Please see attachment for discussion directions and article.
No run on sentence and bad grammar
Must complete on time will dispute.
Please Read All Directions:
1. Briefly summarize what the article is about (2 to 3 sentences)
2. What survey did the data come from? Do an internet search for the survey name and then write 2 to 3 sentences summarizing what it is.
3. List the measures (variables) the researchers used and briefly describe them.
4. What are the researchers' three main findings?
Please write your answers entirely in your own words (no quotes) and use as plain of language as possible. In other words, try to avoid academic jargon and acronyms and describe things using terms that anyone could understand. There will be some things that most likely will not make sense to you (like "poststratification weights," for example) and you do not need to worry about trying to explain what they mean.
Journal of Health and Social Behavior 2017, Vol. 58(3) 272 –290 © American Sociological Association 2017 DOI: 10.1177/0022146517716232 jhsb.sagepub.com
As a recent meta-analysis of cross-sectional and longitudinal studies showed, the negative relation- ship between unemployment and mental health is well established (Paul and Moser 2009), but much less is known about how unemployment translates into mental healthcare (MHC) utilization. The few studies that have explored this relationship used the consumption of health services or medication as a proxy for mental health problems (Schmitz 2011; Virtanen et al. 2008). This is a significant limitation, given that MHC and antidepressant use among the unemployed are not exclusively need based (Buffel, Dereuddre, and Bracke 2015; Buffel, van de Straat, and Bracke 2015). Previous research confirms that the unemployed have higher MHC and medication use than expected based on their mental health sta- tus, which indicates the medicalization of unem- ployment (Buffel, Dereuddre, et al. 2015; Buffel, van de Straat, et al. 2015).
An even more striking limitation of existing research into unemployment, health, and MHC uti- lization is the lack of cross-national comparative research (Bambra and Eikemo 2009). It is crucial to understand whether, how, and why unemployment drives cross-national differences in MHC utiliza- tion given the current context of (a) high unemploy- ment rates and healthcare expenditures in many wealthy democracies and (b) austerity policy implementation in many European countries that has led to public expenditure cutbacks.
716232HSBXXX10.1177/0022146517716232Journal of Health and Social BehaviorBuffel et al. research-article2017
1Ghent University, Ghent, Belgium 2Harvard University, Cambridge, MA, USA
Corresponding Author: Veerle Buffel, Department of Sociology, Ghent University, Korte Meer 5. 9000, Ghent, Belgium. E-mail: [email protected]
The Institutional Foundations of Medicalization: A Cross- national Analysis of Mental Health and Unemployment
Veerle Buffel1, Jason Beckfield2, and Piet Bracke1
Abstract In this study, we question (1) whether the relationship between unemployment and mental healthcare use, controlling for mental health status, varies across European countries and (2) whether these differences are patterned by a combination of unemployment and healthcare generosity. We hypothesize that medicalization of unemployment is stronger in countries where a low level of unemployment generosity is combined with a high level of healthcare generosity. A subsample of 36,306 working-age respondents from rounds 64.4 (2005–2006) and 73.2 (2010) of the cross-national survey Eurobarometer was used. Country-specific logistic regression and multilevel analyses, controlling for public disability spending, changes in government spending, economic capacity, and unemployment rate, were performed. We find that unemployment is medicalized, at least to some degree, in the majority of the 24 nations surveyed. Moreover, the medicalization of unemployment varies substantially across countries, corresponding to the combination of the level of unemployment and of healthcare generosity.
Keywords employment status, healthcare generosity, medicalization, mental healthcare use, unemployment generosity
Buffel et al. 273
In this study, we investigate first whether the rela- tionship between unemployment and MHC use var- ies across European countries. Second, we explore whether these differences are patterned by a combi- nation of unemployment policies and healthcare characteristics, including disability benefits. Third, we analyze how levels of generosity in both policy domains shape the relationship between unemploy- ment and MHC use. Using data from before and after the start of the recent economic recession, which sparked austerity policies in many countries, allows us to shed light on the role austerity politics play in connecting unemployment and mental health (Beatty and Fothergill 2015).
BACKGrOUnD Medicalizing Unemployment and the Gaps in Current Medicalization Research Medicalization describes a process by which nonmedical (social) problems—such as unemploy- ment—are defined and/or treated as medical prob- lems (Conrad 1992). Hitherto, the lion’s share of medicalization research has taken a social construc- tivist approach, focusing on the construction of new medical categories and the subsequent expansion of medical jurisdiction (such as with hyperactivity, meno- pause, and alcoholism; Conrad 2005). Such work has developed an extensive conceptual and empirical lit- erature on the process and effects of medicalization, yet the range of institutional contexts considered is rela- tively limited. This limitation has truncated the range of institutional factors that have been considered as poten- tial foundations for medicalization.
Our theoretical contribution focuses on institu- tional foundations for medicalization, drawing insights from comparative political economy research. We thus address several limitations of existing research that has drawn on Conrad’s (1992) medicalization theory. First, we develop a novel technique for measuring medicalization, building on related cross-national comparative work on medicalization (Christiaens and Bracke 2014; Olafsdottir 2007). Our approach measures and interprets MHC utilization beyond its actual need, as an indicator of medicalization that can be com- pared across societies. In other words, if MHC is used more by the unemployed than the employed, and can only partly be ascribed to poorer mental health status, this indicates the medicalization of unemployment.
Medicalization of unemployment will be explored in one specific domain: the use of medical
care in the mental health field. We recognize that the medicalization of unemployment may take many manifestations, such as in national discourses on unemployment as a personal failure. Unemploy- ment may also be medicalized to allow reliance on disability benefits, which are more stable, less stig- matizing, and often more generous than unemploy- ment benefits (Beatty and Fothergill 2015). Labor market inactivity of those who rely on disability benefits for income support is sometimes known as “hidden unemployment”; arguments in the litera- ture over the boundary between the inactive and the disabled prevent precise identification of hidden- unemployment effects (Bratsberg, Fevang, and Roed 2010). Our analysis addresses this by (1) excluding respondents who report being out of the labor market due to disability and (2) controlling for public expenditures on disability benefits.
Second, we contribute to the developing cross- national comparative approach to understanding medicalization (Olafsdottir 2007, 2010). Our inno- vation, in view of the trend toward deinstitutional- ization in the European mental health sector (Hermans, de Witte, and Dom 2012), is to relax the assumption that physicians and hospitals are the (only) key actors in medicalization and recognize that multiple power actors—the pharmaceutical industry, policy makers, and patients or healthcare consumers—also contribute to the process (Clarke and Shim 2011; Conrad 2005).
Third, we advance the integration of medical- ization research across levels of analysis (Conrad 2005). We apply multilevel modeling to evaluate the hypothesis that medicalization, as a cultural transformation that varies across institutional set- tings, shapes the health behavior of individuals, such as consulting medical professionals.
In addition, multilevel analysis of data collected before and after the onset of economic recession and austerity policies allows us to address the rela- tionship between austerity and the medicalization of unemployment. Previous studies (Antonakakis and Collins 2015; Karanikolos et al. 2013; Kondilis et al. 2013; McKee et al. 2012) indicate that cutbacks in government expenditures impact employment and unemployment conditions, health outcomes, and the consumption of health services.
Unemployment Generosity and Its Relation to Medicalizing Unemployment We investigate whether the relationship between unemployment and medical care use is moderated by welfare generosity in unemployment and
274 Journal of Health and Social Behavior 58(3)
healthcare. Our measure of generosity, taken from Scruggs, Jahn, and Kuitto (2014) captures the level of benefit payments and the conditionality and strictness of entitlements that affect the level of cov- erage (e.g., targeted vs. universal benefits). Healthcare generosity is the degree to which health- care is delivered as a social right of citizenship rather than as something to be purchased in the medical marketplace. In line with the work of Scruggs (2014), who builds on work on decommod- ification by Esping-Andersen (1990), we use the term generosity to engage with institutionalist research on welfare-state effects on population health (Beckfield et al. 2015), highlight improve- ments in measurement since Esping-Anderson’s pioneering work, and emphasize its use in more European countries with recent data.
Countries exemplifying a low level of unem- ployment generosity include the United Kingdom (Bambra 2005a) and several eastern European states. The social protection systems for the unem- ployed in these countries are relatively weak, with less generous income replacement rates and strict entitlement criteria, which may increase financial stress and lessen the feeling of self-control (Strandh 2001). For example, the maximum duration of a standard unemployment benefit in the United Kingdom, Estonia, Lithuania, Slovakia, and Czech Republic is only 26 weeks. Thereafter, the unem- ployed in the United Kingdom rely on means-tested benefits, which are known to be highly stigmatizing (Rodriguez, Frongillo, and Chandra 2001). The unemployed are often considered to be responsible for their situation (Bambra and Eikemo 2009), which can stimulate self-blame, perceptions of fail- ure, and social exclusion. All of these factors may be associated with the medicalization of unemploy- ment. As a result, we can expect that unemployment will be more strongly related to MHC use in coun- tries with low levels of unemployment generosity.
Although a consistent negative relationship between unemployment and well-being, health, and mental health has been observed in previous research (Bambra and Eikemo 2009; Strandh 2001; Wulfgramm 2014), in countries with high levels of unemployment generosity (such as some Scandinavian countries; Esping-Andersen 1990), the medicalization of unem- ployment may be weaker. There is strong protection for the unemployed through highly interventionist governments, which value universalism and social equality (Bambra and Eikemo 2009). In the more gen- erous welfare states, the level of benefits is relatively high and benefits are universalist (rather than means tested), which may result in lower stigma. As a
consequence, unemployment may be less stressful and related to reduced feelings of self-blame and per- sonal failure. Unemployment may be considered more of a social problem, which requires a structural solution.
Healthcare Generosity and Its Relation to Medicalizing Unemployment Institutional conditions and welfare policies may also affect the access to and availability of the healthcare resources needed to make the medical- ization of unemployment possible. Therefore, in addition to the role of unemployment generosity, we also explore the role of healthcare generosity.
Although healthcare is a key dimension of all modern welfare states, it is relatively absent from major welfare-state theories (Bambra 2005a). In response to this limitation, Bambra (2005a, 2005b) introduced the concept of healthcare decommodifi- cation. It accounts for the provision of care, the degree to which this provision is independent from the market, and the extent to which an individual’s access is dependent on his or her market position. The indicators included in our healthcare generos- ity measurement assess the financing, provision, and coverage of the private sector and are, there- fore, useful indicators of the varied role of the mar- ket in a healthcare system. The larger the size of the private health sector, in terms of expenditure and consumption, the larger the role of the market and, therefore, the lower the degree of healthcare gener- osity (Bambra 2005b). For example, the United Kingdom has a relatively high level of healthcare generosity because it has a coverage level of 100% combined with only 3.7% private hospital beds (of the total bed stock) and 1.72% private health expen- ditures (of the gross domestic product [GDP]). In contrast, Belgium (99.0% covered) and Germany (89.1% covered) have, respectively, 61.8% and 59.3% private hospital beds and 2.54% and 2.74% private health expenditures and therefore relatively low levels of healthcare generosity. We can expect that in countries with high levels of healthcare gen- erosity, the unemployed will be less constrained in their medical care use.
Based on this theoretical framework, we hypoth- esize that a combination of low unemployment gener- osity and high healthcare generosity—with the United Kingdom as an illustration—will trigger the medicalization of unemployment. In this situation, the unemployed will possibly perceive greater stig- matization of, and individual responsibility for, unemployment. These correlates of mental illness,
Buffel et al. 275
then, may facilitate increased MHC in generous healthcare systems.
Figure 1 shows the countries according to their combined unemployment generosity and healthcare generosity. Typical countries characterized by the inverse combination of above are Belgium and Bulgaria, the first especially regarding its high unemployment generosity level, and the later con- cerning its low healthcare generosity level. Greece is a typical country with low levels on both gener- osity measurements.
DAtA AnD MEtHODS The Eurobarometer Survey This study used data from the Eurobarometer rounds 64.4 (2005–2006) and 73.2 (2010),1 which included information about a general population ages 15 and above in more than 20 European Union member states. To our knowledge, the Eurobarometer is the
only cross-national survey that combines (1) nation- ally representative samples, (2) measurements of mental health status, (3) measurements of MHC uti- lization, (4) employment status, and (5) broad cross- national institutional variation.
The basic sample design used in all countries comprised a multistage, random (probability) sam- ple of individuals within households within an area. Interviews were conducted face-to-face in the national language. To ensure nationally representa- tive samples, poststratification weights were applied to restore specific town size, age, and gen- der distributions for the general population in each country, using the most recent census data. For our purposes, it was appropriate for small countries, such as Belgium, to be weighted the same as large countries, such as Germany (Frohlich et al. 2001). Unweighted analyses yielded more valid estima- tions. We did not weight the samples according to population size, as the population sizes of the sam- pled countries were highly heterogeneous and
Figure 1. Countries Positioned in a two-dimensional Graph of Unemployment Generosity by Healthcare Generosity Note: Unemployment generosity scores (including the replacement rate, qualifying period, duration of benefits payments, waiting period, and level of coverage) are based on the Comparative Welfare Entitlements Dataset 2 (Scruggs, Jahn, and Kuitto 2014) and calculations using Scruggs’ (2014) formula. Healthcare generosity scores (including level of coverage, private health expenditures, private hospital beds, and household out-of-pocket payments) are based on data from Eurostat (2015), the Organisation for Economic Co-operation and Development (2012), and the World Health Organization (2005, 2011), and calculations using Scruggs’ (2014) formula.
276 Journal of Health and Social Behavior 58(3)
because we were interested in the institutional foundations of medicalization.
We limited our subsample to respondents of working age (20–65 years old; N = 37,477 respon- dents). Missingness was no more than approxi- mately 2% for all variables. We omitted 1,171 cases with missing values from the sample. The final sample consisted of 24 European countries and contained information for 36,306 respondents. Descriptive statistics and the sample size per coun- try are provided in Appendix A, in the online ver- sion of this article.
Measures We constructed two dichotomous outcome vari- ables for MHC use: contacting a general practitio- ner (GP) and/or contacting a psychiatrist (each item coded 1 = yes, 0 = no).
The main independent variables were employ- ment status and mental health status. Employment status contained three categories: employed (refer- ence), unemployed, and nonemployed. Mental health was measured with the five-item version of the Mental Health Inventory (MHI-5), a subscale of the SF-36 Version 2 (Ware and Sherbourne 1992). The scale measured depression and anxiety-related complaints and ranged from 1 (good mental health) to 5 (poor mental health). If one or two items were missing, mean substitution was applied. The internal reliability of the MHI-5 scale was good (Cronbach’s alpha = .803).
Age was measured in years. Period was a cate- gorical variable: 2005–2006 (reference) and 2010. To examine within-country differences in the provi- sion of healthcare services, we controlled for the degree of urbanization using the following catego- ries:2 large town (reference), rural area or village, and small or medium-sized town. This could be considered as a proxy for supply, because the avail- ability of medical professionals varies between urban and rural areas (Saxena et al. 2007). We also controlled for marital status (married [reference], divorced, widowed, or single) and education level. Respondents were asked at what age they finished full-time education, and the European Commission provided a standard categorization for the answers: ages up to 15 (reference), 16 to 19, and 20 and above. This corresponds approximately to primary, secondary, and tertiary education.
At the country level, our central variables were the level of unemployment generosity and health- care generosity. To construct the unemployment generosity measurement, we relied on Scruggs’
updated “unemployment generosity measure” (Scruggs and Allan 2006),3 which was an adaptation of Esping-Andersen’s (1990) original measurement. Scruggs used z scores to combine information on five indicators into a single measurement that facili- tates interpretation. The five indicators were the level of benefits paid to the unemployed (replace- ment rate), the qualifying period, duration of bene- fits payments, waiting period before entitlement, and percentage of the working-age population cov- ered by the program (see online Appendix B). More information and this data set, the Comparative Welfare Entitlements Dataset (CWED 2), are avail- able at http://cwed2.org/.
For comparability with Scruggs’s measure- ment, we adapted Bambra’s (2005a, 2005b) mea- surement of the decommodification of healthcare for the construction of our healthcare generosity measurement, using the same z score technique to combine the following indicators:4 Private health expenditure as a percentage of GDP, private hospi- tal beds as a percentage of total bed stock, the cov- erage of the population by the public healthcare system, and household out-of-pocket (OOP) pay- ments as a percentage of the total health expendi- ture.5 The majority of information is available at http://data.euro.who.int/hfadb/; for coverage per- centages, we used data from the Organisation for Economic Co-operation and Development (OECD; 2012).
For both country variables, we used as much data as possible from the periods 2004–2006 and 2009–2010. We used data for the year of the inter- view and the preceding year because respondents were asked whether they had sought professional help in the year before the interview and because of an expected time lag. This also resulted in the best model fit. Both generosity measurements were interval-level variables, which were grand-mean centered.
We included additional macrolevel control vari- ables to guard against residual confounding. The effects of the nature of welfare policies concerning mental health and the MHC use of the unemployed may partly depend on the condition of the country’s labor market and the general economic capacity (GDP per capita) of a country. A short period of income support for unemployed people, for example, may be less associated with high levels of anxiety and insecurity in countries where the unemployment level is low and unemployment tends to be of short duration (Gallie, Kostova, and Kuchar 2001). We can also expect that in countries with low unemploy- ment, it will be less randomly distributed and thus
Buffel et al. 277
more frequently considered a direct or indirect con- sequence of health selection. Unemployment will be more stigmatizing, different from the norm, and treated as an individualized problem (Clark 2003), which can be a trigger for medicalization. Therefore, GDP per capita (Model 3) and unemployment rates (Model 4) were included (information derived from Eurostat 2015, Table 1).
Disability generosity was included in the mod- els as an additional control variable to take partly into account the possibility of “hidden unemploy- ment” via relying on disability benefits. Generosity in terms of disability benefits is often measured by the level of public spending on disability (Börsch- Supan 2007). This information was available from
Eurostat (2015). Although it was not the objective of this study, we could not ignore the current debate about the claim that there is a movement from a passive toward an active welfare state in several European countries (Bonoli 2010). Central to this are active labor market programs (ALMP; Knotz 2012; Strandh 2001). The level of expenditure on ALMP is often used as an indicator for the activa- tion effort of a country (Knotz 2012). While this is an approximate measurement, we included expen- diture on ALMP (as a percentage of GDP) as a con- trol variable in the multilevel analysis (data retrieved from OECD 2015).
Based on the economic crisis literature (Antonakakis and Collins 2015; Karanikolos et al. 2013; Stuckler
Table 1. Country Scores on the Generosity Measurement of Unemployment and Healthcare, and Countries’ national Unemployment rate and GDP.
Group Categorization Country Unemployment
Generositya Healthcare Generosityb
GDP per Capitac
1 High unemployment generosity and high healthcare generosity
Austria 1.42 12.04 5.5 31,450 Denmark 1.60 15.24 5.4 39,400 netherlands 11.94 14.12 5.3 33,100 Sweden 1.08 13.52 7.8 32,250 France 11.24 13.50 9.0 28,300 Ireland 1.64 13.39 7.0 37,500
2 High unemployment generosity and low healthcare generosity
Belgium 13.72 9.45 8.3 30,300 Germany 1.30 9.89 9.7 28,000 Portugal 1.50 9.31 9.1 15,250 Spain 11.04 1.70 12.7 21,900 Latvia 11.35 8.30 13.1 7,200 Bulgaria 11.55 7.46 9.7 3,800
3 Low unemployment generosity and high healthcare generosity
United Kingdom 8.70 15.77 5.7 28,350 Slovenia 9.51 14.79 6.2 15,850 Czech republic 9.73 15.95 7.6 11,900 Finland 9.22 13.10 8.5 31,150 Estonia 8.37 12.94 1.5 9,350 romania 5.68 14.45 7.2 4,750
4 Low unemployment generosity and low healthcare generosity
Greece 7.38 7.39 1.1 19,050 Hungary 7.59 11.62 7.8 8,950 Italy 5.64 11.72 7.8 24,850 Lithuania 7.51 11.62 11.0 7,350 Poland 6.79 11.08 15.0 7,250 Slovakia 7.50 11.49 15.6 9,350
Note: GDP = gross domestic product. aData from the Comparative Welfare Entitlements Dataset 2 (Scruggs, Jahn, and Kuitto 2014) and calculations using Scruggs’ (2014) formula. bData from Eurostat (2015), the Organisation for Economic Co-operation and Development (2012), the World Health Organization (2005, 2011), and calculations using Scruggs’ (2014) formula. cData from Eurostat (2015).
278 Journal of Health and Social Behavior 58(3)
et al. 2009), cuts in government expenditure on domains such as health, unemployment, ALMP, family, and housing were used as proxy of fiscal austerity. Data for general government final con- sumption expenditure as a percentage of GDP were collected from the World Bank’s (2015) World Development Indicators database. In line with Antonakakis and Collins’ (2015) work, we divided general government final consumption expenditure by real GDP, as the expenditure measurement might have been biased during a period when nominal GDP was falling. Because of a time lag of at least one year, information from 2005 for wave 64.4 (2005–2006) and 2009 for wave 73.2 (2010) was used to calculate the mean government expenditure over the periods per country. The operationalization of the change in government expenditure is explained in the next section, as it relates to the sta- tistical procedures used. Country scores on the macro variables are presented in online Appendix C.
Estimation Our analyses consisted of two parts. The first part focused on country-specific differences in the MHC use of the unemployed and how these differences were patterned by unemployment and healthcare generosity levels. In the second part, we looked for general trends regarding the impact of the level of unemployment generosity and healthcare generos- ity on the relationship between employment status and MHC use, while taking several key institutional and macroeconomic factors into account.
In the first part, on the basis of country-specific logistic regressions, we tested the relationship between employment status and MHC use and to what extent this association could be ascribed to mental health. To compare the MHC use of the unemployed with that of the employed between countries, predicted probabilities (PPs) for the unemployed and employed were calculated based on the odds ratios (ORs) resulting from the logistic regression analyses.6 The differences (PPunemployed – PPemployed) between both PPs are presented in Table 2. PPs are preferable to reporting differences in logistic regression coefficients because PPs do not require the assumption that the error variance is identical across countries. First, the PPs were based on the models controlled for age, gender, marital status, education, and period. Second, they were based on the adjusted models including mental health. The results based on Model 2 are presented in a bar chart (Figure 2) and related to country scores on the generosity measurement of unemployment and of healthcare. We categorized the countries into
groups depending on whether a country’s score was above or below the median score of all countries included in the study on the unemployment generos- ity and the healthcare generosity measure (see Figure 1). This resulted in four groups of countries with a specific combination of scores: a relatively high level on both generosity measures, a relatively low level on both measures, and a relatively high score on one measure and a relatively low score on the other (and visa versa). The four groups of coun- tries do not represent a typology, nor are they the results of a cluster analysis.
In the second part, to test whether unemploy- ment and healthcare generosity have a significant moderating effect on the relationship between employment status and MHC use, we performed logistic multilevel analyses that included cross- level interaction effects of employment status and the two generosity measurements. Multilevel anal- ysis enabled us to take the clustering of our data in periods, as well as countries, into account. However, two periods were not enough to use period as a separate level, and thus—like most repeated cross-sectional surveys––we faced a prob- lem of obtaining an adequate number of higher- level units at the period level. Given the cross-national nature of the Eurobarometer, there was a solution to this lack of sufficient repeated waves, as has been described by Fairbrother (2014): considering the clustering of different waves clus- tered within countries. National-level time-series cross-sectional data enable simultaneously model- ing cross-sectional effects that explain between- country differences and longitudinal effects that explain within-country differences over time.
In sum, respondents, as units of the individual level (level 1), were nested within country-years ranging from 2005–2006 to 2010 at the period level (level 2), which were in turn nested within countries (level 3; see online Appendix D). To control for aus- terity measurements, we estimated an additional model that took the average level of government expenditure into account as well as the change in expenditures. To include longitudinal effects at the period …