Prevalence of attention deficit hyperactivity disorder among children and adolescents in Spain: a systematic review and meta-analysis of epidemiological studies

Attention deficit hyperactivity disorder (ADHD) can be defined as a condition starting in childhood, that comprises a persistent pattern of symptoms of hyperactivity, impulsiveness and/or lack of attention, more frequent and severe than usual for that age, and causing a significant impairment in school or work performance and in the activities of daily life. ADHD is a common neuropsychiatric disorder, with a high impact on the health system and the community in terms of economic costs, family stress, academic and vocational adversity and a clear negative effect on the self-esteem of the subject affected [1].

Currently, there exist two diagnostic criteria in regular use to diagnose ADHD in children and adolescents, DSM-IV and ICD-10. Both classifications utilise lists of behaviours to consider in the process of diagnosing hyperactive conditions. The main differences between DSM-IV and ICD-10 pertain to the concomitance of the three domains (inattention, hyperactivity and impulsivity), the exclusion of comorbidity and the degree of pervasiveness. The ICD-10 criteria require a full set of symptoms in all three domains, whereas the DSM-IV recognizes three subtypes of the disorder – the predominantly inattentive type, the predominantly hyperactive-impulsive type and the combined type.

Despite its relevance in terms of public health, the frequency of the disorder is not well established in many countries, including Spain. This information may be necessary to improve the design of future studies on aetiological factors and disease distribution in the population, evaluate the effectiveness and cost-effectiveness of various interventions or programmes and provide representative reference values for evidence-based health services planning. In recent decades, several observational studies have been performed in different population groups and geographic areas. Epidemiological studies in several countries have used questionnaires and scales based on symptoms as a criterion for ADHD. According to previous studies, in Spain the prevalence of ADHD would be 3-14% in children aged 8-15 years in Valencia [2, 3], 4-6% in children aged 6-15 years in Seville [4] and 1% in children aged 6-8 years in Navarre [5]. Therefore, it would be relevant that the data provided in the scientific literature were analysed through integrated approaches which allow for establishing the extent of ADHD and its epidemiological characteristics for the whole children and adolescent population.

In this context, the objective of this study was to perform a systematic review of the studies performed in Spain on the prevalence of ADHD in children and adolescents and combine its results in an overall estimation through meta-analysis techniques.

This study reviewed 14 epidemiological observational studies that analyse the prevalence of ADHD in children and adolescents in Spain, over more than 13,000 people. The main result was the identification of the epidemiological information for providing an estimation of the overall prevalence of disorder in the country. It also provides a greater precision than that resulting of the studies individually considered. Specifically, the results of the meta-analysis show a pooled-prevalence of ADHD of 6.8% in children and adolescents. These values would be generally consistent with those found in the European Union [26, 27]. Wittchen et al. [27] recently estimated that 3.3 million children and adolescents aged 6-17 years have ADHD in the European Union, with a prevalence of 5%. The results of the meta-analysis by Polanczyk et al. [26] yielded a world prevalence equivalent to 5.3%, with similar values for developed regions.

ADHD is a major cause of personal impairment to patients and their families and place a substantial burden on the healthcare services [1]. Given the large amount of literature on ADHD in recent years, healthcare professionals and decision-makers must have up-to-date epidemiological information gathering the best scientific evidence that is useful for service planning and public health policy. Particularly, this information may be necessary to improve the design of future nationwide epidemiological studies, evaluate the impact of interventions or programmes on ADHD. With this regard, the prevalence measures provided can be used to further assess the non-fatal consequences of ADHD integrating this information in summary measures of population health, such as disability-adjusted life years [28]. Furthermore, where scientific evidence exists on the (cost-) effectiveness of particular health services and treatments, our results can also provide data inputs necessary for models of health impact assessments considering the implementation of alternative interventions.

This study has some limitations that must be considered. On the one hand, it shows the generic limitations of any systematic review and meta-analysis, particularly that its quality depends on the studies included and that, although the variability is controlled statistically by random effect models, strata are weighed by their characteristics and heterogeneity is measured, it must be always considered that the meta-analysis is a combination of sometimes disagreeing results. With this regard, the prevalence of ADHD ranged from 1% to 14% in the studies reviewed. Specifically, the study by Farré and Narbona [5] reported the lowest frequencies (1%). These values are probably influenced by the definition of a high critical score as hyperactivity index (> 18 of 30 points in the hyperactivity index of the Conners’ scales). In the same study, the authors recognised that, considering subjects with a critical score in at least one of the two questionnaires, the prevalence of ADHD was 6.4% (2.1% in parents, 3.4% in teachers and 0.8% both). Furthermore, in two studies [14, 17] using a single informer (the teacher) the prevalence rates were higher (12-14%). This is consistent with the information from other authors reporting that using two informers instead of one usually provides a lower prevalence [26]. Along this, previous studies [29, 30] found that teacher and parents agreement on questionnaires is often low, with teachers reporting more symptoms than parents. Both the DSM-IV [31] and the ICD-10 criteria [32] require that the main symptoms (attention deficit, hyperactivity and impulsiveness) occur in more than one setting (e.g., home and school). For example, Rojo et al. [33] analyzed the risk of ADHD characteristics among obese adolescents using a self-administered report of the Strengths and Difficulties Questionnaire (SDQ). As they mentioned, the sensitivity for predicting ADHD conditions was reported to be very low because they were based only on data from one informant (adolescents) and a clinical evaluation of the probable cases was not conducted. Although the aim of that study was not to establish prevalence estimates of ADHD, the authors provided a proportion of ADHD characteristics equals to 21.4% of adolescents aged 13-15 years. It is noteworthy that these results diverge somewhat from those obtained by Rodríguez-Hernández [18] and included in our meta-analysis (3.9% of children aged 7-10 years using parents/teachers’ SDQ).

As with other systematic reviews and meta-analyses [26, 34] a significant heterogeneity was found in the prevalence measures, that were explained incompletely by analysis of subgroups and univariate meta-regressions. In our study, the variables “geographic area” (e.g. South region), “among informers only teachers are included” and/or “one informant” would allow for explaining very partially part of the heterogeneity found. Meta-regression on all epidemiological studies showed only a non statistically linear trend to explain effect variation by year of publication and male-to-female ratio, perhaps because of the limited power of our analysis (e.g. number of observations). These findings also suggest that other unknown factors could be important in accounting for between-study variations. For example, it is noteworthy that none of the studies included in our systematic review has studied the overall prevalence within the national population. The lack of whole population studies has been criticised in the past because selected populations (e.g. subnational/local different samples) and settings (e.g. population-based versus school-based studies) might introduce bias and some degree of uncertainty to the estimates. Previously, the revision by Skounti et al. [34] suggested that the characteristics of the population, the study methods and the diagnostic criterion differences could explain part of the changes in the ADHD prevalence rates.

The existence of nonindexed epidemiological studies in the databases consulted may have involved the loss of some locally relevant studies despite the extensive data searches we made (e.g. PubMed/MEDLINE, IME, IBECS and TESEO). Although an attempt was made to minimize this possible screening bias with specific searches in national databases and thesis dissertations, there may be other studies which have not been identified. However, publication bias is not anticipated (as denoted by funnel plot) because of we obtained a substantial proportion of data from unpublished studies. We also conducted subgroup and sensitivity analyses to assess uncertainty assumptions on the pooled ADHD-prevalence for study characteristics. Such an approach is important in assessing the validity of the assumptions made for the statistical calculations in meta-analyses. Unfortunately information on methodological quality of selected studies was insufficient to allow a detailed analysis of their quality.

The diagnosis of ADHD is complex and should be based on the clinical assessment confirmed by an expert on the recognition and treatment of it. In epidemiological studies there is no agreement about the instruments to be used for evaluating children with potential ADHD; there is also some controversy about the criteria to be used for defining a so-called “case”. These difficulties in the detection, diagnostic process and methods affect the epidemiological studies performed, originating changes that can lead to under- or over-diagnosing ADHD. In our meta-analysis, although based on a small subgroup of studies, the inclusion of those epidemiological studies that were restricted to clinically confirmed ADHD cases (e.g. DSM diagnostic criteria) led to a reduction of the pooled-prevalence from 6.8% (95% CI 4.9 – 8.8%) to 6.4% (95% CI 4.9 – 7.9%), which is even more consistent with the prevalence rates worldwide [26]. Similarly, people who screen negative do not undergo the clinical ascertainment by the specialists, therefore false negatives might have occurred in some epidemiological studies. In this revision, a number of studies did not mention any reference diagnostic criterion [5, 17, 18, 24]. In addition, some used only screening scales with a low sensitivity and specificity and that are not valid as a single measurement for the diagnosis. Actually, DSM-IV and ICD-10 are the diagnostic criteria most commonly used. Both classifications describe the clinical condition of hyperactive children (ADHD/Hyperkinetic disorder) and use similar operative criteria for diagnosing it. However, as the ICD-10 diagnostic criteria are more restrictive, the diagnoses according to this classification will correspond to the most severe cases of ADHD according to DSM-IV criteria. Therefore, the prevalence studies taking ICD-10 as reference will probably yield lower rates than those using DSM-IV. ICD-10 and DSM-IV are also different when considering the subtypes of disorder.

The disagreement between the different studies in terms of case definition criteria involves the need for performing separate analyses for each criteria used for the diagnosis of ADHD. In our revision, no study reported expressly ICD-10 reference diagnoses. There exist evidences [35‐37] that the prevalence of ADHD as defined in the DSM-IV can be somewhat higher than when defined according to DSM-III-R criteria, due to the inclusion of the types “with hyperactive-impulsive predominance” and “with attention deficit predominance” (that had been diagnosed as ADHD not specified in the DSM-III-R) [38]. For the diagnosis of ADHD, the DSM-III-R requires the presence of at least 8 symptoms of a total of 14; it does not include the requirement that they must occur in at least two settings and does not give a division into subtypes, and the severity criterion is based on the number of symptoms. In our subgroup meta-analysis, we confirmed that the use of DSM-IV vs DSM-III-R increases the prevalence rate. However, it is note worthy that the analyses of subgroups involve a loss of statistical power and, therefore, of precision and the unfeasibility to analyse population subgroups that would have been interesting (e.g. analysis by age or social class). As with other revisions [26, 34] in virtually all the studies reviewed, regardless of the methods used, the prevalence rates of ADHD were significantly higher in men than in women. On the contrary, in the study by Cardo et al. [19] ADHD prevalence rates were slightly higher among women. This could be due to the fact that retained students, children with special educational needs and those with some known psychopathological diagnosis were excluded from the study, which would have underestimated prevalence in males.