children gave written permission, and teachers completed a behavioral checklist to rate ADHD symptoms and impair- ment (Rowland, Umbach, Bohlig, Stallone, & Sandler, 2007) on 6,072 children. After excluding 411 children with severe medical disabilities or low English proficiency, there were 5,661 children eligible for the parent interview. We identi- fied 1,414 children as potential cases; these were children taking medication to treat ADHD or who often exhibited at least 3 DSM-IV hyperactive/impulsive or inattentive behav- iors and impairment at school on teacher ratings. We also selected a random sample of the eligible children (N = 706) as potential controls. These two groups had 169 children in common. At this stage, before parental interviews had been completed, both groups contained some children who would later screen positive and become ADHD cases. (A detailed explanation and sampling diagram is available.) (Rowland et al., 2013).
Of the 1,951 potential cases and potential controls, we attempted to interview parents of 1,619 children (randomly eliminating 332 to meet budget constraints) and completed 1,160 telephone parental interviews (71.6%). The parental interview included the ADHD module of The NIMH Diagnostic Interview Schedule for Children (DISC-2.3) (Shaffer et al., 1996).
After the parent interview, we combined parent and teacher reports to determine final case status using full DSM-IV criteria of six hyperactive/impulsive symptoms or six inattentive symp- toms, impairment in two settings and severe impairment in at leastonesetting (Rowlandet al., 2013). Incombiningparentand teacher ratings, we required six or more symptoms in all with three or more symptoms from the teacher screener and three or more symptoms from the parent interview; symptoms reported by both informants were only counted once.
We did not use the DSM-IV ADHD age 7 age-at-onset criterion because of difficulty establishing an accurate age at onset, particularly for the Predominately Inattentive Subtype (Barkley & Biederman, 1997), which may have inflated our prevalence estimates slightly. Youth taking stimulant medica- tion were treated as cases if they met DSM criteria while taking medication or if their parents reported they met symptom and impairment criteria the year prior to beginning medication treatment.
Children in the random sample of potential controls, who met ADHD case criteria after the phone interview, were classified as cases. After removing these cases, those remain- ing were our controls; they represent a random sample of the entire population who did not meet ADHD case criteria.
Some of the children eligible for parental interview were siblings. We randomly selected an index child from each family with multiple siblings. We collected a full interview from the
parent on the index child but, to reduce respondent burden, only an abbreviated interview on other siblings.
Of the 1,160 completed parental interviews, we had 967 full interviews from birth mothers and 127 interviews from non- birth parents (i.e., foster parents, step-parents and grandpar- ents), and 66 children with abbreviated parental interviews. We restricted our sample to the 967 index children living with their birth mothers because only for these children had respondents been asked about parental history of ADHD. The final sample contained 392 ADHD cases, 393 controls, and 182 children who had been classified as potential cases based on the teacher rating, but did not meet case criteria after the parent interview, and were not in the random sample of potential controls.
Validation of ADHD Case definition: We validated our case definition using three approaches. First, we compared our ADHD prevalence estimate to the estimate obtained from a pilot study the previous year on our same population; very similar estimates were obtained (Rowland et al., 2001, 2013). Second, we compared our prevalence estimate to estimates reported in two population-based studies of the prevalence of ADHD in South Carolina and in Oklahoma (Wolraich et al., 2012). When we used similar assumptions about counting ADHD medication use and age of onset, our prevalence estimate of 9.3% (Rowland et al., 2013) fell between the estimates of those two studies. Finally, we conducted a validation study of 34 cases comparing our epidemiologic case definition to a clinical consensus diagnosis process that included a semistructured clinical interview (KSADS-PL) with parent and child, as well as rating scales and extra testing. Three cases identified as ADHD by our epidemiologic method were found to have been misidentified (Rowland et al., 2013), indicating a false positive rate near 9%.
Parental history of ADHD. We measured parental his- tory of ADHD in two ways.
Parental history of ADHD diagnosis: Birth mothers were asked ‘Has a doctor or psychologist ever said that the Child’s birth father had ADHD, Attention Deficit Disorder (ADD) or hyperactivity?’ and ‘Has a doctor or other health provider ever told you that you had ADD or Attention Deficit Hyperactivity Disorder (ADHD)?’ If they answered yes to either question, their child was considered to have a positive parental history of ADHD.
Parental history of ADHD symptoms: Birth mothers were asked four questions: (a) ‘When you were a child, did you often have trouble paying attention or concentrating in school?’ (b) ‘When you were a child, did you often have problems because you were overactive, fidgety or impulsive (that is doing things that might be dangerous or get you into trouble without thinking about them first)?’ The next questions are about Child’s birth father’s activity and attention. (c) ‘As far as you know, when Child’s father was a child, did he often have trouble paying attention or concentrating in school?’ (d) ‘As far as you know, when Child’s father was a child, did he often have problems because he was overactive, fidgety or impulsive (that is doing things that might be dangerous or get him into trouble without thinking about them first)?’ If respondents answered yes to any of these four questions, their child was considered to have a positive parental history of ADHD symptoms.