The ontology of ADHD

Historically, the debate about ADHD has been rather polarized. In psychiatry and the natural sciences, ADHD has been conceptualized as a universally occurring medical condition that exists regardless of societal conditions and cultural settings [7]. Within this paradigm, researchers have sought to investigate neurobiological components associated with possible functional [8, 9], structural [10, 11], and neurotransmitter [12–14] alterations in various regions of the brain. Accordingly, ADHD has been treated as a natural kind, i.e., as a universal entity that exists independently of human systems of classification [15]. On the other hand, sociologists [16] and critical psychiatrists [17] have emphasised how societal and cultural processes have led to an expanded use of the diagnosis, and the scientific integrity of ADHD has also been questioned [18]. Others [19] have nuanced the tension by emphasizing how considering a diagnosis as a societal and cultural process, does not suggest that the diagnosis is not real in a specific context. Furthermore, the controversies around ADHD have been addressed as a problem of reification, i.e., the belief that if something is given a name, we have acquired an explanation [20]. Te Meerman and coworkers thus reminds us that “The descriptive classification Attention-Deficit/Hyperactivity Disorder (ADHD) is often mistaken for a disease entity that explains the causes of inattentive and hyperactive behaviors, rather than merely describing the existence of such behaviors”. Accordingly, there is a risk that not only scientists, but also laymen and politicians tend to regard ADHD as an entity, rather than a mere name for observed behaviors. Such a point of departure does not diminish the suffering and seriousness of the symptoms, but when it comes to scientific understanding, problems of reification can illustrate the fallacies that systematically run through psychiatric thinking. As an example, in an analysis of the increased prescription of pharmacological treatment of ADHD, The Swedish National Board of Health and Welfare [21] stated that the proportion of people receiving a diagnosis eventually will be reflected in how many that actually have ADHD in the population, and when that happens, the curve for new users of medications will flatten out. The fallacy in this sort of thinking is that the diagnosis is conceptualized as a thing one simply has or not. There is a risk that an excessively narrow perspective on epidemiology will end up in a discussion about whether ADHD is over- or underdiagnosed. Such a focus assumes that there exists an objective prevalence, but since most traits and behaviours, including those applied to define ADHD, are continuously distributed within the population, the prevalence is ultimately dependent on social consensus regarding the boundaries of what fits into the diagnosis and what does not. In other words, there is no ‘natural’ prevalence to be found, regardless of scientific rigour. ADHD as a category is so heterogenous, multi-factorial, and even ambiguous [22], that it could implicate an all too broad spectrum of idiopathic symptoms of suffering, depending on the intentions and bias of its advocates.

According to meta-analyses [23, 24], there has been no evidence to suggest an increase in the number of individuals who meet criteria for ADHD in the past three decades. Within the general Swedish population, there are no indications that ‘ADHD traits’ have increased, despite the obvious rise in the number of diagnoses. This discrepancy of course raises questions. Has society become better in discovering more cases of a valid disease entity, among individuals who have previously been undiagnosed? Or has it changed the view on what should be considered a disorder [25, 26]? Psychiatric diagnoses are flexible objects that arise, evolve, and even disappear, in line with historical processes [27, 28]. Something that started out as a label for hyperactive children, nowadays serve as a hypothetical cause for a tremendously broad range of behavioral and cognitive features, in both children and adults: difficulties with concentration, aggression, carelessness, burn-out syndrome, addiction, apathy, impulsivity, antisocial behavior, and so on–basically a majority of psychological attributes associated with impaired functioning. To treat ADHD as a more or less biological disease entity causing an impairment, can be a logical fallacy of reification, prohibiting a more nuanced understanding of human functioning. Therefore, it is important to map the contextual factors that are associated with individuals that become diagnosed. This is where sociological and philosophical perspectives have the potential to enrich the understanding of psychiatric epidemics such as ADHD [16, 27, 29–31].

Ecological niches as a conceptual framework

Rather than seeking to establish the objective prevalence of ADHD, Hornborg and Merlo [25] have advocated that the focus should be to identify the way diagnoses and prevalence are moulded by how societal, economic, and ideological currents change the assumptions, conceptual apparatus and explanatory models of clinicians and researchers. What type of social and cultural forces may have increased the prevalence and in which settings do ADHD occur? One way of understanding social and psychiatric epidemiology on the basis of these issues is by applying the philosophy of Ian Hacking [27]. Instead of perceiving controversial diagnoses as either medically valid or merely socially constructed, Hacking sought to understand the contextual factors that allowed the diagnosis to arise and thrive over time and space. To do this he used the concept of ecological niche, a metaphor derived from biology. Just as biological organisms only will survive in a niche with the right conditions, the same applies to certain types of diseases. For example, the diagnosis of dyslexia would not flourish in a society without a written language. It is possible to argue that the symptoms of ADHD have always existed, but the idea of an ecological niche helps to illuminate in which conditions they become classified as a medical diagnosis. Hacking’s metaphor is epidemiologically relevant since it acknowledges both a disorder (analogous to the organism) and its context (defined by the factors that forms the ecological niche in which the organism thrives), and it can be particularly useful in social epidemiology. In drastic epidemiological changes, it is of great importance for research to study the contextual forces that underly the act of diagnosing. This can be done by investigating vehicles on a macro level, such as the influence of the pharmaceutical industry, an increased biological focus within psychiatry, transitions from ICD to DSM and the role of the Internet [32], i.e., factors that affect the entire way research is conceptualized, as well as the general perception of laymen and media. But it can also be done by examining how different micro niches, domains within society, is affected differently by the changing perceptions of what constitutes a disorder, acknowledging how the act of turning human behaviour into disease always takes place within a particular context. If the national ADHD prevalence has a considerable heterogeneity in the distribution of diagnoses, this could be framed as an existence of smaller ecological niches in which demographical and socioeconomic factors condition the setting where diagnoses occur. This approach is analogous to the traditional study of inequalities in health where socioeconomic differences are investigated with an intersectional approach [33].

Intersectional analysis of individual heterogeneity and discriminatory accuracy (AIHDA)

Intersectionality was originally a theory that mainly applied to qualitative methodology. However, an intersectional quantitative approach is being increasingly utilized in epidemiology and public health [33–37] as it provides an ideal framework for understanding the demographic and socioeconomic heterogeneity in diagnoses existing within a community. The Oxford English Dictionary 2015 defines intersectionality as “The interconnected nature of social categorizations such as race, class, and gender, regarded as creating overlapping and interdependent systems of discrimination or disadvantage”. Thereby, the demographic, and socioeconomic dimensions that condition health and disease risk need to be considered as interlocked rather than as unidimensional. By using intersectionality, numerous intersectional groups (i.e., strata) can be defined by a combination of several demographic and socioeconomic dimensions (e.g., age, gender, income, and country of birth). In this way, the intersectional perspective provides an improved mapping of disadvantage that better illustrates the heterogeneous socioeconomic distribution of health in the community. Such intersectional strata could be conceptualized as micro niches that facilitate or hinder the construction of ADHD diagnoses by providing different thresholds for dichotomizing traits and behaviours. For instance, many of the questions included in self-assessment forms for ADHD such as ASRS-v1.1 [38] can be interpreted differently in different intersectional contexts. Also, schoolteachers and legal guardians from different niches may have different expectations that condition the evaluation of the symptomatology [39–41]. The influence of intersectional niches could be perceived as an addition of influences from separate social dimensions or as an interaction where the propensity of receiving an ADHD diagnosis is larger or lower than the sum of the influences of separate social dimensions.

The present study develops an intersectional perspective into an intersectional ecological niche (IEN) framework. The IEN framework benefits from intersectionality by allowing an improved use of interlocked socioeconomic dimensions, but simultaneously allows the considerations of whatever other relevant ecological factors that would not traditionally be included in a formal intersectional analysis. Thereby, the IE framework gains flexibility when it comes to investigating overlapping and interdependent systems of discrimination or disadvantage. Utilizing the concept of ecological niches is a way of counteracting essentialist narratives and developing a more constructionist framework on how disorders such as ADHD flourish due to contextual factors. Instead of conceptualizing ADHD as a universally occurring entity, it is approached as something that emerges more or less in a given context. Accordingly, the IEN-AIHDA framework that we propose may be a way of addressing the emergence of mental illnesses both in societies and in individuals [cf. 42].

By applying a methodology denominated multilevel analysis of individual heterogeneity and discriminatory accuracy (MAIHDA) [36], the individual is considered as nested within intersectional ecological niches. Such a methodology can be applied by using both formal multilevel regression [43, 44] as well as traditional regression analyses [33]. MAIHDA provides a detailed mapping of the risk of ADHD diagnosis in the population and allows an evaluation to which degree micro niches are with accuracy able to discriminate individuals with ADHD from individuals without ADHD [36, 45]. Traditional methodologies assign the same average value (i.e., probability) to all the members in the group (i.e., intersectional niche) without considering the individual heterogeneity around the group average and the overlapping between categories [46]. If the discriminatory accuracy is low, the benefit of focusing on specific intersectional niches for understanding differences in ADHD diagnosis can be discussed.

Aim

The aim of this study is twofold. Firstly, we want to obtain a better understanding of the demographic and socioeconomic distribution of ADHD diagnosis in the Swedish population. We hypothesised that different intersectional niches (i.e., contexts, strata) may promote or hinder the risk of getting an ADHD diagnose in Sweden. That is, the same individual would receive or not receive the diagnosis if he/she was exposed to a different niche. If this is true, differences between the intersectional niches would be high. Secondly, we want to contribute to the ongoing critical discussion on medicalization, i.e., the tendency to increasingly define and treat behavioural and social problems as medical entities. Our analysis of intersectional niches is therefore grounded in a nominalist point of departure, highlighting the importance of critically examining current discourses on disease categories such as ADHD.

• How is the prevalence of ADHD diagnoses distributed across intersectional niches defined by gender, age, country of birth, and income?
• What is the discriminatory accuracy of the intersectional niches for classifying individuals regarding the existence of an ADHD diagnosis?
• In what way can the findings contribute to the ongoing discussion about ADHD as a historically situated diagnostic trend of epidemic proportions?

Databases

We used a database composed of three linked registers covering the whole country of Sweden: The Register of the Total Population (TPR) [47], the Longitudinal Integration Database for Health Insurance and Labour Market Studies (LISA) administered by Statistics Sweden [48], and The National Patient Register (NPR) [49] administered by the National Board of Health and Welfare [50]. We performed a record linkage using the unique 12-digit social security number (SSN) assigned to each individual residing in Sweden. In order to pseudonymize the research database, Statistics Sweden in coordination with The National Board of Health and Welfare replaced the SSN with an arbitrary serial number before delivering files. The information is registered without formal individual consent. However, while the data is confidential, the Swedish authorities delivers pseudonymize information for research purposes after evaluation.

The TPR [47] registers data on all Swedish residents regarding socioeconomic information about gender, civil status, migration, country of birth and family relationship as well as general demographics for those who stayed in Sweden for at least 12 months. The LISA database is administered by Statistics Sweden [48] and combines information on all Swedish residents above the age of 15 from several registers, including the labour market, social sectors, and educational sectors. It includes status on living conditions, demographics, and socioeconomic factors, like income, education, and employment. The NPR [50, 51] records more than 99% of inpatient and approximately 80% of specialized outpatient diagnoses from hospitals in Sweden. At the time of the study, all diagnoses were coded according to the International Classification of Diseases, 10th version (ICD-10). The NPR collects data from both private and public health care providers with an 85–95% positive predictive value on inpatient diagnosis [49]. Data from primary care is not included in the NPR.

Study population

According to the TPR, 9.420.128 individuals resided in Sweden at the baseline date of December 31^st, 2010 (Fig 3). When comparing data with the Death Register, we excluded 5.630 individuals that had passed away but with delayed registration, making the correct number of individuals 9.414.498. Based on the TPR, 444.596 individuals that had stayed in Sweden for less than 5 years before baseline were excluded, since recently arrived immigrants might have incomplete register information. Finally, we excluded 2.882.109 individuals older than 60 or younger than 5 years, leaving the database containing information on 6.093.423 individuals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Flow chart informing on inclusion and exclusion criteria, as well as number of individuals included in the study population.

https://doi.org/10.1371/journal.pone.0294741.g003

Assessment of the outcome variable

The outcome variables used were presence or absence of ADHD diagnosis between 2005 and 2010. We defined ADHD diagnosis as any variant of International Classification of Diseases, 10^th version (ICD-10) codes F90.0-F90.8.

Assessment of demographic and socioeconomic variables

Gender was binary coded as either man or woman according to legal status. We categorized age at the baseline date into eight intervals: 5–9, 10–14, 15–19, 20–24, 25–29, 30–39, 40–49 and 50–59 years of age. It is known that an ADHD diagnosis is more common in boys than girls [52–57] and studies have also concluded that it is more common amongst 10–14-year-old children [58, 59]. Furthermore, we used information from Statistics Sweden to classify country of birth into a binary variable. We labelled individuals born in Sweden as ‘natives’ and individuals born outside of Sweden as ‘immigrants’.

In order to achieve accurate measures of the individuals’ income, the income variable was created using the individualized cumulative disposable income of the household. The disposable income is the income that remains after taxes has been accounted for. We computed a variable for the years 2000, 2005, and 2010, and divided the total disposable income of a family by the number of individuals in that family. The different consumption weights of adults and children of different ages were used according to a specific formula [60]. We sorted groups by 25^th quantiles using the TPR in the years 2000, 2005, and 2010. Thereafter, were added the scores from the three years ranging from 1 to 25 and obtained values from three (lowest cumulative income) to 75 (highest cumulative income). Finally, we categorized the cumulative income into three groups (high, middle, and low) by tertiles. The 1002 individuals with missing values of income during 2000 or 2005 were assigned the tertile values of the year 2010. No individuals in our study population had missing income data in that particular year.

Assessment of intersectional niches

As a way of including the intersectional niches, we created a multi-categorical variable consisting of 96 strata. The strata were created by combining the two categories of gender, the eight categories of age, the two categories of country of birth and the three categories of cumulative income. We used a reference stratum constructed of 10-14-year-old women born in Sweden with high income for comparison. This choice was somewhat arbitrary but based on previous studies indicating the characteristics of the individuals with the lowest ADHD risk.

Statistical analysis

The first step in our analysis was to map the prevalence (i.e., absolute risk) of ADHD across the intersectional niches by graphical stratified analyses. Thereafter, we modelled the binary ADHD variable as the dependent variable. We applied Cox proportional hazard regressions with a constant follow-up time equal to one. In this way, we formally attained prevalence ratios (PR) or relative risks [61].

We performed six different models. Model 1 included only age, model 2 only gender, model 3 only income, and model 4 only country of birth. In model 5 all four single variables were entered simultaneously. Finally, in model 6, we included the same information as in model 5 but with the multi-categorical variable with the stratum of 10–14-year-old native women with high income as reference. In all models we calculated PR with 99% instead of 95% confidence intervals (CI) to counteract the multiple comparisons problem.

For each model, we quantified its discriminatory accuracy (DA) by means of the area under the receiver operator characteristics curve (AUC) [45]. The AUC is calculated by plotting the true positive fraction (i.e., sensitivity) against the false positive fraction (i.e., 1 –specificity) for different binary classification thresholds of the predicted probability of ADHD diagnosis. Thus, the AUC measures the accuracy of the information provided by the variables in the model used to discriminate individuals with ADHD from those without ADHD. The AUC obtains a value between 0.5 and 1, where 1 indicates perfect discrimination and 0.5 means that the studied variables have no discriminatory accuracy at all. Rather than only evaluating the differences between absolute risks of strata values, the DA also assessed the overlapping of the individual risk predictions (based on the intersectional niches) between individuals with and without ADHD. There is no fully established practical guideline for the interpretation of the size of the AUC as a measure of discriminatory accuracy when analysing IEN inequalities. However, based on the classification provided by Hosmer and Lemeshow [62] we defined the discriminatory accuracy as (i) ‘absent or very low’ (AUC = 0.5–0.6) corresponding with ‘absent or very low’ strata inequalities, (ii)‘poor’ (AUC >0.6–≤ 0.7), corresponding with ‘small’ strata inequalities, (iii) ‘acceptable’ (AUC >0.7– ≤ 0.8), corresponding with ‘large’ strata inequalities, (iv) ‘excellent’ (AUC >0.8–≤ 0.9) or (v) ‘outstanding’ (AUC > 0.9–1) corresponding with ‘very large’ strata inequalities.

We further calculated the gradual change in the AUC value (Δ-AUC) between the models. The Δ-AUC quantifies the improvement in the discriminatory accuracy obtained by a model, in relation to a reference model [63]. The categorical intersectional variable in model 6 allows us to capture possible interaction of effects between the variables that define the strata so if any interaction exists, the discriminatory accuracy of model 6 in comparison with model 5 will increase and the Δ-AUC be positive [33].

All statistical analyses were performed using IBM SPSS (Statistical Package for the Social Sciences) version 24 and Stata v15 (StataCorp, College Station, TX).