Introduction

Psychotropic medications are among the most prescribed medications in Europe [1]. In recent decades, there has been a general increase in their use in Europe [2]. Also in Belgium, use rates have risen over time [3]. Latest figures show that about 21.5% of the Belgian population in 2019 used psychotropic medications [4], making Belgium one of the leading countries within Europe as concerns psychotropic medication use.

To date, several studies have shown socioeconomic inequalities in the use of psychotropic medication, with people with a lower socioeconomic background having higher uses rates compared to people with a higher socioeconomic background. The majority of these studies are inspired by Andersen’s [5] behavioural model of health care use and inequalities therein and focus on horizontal equity [6], based on the principle of equal access for equal need [7, 8]. In his original model, Andersen differentiates between predisposing (e.g., education, gender, age) and enabling factors (consisting of community and family enabling resources, e.g., health insurance, social support), which both determine perceptions of need and subsequent health care use.

Despite its widespread use, Andersen’s model of health care use is often criticised for its individualistic orientation, mainly because of its strong focus on individuals’ needs [9, 10]. The initial model has, therefore, been refined over the years, with the current modified model paying more attention to contextual factors, such as the organisation of the health care system. Nevertheless, health care use is still seen as an individual choice and responsibility [10,11,12], with need having a determining role. Little attention is paid to interactive processes that can influence health care use, recognising the context dependence of individuals, and insufficient attention is paid to health care use that is not based on need.

Medicalisation, however, is one example of an interactive process, focussing on the social context, that impacts both need and non-need determinants of health care use. Conrad [13] depicts medicalisation as a social process in which normal biological processes or behaviours are increasingly described, accepted and treated as medical problems, leading to increased attention for, and growing consumption of medical treatments [14, 15]. Medicalisation, as a critical theory, also points to the presence of inequalities in the process of medicalisation [16]. It recognises that medicalisation is a heterogenous process that affects socioeconomic groups in different ways. This is referred to as stratified medicalisation.

Furthermore, medicalisation interacts with what is declared and experienced as ‘need for care’, since tolerance towards health-related problems and discomfort has lowered [14, 17]. Increased feelings and observations of need have an impact on health care use. But medicalisation processes also interact with non-need determinants that affect health care use. For example, the interaction with institutional factors.

To elaborate, two important institutional factors, that might interact with the medicalisation of mental health symptoms, are discussed. Both factors are typical of the Belgian case. First, the organisation of the Belgian health care system in which general practitioners (GPs) play an important key position. For instance, GPs prescribe the most of the psychotropic medication in Belgium [18, 19], and figures show that there is a pro-low socioeconomic status (SES) bias in GP contact in Belgium [3], which means that people with a lower SES have higher chances of being prescribed medication. In addition, there are also differences in GP–patient interaction, with patients with a lower SES being more likely to be prescribed medical treatment, because GPs perceive those patients as lacking the financial but also social and personal resources to handle a more ‘active’ treatments [20, 21].

The second important factor relates to the reimbursement structure of the Belgian health care system. In Belgium, psychopharmaceutical treatments are reimbursed on prescription, but active treatments, such as psychotherapy, are only reimbursed for a small proportion and a limited number of sessions [22]. The extensive reimbursement of psychotropic medication makes them a straightforward prescribing option, especially since GPs often consider the socioeconomic context of their patients when prescribing [23]. Prescribing usually happens without first having psychological counselling about the underlying causes of the complaints.

Both factors lead to the hypothesis that the institutional context may affect the medicalisation of mental health symptoms (due to the great accessibility of GPs and the comprehensive reimbursement of psychotropic medications) and more important inequalities therein. Therefore, this study will examine if psychotropic medication use is stratified, and whether this is not simply the results of differences in need for care but also influenced by factors not based on need, initiating the stratified medicalisation of mental health symptoms. Medicalisation is thus interpreted and measured as psychotropic medication use beyond actual ‘need’. Moreover, if psychotropic medication use differs between socioeconomic groups, and it can only be partly ascribed to differences in need, whilst also controlling for enabling determinants of relevance, evidence is proven for the stratified medicalisation of mental health symptoms. However, as mentioned earlier, we do recognise that medicalisation processes also interact with what is experienced and declared as ‘need’. For example, treatment may affect problem definition and, therefore, assessed need. As such, we do recognise that ‘need’ is only partially addressed. Since (stratified) medicalisation is an ongoing process, time trends are also incorporated, analysing inequalities in use over time.

Methods

Sample

Data are obtained from the Belgian Health Interview Survey (BHIS). The BHIS is a repeated cross-sectional survey coordinated by Sciensano, the Scientific Institute of Public Health of the federal Belgian State. This study covers information from five successive waves: 2001, 2004, 2008, 2013, 2018. Households and their members are selected from the National Register following a multi-stage stratified sampling procedure. Information is collected through face-to-face interviews, as well as through a self-administered questionnaire. This study includes 31,493 Belgian respondents aged between 25 and 85 years old. Information on data-cleaning is available in appendix (see Appendix 1).

Variables

Dependent variable

The question used to operationalise psychotropic medication use gauges the use of prescribed psychotropic medication in the past 2 weeks, where psychotropic medication consists of antidepressants, sleeping medication and tranquillizers. The variable is recoded as dichotomous (0 = no).

Independent variable

Education is used as independent variable. Education has become the most commonly used indicator of SES in health research for several reasons [24, 25]; its stable, less subject to reverse causality and acquired first over the life course. Education is measured as the highest level of education completed and is recoded into three categories according to the International Standard Classification of Education of 1997: 0 = shorter education (pre-primary or primary education), 1 = moderate education (lower and upper secondary education), and 2 = longer education (post-secondary or tertiary education) [ref.cat].

Mediating variables

Mediating variables are grouped into ‘enabling’ and ‘need’ determinants, where enabling determinants—encompassing community and family enabling resources—consist of GP contact in the past twelve months (0 = yes), regular GP (0 = yes), frequency of social contact (0 = less than once a week) and household composition (0 = single [ref.cat], 1 = partner, 2 = other). Need determinants are the mental health scale, measured by respondents’ score on the GHQ-12 (Cronbach’s alpha = 0.88). Likert operationalisation is used, with higher scores indicating lower mental health. The GHQ-12 is a standard and extensively used measure of general well-being [26, 27]. Also, a variable of whether respondents suffer from a chronic disease and/or long-term illness is added (0 = no).

Control variables

To control for associations with psychotropic medication use, following socio-demographic control variables are added: gender, nationality, region, urbanisation and wave. Gender is included as a dichotomous variable (0 = man), together with nationality (0 = Belgian). Region is recoded into three categories: 0 = Flanders [ref.cat], 1 = Brussels, 2 = Wallonia, as well as urbanisation: 0 = cities-agglomerates [ref.cat], 1 = urban-suburban, 2 = rural. Wave is incorporated as a categorical variable: 0 = 2001 [ref.cat], 1 = 2004, 2 = 2008, 3 = 2013, 4 = 2018.

Statistical procedure

Prevalence rates of psychotropic medication use are reported using weighted proportions and are stratified by wave. After carrying out bivariate statistics, Poisson regression models are tested by estimating adjusted prevalence ratios (APRs) and their corresponding P values. Robust variance estimators are used to ensure that errors are not heteroskedastic. Since psychotropic medication use is highly dependent on age, analyses are run separately for three different age categories (25–44 years, 45–64 years, 65–85 years). However, analyses with the three age categories together are available in Table 3 (see Appendix 2). Models are built stepwise; where the first model contains the dependent variable education and the control variables; in the second model, the enabling determinants are added; and in the third model, the need determinants are included. Each model is plotted to detect evolutions over time, using marginal means postestimation. Analyses are weighted for survey sampling and non-participation bias and conducted with SPSS 22 and STATA 15.

Results

Bivariate results (see Table 1) show that overall psychotropic medication use is lower among those with the longest education, compared to those with moderate or shorter education. This educational gradient is persistent over time. Furthermore, psychotropic medication use strongly increases with age. Although time trends partially differ between age categories, we notice an overall increase of psychotropic medication use during the observed period, with a hint towards a decrease in the latest wave. In addition, educational inequalities in use are more pronounced in the youngest age category (odds shorter vs. longer education: 13.1/7.1 = 1.85), compared to the middle age category (odds shorter vs. longer education: 23.7/16.3 = 1.45) and oldest age category (odds shorter vs. longer education: 34.4/24.0 = 1.43). Nevertheless, in all age categories, a clear educational gradient is observed.

Table 1 Weighted proportionsa of psychotropic medication use in the past 2 weeks among the different age categories, according to education
Full size table

The Poisson regression results (see Table 2) reveal several common trends visible in all three age categories. To begin with, and in line with the descriptive statistics, those with the longest education report lower levels of psychotropic medication use compared to those with moderate or shorter education. Next, women and people with a Belgian nationality report higher use rates compared to men and people with a non-Belgian nationality. Furthermore, regional differences show that in Brussels and Wallonia use rates are higher compared to Flanders. Concerning the enabling determinants, the results show that having contact with a GP relates to more use. On the other hand, having social contact and living with a partner or having another household composition apart from being single, relates to less use. Regarding the need determinants, the results show the importance of mental health in relation to psychotropic medication use, where people with poorer mental health status report significantly higher levels of use. Also, having a chronic condition and/or longstanding illness highly impacts the use of psychotropic medication.

Table 2 Weighted APRs and corresponding P-values for psychotropic medication use in the past 2 weeks among the different age categories, according to characteristics of relevance
Full size table

As for the age-specific Poisson regression results; in the youngest age category, educational differences are no longer significant after entering the need determinants (model 3a), which makes clear that educational inequalities in psychotropic medication use are mostly related to the shorter educated having a worse mental health, a longstanding illness and/or chronic condition. In the middle and oldest age category, educational inequalities remain significant even after entering the enabling and need determinants (model 3b and model 3c). Also, in the middle age category, a significant increase in use over time becomes visible in models 1b and 2b, whereas this is not the case in the other age categories.

The plotted graphs of the Poisson regressions (see Fig. 1), displaying trends over time, also reveal some common and age-specific results. In general, the graphs show that the education gradient is clearly visible and persistent over time (see Fig. 1, models 1a, 1b and 1c). After entering the enabling determinants, educational inequalities diminish but remain evident (see Fig. 1, models 2a, 2b, 2c). When the need determinants are inserted, educational inequalities strongly decrease (see Fig. 1, models 3a, 3b, 3c), indicating the importance of (mental) health in relation to psychotropic medication use. Besides, the effect of entering the enabling and need determinants remains stable over time. Furthermore, the graphs show that psychotropic medication use steadily increases by age, with the oldest age category having visibly higher uses rates compared to the other age categories. In addition, the graphs reveal an overall increase in use, but that increase flattens and is even replaced by a decrease when the need determinants are entered. The decrease in use is most visible in the youngest and middle age category, starting from 2008. In the oldest age category, time trends are less clear, however, the level of use and inequalities in use are most prominent here.

Fig. 1
figure 1

Psychotropic medication use in the past 2 weeks according to education, weighted prevalence (min.–max.: 0–1) among the different age categories for each tested Poisson regression model (see Table 2). Models 1: APRs adjusted for education and socio-demographic control variables; Models 2: APRs adjusted for education, socio-demographic control variables and enabling determinants; Models 3: APRs adjusted for education, socio-demographic control variables, enabling and need determinants

Full size image

Discussion

Before discussing the main findings, some limitations of this study should be addressed. First, the dichotomous character of the dependent variable ‘prescribed medication use’; the variable does not provide information about duration and amount of use. Information on both indicators would provide a more accurate operationalisation. Also, the variable does not offer information about non-prescribed use. The second limitation concerns the measurement of ‘need’; information about mental health is self-reported. The use of self-reported health has been subject of controversy, as its origins lie in the perceptual framework of the individual respondent [28]. Nevertheless, many studies support the validity and reliability of self-reported health [29, 30]. Also, no information was provided on the intensity or possible comorbidity of chronic conditions and/or long-term illnesses. Last, the time period between the scale that assesses mental health and the dependent variable psychotropic medication use is rather limited. The mental health scale gauges mental health complaints in the last month, while the dependent variable asks for psychotropic medication use in the past two weeks.

Despite these limitations, our study provides evidence that educational inequalities in psychotropic medication use are significant and persistently visible over time. In the middle and oldest age category, educational inequalities remain significant even after entering need for care. Moreover, as concerns our operationalisation of medicalisation, evidence is proven for the stratified medicalisation mental health symptoms. However, entering the need determinants lead to a strong reduction in inequalities in use, showing the importance of differences in need in relation to inequalities in psychotropic medication use.

Second, after entering the need determinants, the visible increase in psychotropic medication use over time is being flattened and even replaced by a decrease in use over time. The initial increase in use might thus be explained by an increase in need for care. However, medicalisation processes also impact feelings and observations of need, since medicalisation also happens at the interactional level where complaints are increasingly defined as medical [13, 14], indirectly influencing health care use. In consequence, ‘medicalisation of needs’ may also be involved here. Despite that, regarding absolute numbers, psychotropic medication use declined in the last wave.

Third, this study reveals the importance of age-specific analyses. Age differences in psychotropic medication use are clearly observable, where use rates in the oldest age category are rather high comparing to the other age categories. This finding is in accordance with existing research from Belgium [31], as well as from other European countries [32,33,34]. Even with need for care incorporated, use rates remain remarkably high in the oldest age category. The worryingly high level of use by older people is also supported by several Belgian researchers and experts stressing the misuse and overuse of psychotropics in that age category [35, 36].

Fourth, the results reveal consistent regional differences in psychotropic medication use, with Brussels and Wallonia having higher levels of use compared to Flanders, even when controlling for socio-demographic and composition determinants of relevance. This might partly be the result of regional differences in institutional arrangements, such as the organisation of mental health care, which is a regionalised policy domain in Belgium [37].

Fifth, GP contact significantly relates to psychotropic medication use. In line with previous Belgian research [18, 19], the influential role of GPs in prescribing psychotropic medication is again confirmed. Also in other European countries, GPs prescribing behaviour receives attention [38,39,40,41,42]. In Belgium, health experts and policy makers have joined forces to address GPs prescribing behaviour regarding psychotropic medication. For example, Belgian GPs repeatedly receive voluminous reports with individual scores on different consciously chosen domains, where psychotropic medication prescribing is often one of them [43]. The scores allow them to compare their own prescribing behaviour with that of their colleagues, with the aim of bringing GPs to repentance where necessary.

To conclude, our study shows that psychotropic medication use is indeed stratified, and that this is not simply the result of differences in need, but also influenced by factors linked to the stratified medicalisation of mental health symptoms. Therefore, research should move beyond analysing inequalities in psychotropic medication use merely in relation to differences in individual needs. Future studies need to reflect upon non-need determinants that can impact psychotropic medication use too, as this study tried to stress the importance of the institutional context. Incorporating a multilevel approach in future studies, stressing the institutional context in relation to the stratified medicalisation of mental health symptoms use might be useful [44]. However, research also needs to recognise that ‘need’ in itself can also be medicalised. This study only partially addresses medicalisation, since all needs are considered as ‘real complaints’, and medicalisation processes only take place irrespective of ‘needs’. As such, medicalisation is not totally measured; For example, if someone feels tired all the time, it may be thought of as a physical health problem, but if a GP prescribes antidepressants as response, the individual might come to define the problem as depression. Future research should incorporate a measure of medicalisation which recognizes that need for care may also be medicalised and that it may also be subject to inequalities.

Another suggestion for future research is to focus on the recent downward trend in absolute numbers of psychotropic medication use. In recent years, more attention is being paid to the side effects of psychotropic medication [33, 45, 46]. Medical treatments are, therefore, less often chosen as ideal treatment option, where alternative treatment options, such as psychotherapy, are promoted. The decline in use could indicate that this attention is effective. Psychotherapy is also becoming institutionally stimulated in Belgium. For example, recently, a new policy initiative came into practice in which the reimbursement of psychotherapy is being raised, together with the number of sessions that is being reimbursed [22]. The aim is to make psychotherapy more accessible, because currently individuals face significant financial barriers to accessing it. Future research might thus also incorporate psychotherapy in its analyses, by investigating how psychotherapy use evolves over time and how this interacts with psychotropic medication use, and more particular how inequalities in both processes relate to each other and evolve over time.