Social adversity predicts ADHD-medication in school children – a national cohort study

Authors

  • A Hjern,

    1. Centre for Epidemiology, National Board of Health and Welfare, Stockholm, Sweden
    2. Centre for Health Equity Studies (CHESS), Karolinska Institutet, Stockholm, Sweden
    Search for more papers by this author
  • GR Weitoft,

    1. Centre for Epidemiology, National Board of Health and Welfare, Stockholm, Sweden
    2. Department of Clinical Neurosciences, Karolinska Institutet, Stockholm, Sweden
    Search for more papers by this author
  • F Lindblad

    1. Department of Neuroscience, Child and Adolescent Psychiatry, University Hospital of Uppsala, Uppsala, Sweden
    2. Stress Research Institute, Stockholm University, Stockholm, Sweden
    Search for more papers by this author

Anders Hjern, Centre of Epidemiology, Swedish National Board of Health and Welfare, 106 30 Stockholm, Sweden.
Tel: +46-8-55 55 31 69 |
Fax: +46-8-55 55 33 27 |
Email: anders.hjern@socialstyrelsen.se

Abstract

Aims:  To test the hypothesis that psychosocial adversity in the family predicts medicated ADHD in school children.

Method:  ADHD-medication during 2006 was identified in the Swedish Prescribed Drug Register in national birth cohorts of 1.1 million 6–19 year olds. Logistic regression models adjusted for parental psychiatric disorders were used to test our hypothesis.

Results:  There was a clear gradient for ADHD medication with level of maternal education, with an adjusted odds ratio of 2.20 (2.04–2.38) for the lowest compared with the highest level. Lone parenthood and reception of social welfare also implied higher risks of ADHD-medication with adjusted ORs of 1.45 (1.38–1.52) and 2.06 (1.92–2.21) respectively. Low maternal education predicted 33% of cases with medicated ADHD and single parenthood 14%.

Conclusions:  Social adversity in the family predicts a considerable proportion of ADHD-medication in school children in Sweden.

Introduction

There is considerable evidence that genetic factors are important in the aetiology of ADHD, although the specific genes involved are yet to be identified (1). Twin studies demonstrate a high concordance rate, indicating an important role for genetic factors in the aetiology of ADHD (2,3). Studies of anti-social behaviour, with a gene-environment design, have widened the perspective of heritable mechanisms considerably by suggesting that preventable environmental factors such as ‘parenting style’ and neglect, to a certain extent, may mediate heritable effects on antisocial behaviour (4). A recent study by Lasky-Su et al. (5) suggests a more direct interaction between environmental and genetic factors in the development of ADHD. In this study, the socioeconomic status of the family predicted inattentive, but not hyperactive-impulsive, symptom counts in children exposed to the genetic risk factor brain-derived neurotrophic factor (BNDF). Case–control studies of ADHD in comparatively small samples (2,6) as well as one population-based national survey of hyperkinetic syndrome in the UK (7) have demonstrated an association between psychosocial adversity and these syndromes in school children.

Swedish national registers offer good prerequisites for studies of family-associated risk factors without the selective attrition, which so often complicates the interpretation of studies in this field. Whole age cohorts of children and youth can be identified and children linked to psychiatric and socioeconomic risk factors of the household. Since July 2005, these registers also contain information on individually linked prescriptions of drugs, opening up for identifying study groups who are usually not admitted to in-patient care, but can be identified by prescriptions of specific drugs or groups of drugs. ADHD is such a group, characterized by treatment based on a few drugs, used exclusively for this indication. In this national cohort study, we wanted to determine to what extent socioeconomic characteristics of the family predict ADHD-medication in Swedish school children and to what extent these social factors interact with exposure for parental traits for psychiatric disorders.

Methods

This study was based on Swedish national registers held by the National Board of Health and Welfare and Statistics Sweden. All Swedish residents are assigned a unique ten digit ID number at birth or immigration. This ID was used to link information from different register sources.

Study population

The study population was created from the Swedish birth cohorts of 1987–2000, who were alive and residents in Sweden on December 31st 2005 according to the Register of the Total Population (RTP). From these individuals, we excluded the 24.2% with a record of having at least one foreign-born parent according to the Multi-Generation Register, thus leaving 1 162 524 native Swedish children to be included in the final study population. Offspring of foreign-born parents were excluded from the study population because of the uncertain quality of the information about parental education in these families.

Sociodemographic variables

Sociodemographic variables were created from several register sources. Information on the educational level of the mother, or if not present in the register, the father, in 2005 was obtained through linkage to the Swedish National Education Register. The Swedish National Education Register was established by Statistics Sweden in 1985 and is annually updated with information on the highest formal education attained by each individual, from elementary to post-graduate level. Educational attainment was classified into four categories: having 9 years or less of schooling, representing primary school only, 10–12 years, equivalent to secondary school, 13–14 years i.e. short post-secondary education and 15+ representing a university education. A dichotomous variable of high education, indicating post-secondary education was created for interaction analyses.

Lone parenthood of the head of the household of the children in the study was obtained from RTB in 2005 and income from social assistance was obtained through linkage to the Total Enumeration Income Survey of 2005. Social assistance in Sweden is a form of cash income allowance from local social authorities, after a thorough means investigation, with the purpose to guarantee the applicant a minimum standard of living.

Parental psychiatric/addictive disorder

A dichotomous variable (yes/no) of psychiatric and/or addictive disorders in at least one of the parents was created through linkage to the Swedish Hospital Discharge Register (SHDR) for the years 1973–2005 and the National Cause of Death Register (NCDR) from 1986 to 2005 with the aid of the, 9th and 10th revisions of the WHO International Classification of Diagnosis (ICD 8–10). Yes included the categories; Suicide death or attempt, defined as an underlying cause of death in the NCDR or a hospital discharge with an external cause diagnosis of E950-E959 or E980-E989 (ICD-9) during 1973–1996 and X60-X84 or Y10-Y34 (ICD-10) in 1997–2005, Psychotic and affective disorders, defined as at least one discharge or cause of death diagnosis of F20-F39 in ICD 10), and Addictive disorders. The addictive disorders included a discharge from a hospital or a cause of death diagnosis indicating a psychiatric or somatic disorder associated with alcohol addiction, excessive alcohol consumption or illicit drug abuse.

Drug variables

The Swedish Prescribed Drug Register contains data, with unique patient identifiers for all drugs prescribed and dispensed to the whole population of Sweden (more than 9 million inhabitants) since July 2005. The purchase of at least one prescription of a drug with an Anatomical Therapeutic Chemical (ATC)-code of NO6BA01- NO6BA04 during the calendar year 2006, according to this register, was used to create the outcome variable of the study, ADHD-medication.

There were considerable regional differences in the consumption of ADHD-medication (see Table 1). As these differences did not follow any obvious demographic or geographical patterns, we assumed that this pattern mirrored varying prescription patterns in different counties rather than variations in the prevalence of ADHD. The counties were classified into four categories according to the proportion of children having purchased ADHD-medication during 2006: high prescription rates (>0.8%), high average prescription rates (0.7–0.8%), low average prescription rates (0.5–0.6%) and low prescription rates (<0.5%).

Table 1.   Crude rates of ADHD medication in school children by sociodemographic variables
 BoysGirls
Total (N)ADHD (%)Total (N)ADHD (%)
County category
 Cat 164 4701.6561 1230.49
 Cat 2288 9041.11273 5130.32
 Cat 3150 0740.98142 6060.23
 Cat 493 3850.6188 4490.16
Maternal education (years)
 0–950 4491.9747 8470.55
 10–12202 8031.32191 8780.34
 13–14143 6210.92136 4550.27
 15+196 2460.64186 0830.18
 Missing37141.9934280.50
Single parent
 Yes142 6671.72137 1810.47
 No454 1660.85428 5100.23
Social welfare
 Yes31 4043.1829 4300.95
 No565 4290.94536 2610.25
Parental psychiatric or addictive disorder
 Yes35 6102.6133 5370.81
 No561 2230.97532 1540.26
 Total596 8331.06565 6910.29

Statistical analysis

Logistic regression was used to calculate odds ratios (OR) with 95% confidence intervals (CIs) as estimates of effects, with ADHD-medication, defined above, as the outcome variable. We used three models to investigate the effects of three socioeconomic variables (lone parenthood, social welfare and maternal education), on medicated ADHD. In all models, we entered gender, and age in 2006 in three categories (6–9, 10–15, 16–19); in Model 2, we added county of residence categorized according to level of consumption and in Model 3, parental psychiatric/addictive disorder. Interaction analyses were made in a simplified Model 3 where the four category maternal education variable was replaced by the dichotomous ‘high education’ described above.

To estimate possible gender differences in effects, interaction analyses were made for each socioeconomic variable separately in a logistic regression model that included age and county of residence only. To supplement risk comparisons, we calculated attributable fractions [(OR −1/OR)*proportion of exposed cases*100], to estimate the number and percentage of cases that theoretically would not have occurred if the numbers with medicated ADHD had been the same among exposed and non-exposed children and youth. This calculation was based on Model 2 in Table 2 which included all socioeconomic variables.

Table 2.   Logistic regression of ADHD-medication at least once in 2006
 Population share (%)Model 1*
OR (95%CI)
Model 2
OR (95%CI)
Model 3
OR (95%CI)
  1. *Model 1 is adjusted for age and gender.

  2. Model 2 is adjusted for age, gender and county of residence.

  3. Model 3 is adjusted for age, gender, county of residence and paternal addictive or psychiatric disorder.

Maternal education (years)
 0–98.52.30 (2.12–2.48)2.32 (2.15–2.51)2.20 (2.04–2.38)
 10–1234.01.77 (1.66–1.88)1.79 (1.69–1.91)1.74 (1.64–1.85)
 13–1424.11.37 (1.28–1.47)1.38 (1.28–1.47)1.37 (1.28–1.47)
 15+32.9111
 Missing0.62.08 (1.68–2.59)2.08 (1.68–2.59)1.89 (1.52–2.35)
Single parenthood
 Yes24.11.54 (1.47–1.62)1.53 (1.46–1.61)1.45 (1.38–1.52)
 No75.9111
Social welfare
 Yes5.22.35 (2.20–2.51)2.37 (2.22–2.54)2.06 (1.92–2.21)
 No94.8111
Any paternal addictive or psychiatric disorder
 Yes5.91.75 (1.64–1.88)
 No94.11

All statistical analyses were performed in SPSS version 15.0 for Windows (SPSS Inc., Chicago, IL, USA).

Results

There were 7960 children in the study population with a registry entry indicating consumption of ADHD-medication. The most commonly purchased drug was methylphenidate (87.8%), followed by atomoxetine (9.2%) and amphetamine (3%). Figure 1 shows the variation in medicated ADHD by age and gender, demonstrating the highest incidence in boys in the age 10–15 years.

Figure 1.

 Children with ADHD-medication during 2006, by age and gender.

Boys more often consumed ADHD-medication than girls, 1.06% compared with 0.29% in the entire study population. Consumption of ADHD-medication was more common in offspring of mothers with a short education as well as in children in single parent households and households of social welfare recipients in boys as well as girls (Table 1).

Table 2 presents the multivariate analysis. Model 1 that included socioeconomic variables adjusted for age and gender only showed that living in a household that received social welfare was associated with an OR 2.35 and living in a lone parent household with an OR of 1.54. There was a stepwise gradient by level of maternal education with ORs: 2.30, 1.77, 1.37 in comparison with university education. Adding county of residence in Model 2 had no or marginal influence on the effect of the socioeconomic variables. Model 3 demonstrates that adding the indicator of parental psychiatric/addictive disorder attenuated the odds ratio for social welfare to a certain degree, but had a marginal effect on maternal education and lone parenthood.

The gender differences in effects of low education and single parent household on ADHD were marginal with p-values in interaction analyses of 0.94–0.95. There was a slight tendency for a stronger effect in girls of living in a household that receives social welfare (p = 0.21).

Table 3 gives an overview of how the circumstances presented above affected the risk of medication in the offspring, independent of each other. The Table shows how many fewer cases would theoretically have occurred if the risks caused by these circumstances had been eliminated. Low maternal education had the strongest effect, 33.2% of the cases would not have occurred if all children had had the same prevalence as offspring of highly educated mothers while single parent household predicted 13.5% of the medicated cases. For the less frequent indicators, social welfare in the household and psychiatric and/or addictive disorder in parents, the number of avoidable cases was 9.3% and 3.8% respectively.

Table 3.   Consumption of ADHD-medication in school children in different types of household 2005. Number of cases associated with family situation
Type of householdConsumption of ADHD-medication
No. of avoidable casesProportion of all cases that would be avoided (%)
Low maternal education (<15 years)264333.2
Single parenthood107213.5
Social welfare7149.3
Parental psychiatric or addictive disorder3063.8

Interaction analyses demonstrated a significant interaction effect of single parenthood and parental psychiatric or addictive disorder (p < 0.001) on ADHD-medication, while no such effect was identified for maternal education.

Discussion

This study, covering more than one million children in primary and secondary school age, demonstrates that socioeconomic indicators such as education, lone parenthood and reception of social welfare are associated with medicated ADHD in a Swedish context. Low maternal education alone predicted 33% of cases with medicated ADHD, single parenthood 14% and social welfare 10%, while psychiatric or addictive disorder in the parents predicted less than 4%.

The associations found between ADHD-medication and single parenthood as well as low parental education are quite similar to the associations found for hyperkinetic syndrome in the British National Survey of mental health in children 5–16 year old in 2004 (7). In this nationally representative British study, the prevalence of hyperkinetic syndrome was 2.5% with rates being 71% higher in offspring of parents with no educational qualifications and 87% higher in single parent households compared with the rest of the population.

Limitations

To the best of our knowledge, this is the first study of risk factors for ADHD in a national cohort of school children. ADHD is a difficult concept to define in a reliable way in epidemiological studies because of the subjective and context-bound nature of the impairment criteria built into existing diagnostic classifications (8,9). In this study, we used purchased prescriptions of ADHD-medications as an indicator of ADHD. National guidelines for medication of ADHD were issued by the National Board of Health and Welfare in 2002, which stated that medication should be reserved for cases where other supportive interventions have failed. The right to prescribe stimulants for ADHD in Sweden is restricted to specialists with particular familiarity with treatment of this disorder. With this background, and as the Swedish health care system provides free medical and psychiatric care for all children on equal economic premises, we believe that medication is a useful indicator of the more severe cases of ADHD in a Swedish context. Indirect support for this assumption can be cited from the British survey cited above (7) where 95% of the children who fulfilled the criteria for hyperkinetic syndrome had sought professional help during the last 12 months and medication was present in two of five of these children.

The Swedish health care system is administered in 21 separate counties, each county having its own health organization. The existence of a few counties where medication rates are higher, and a few where they are lower, than the average county, indicates that other factors, apart from the severity of the disorder, determine medication to a certain extent. Such factors could be associated not only with geographical differences in availability of services for children with ADHD but also with differences in interpretation and implementation of the national guidelines in different health organizations. The fact that risk estimates were quite marginally affected by the inclusion of the county variable into the multivariate analysis, however, indicates that this variation was not an important source of bias in relation to the socioeconomic variables in the study.

More problematic in this particular study is that risk factors related to the sociodemographic characteristics of the family may influence the detection rate of children with ADHD. Children from exposed families may more often have academic problems, for instance, that may increase the chance of their symptoms being detected in the school setting. It is also possible that these factors may have the opposite effect through delayed help-seeking behaviour that eventually leads to medication. Thus, it is difficult to assess whether unfavourable sociodemographic conditions lead to an under- or overestimation of their true associations with ADHD in the offspring. Further studies that assess consumption of medication in relation to these factors with access to well-defined diagnostic data would be very helpful in this respect.

The indicator for psychiatric and addictive disorder in parents in this study is based on hospital care and thus only captures the most severe cases. It seems most probable that a more sensitive indicator would have predicted a considerable number of cases of ADHD in the offspring.

Gender

This study confirms the pioneering study by Biederman et al. (6) in demonstrating that socioeconomic factors are strong predictors of ADHD. It could not, however, confirm the stronger effect of socioeconomic factors in boys compared with girls; there were no statistically significant gender differences in our study. It is possible that a more nuanced outcome variable, such as the global functioning scales used by Biederman et al. in their study, is needed to identify such gender differences.

Mechanisms

There are several ways in which risk factors in the family environment may be influential in increasing the risk for ADHD in school children. Low parental education is associated with social disadvantage in general, and through this route associated with exposure to a higher number of stressors (10) and a higher risk of childhood adversity (11). Lack of time and money are more common features of everyday life of the single parent, as are lack of social support, family conflict including divorce/separation and parental absence. In a study of more than a million Swedish children and youth, offspring in lone parent families were shown to have increased risks of psychiatric disease, suicide and suicide attempts and addiction (12). Such stressful living conditions may be expressed in altered physiological function, such as dysregulation of the HPA (Hypothalamus, Pituitary, Adrenal cortex) axis (13) and also increase the risk of behavioural and psychiatric symptom formation (10,14). Education and single parenthood also affect the quality of parent-child interactions in a family, with long-term effects on the risk of anti-social behaviour in children and youth (15).

The strong interaction effect of single parenthood and parental psychiatric and addictive disorder demonstrates that socioeconomic and heritable factors are not independent in relation to ADHD-medication. Such interactions may also explain how the high attributable risk of social factors in this study may coexist with the high heritability of ADHD found in twin studies (16). Heritable and environmental factors may interact in many intricate ways. Lasky-Su et al. (5) have demonstrated that environmental factors may enhance biological factors in the development of the inattentive symptoms of ADHD, but parental traits may also act as environmental influences in the development of ADHD symptoms and thereby enhance other environmental factors (4). Of particular relevance for this study is the possibility that effects of environmental factors like maternal education and need for social welfare to some degree may be genetically mediated: they may indicate expressions of heritable personality characteristics of importance for the development of ADHD. They may even be expressions of parental ADHD.

Implications

The identification of environmental risk factors as important predictors for medicated ADHD in this study and a diagnosis of ADHD/hyperactivity in previous studies (6,7), despite a high heritability, opens up a theoretical possibility for prevention. Such prevention could be universal by prevention of adversity as such, or indicated through family-centred interventions for young children at risk to prevent the development of ADHD (17). Preliminary evaluations of a basic parenting skills programme for preschool children with conduct disorder and signs of ADHD suggest that interventions aimed at modifying parent-child interactions might be a successful strategy for secondary prevention.

Conclusions

Further studies of socio-demographic patterns of ADHD-medication in the presence of independent information of ADHD-diagnosis are needed to validate ADHD-medication as an indicator of severe ADHD. Socioeconomic indicators are potent predictors of ADHD-medication in Swedish school children. This suggests that gene-environment interaction models may provide a useful framework for future research into the aetiology of ADHD and subsequent development of preventive interventions.

Ancillary