impact of the european financial crisis on health outcomes: a … · 2016-09-07 · risk for bias,...

Confidential: For Review O

nly

Impact of the European Financial Crisis on Health

Outcomes: A Systematic Review of the Literature

Journal: BMJ

Manuscript ID BMJ.2016.032729

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the Author: 31-Mar-2016

Complete List of Authors: Parmar, Divya; City University London, School of Health Sciences

Stavropoulou, Charitini; City University London,

Ioannidis, John; Stanford University, Stanford Prevention Research Center,

Department of Medicine and Department of Health Research and Policy

Keywords: European financial crisis, health outcomes, systematic literature review

https://mc.manuscriptcentral.com/bmj

BMJ


nly

1

Impact of the European Financial Crisis on Health Outcomes: A Systematic Review of the

Literature

Divya Parmar, Charitini Stavropoulou, John PA Ioannidis

Divya Parmar, Lecturer, School of Health Sciences, City University London, EC1V 0HB London, UK

Charitini Stavropoulou, Senior Lecturer, School of Health Sciences, City University London, EC1V

0HB London, UK

John PA Ioannidis, Professor, School of Medicine, Stanford University, CA 94305 Stanford, USA

Correspondence to: [email protected]

Page 1 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

2

ABSTRACT

Background: Despite the immense interest in the impact of the European financial crisis, which

intensified in 2008, evidence on its impact on the health of the population is still unclear and

fragmented.

Aim: To systematically identify, critically appraise and synthesise empirical studies about the impact

of the European financial crisis on health outcomes.

Methods: We conducted a systematic literature review by performing structural searches of key

databases, healthcare journals and organization-based websites. Empirical studies reporting on the

impact of the European financial crisis on health outcomes, published from January 2008 to

December 2015, were included. All selected studies were assessed for risk of bias. Owing to the

heterogeneity of studies in terms of study design and analysis and the use of overlapping datasets

across studies, we analysed the studies thematically per outcome and synthesised the evidence on

different health outcomes without formal meta-analysis.

Results: Thirty-nine studies met the inclusion criteria. These focus on suicide, mental health, self-

reported health, mortality and other health outcomes. Of those, 29 (74%) were deemed to be at high

risk for bias, eight (21%) at moderate risk for bias, and only two at low risk for bias. Although there

are differences across countries and groups, the evidence largely points towards an increase in

suicides and the deterioration of mental health. The crisis did not seem to reverse the trend of

decreasing overall mortality. Evidence on self-reported health and other indicators is mixed.

Discussion and Conclusions: The European financial crisis has had heterogeneous effects on health

outcomes, with the evidence being most consistent for suicides and mental health. Most published

studies have a substantial risk for bias. There is a need for better empirical studies, especially those

focused on identifying mechanisms that can mitigate the adverse effects of the crisis.

Page 2 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

3

INTRODUCTION

In 2008, Europe entered a period of unprecedented financial crisis following a global economic

downturn. A number of EU countries faced declining GDP, increasing public debt, and rising

borrowing costs, while individual households experienced financial insecurity created by job loss,

reduced salaries and plummeting house prices.1 The situation worsened by early 2010, and Greece

became the first EU country to get a bailout package jointly from the International Monetary Fund,

the EU and the European Central Bank. Ireland, Portugal and Cyprus followed a few months later.

The effect of the financial crisis on European health systems was inevitable. Different

countries responded with different policies. Many countries reduced their health budgets, and some

had to introduce structural changes and tough austerity measures.2 Fears about the adverse impact of

the financial crisis on health outcomes have been increasing, and, consequently, the number of studies

investigating this impact has grown in the last few years.3 4 It has been argued that the impact of a

financial crisis is not always uniform across countries, but may depend on the duration and severity of

the crisis, the type of austerity measures introduced by the government, and whether the populations

are covered by social protection schemes.5

Yet—and despite growing interest in the impact of the European financial crisis on health

outcomes—the evidence so far has been fragmented. Studies focus on particular countries (e.g.,

Simou and Koutsogeorgou6 on Greece), on particular health outcomes (e.g., van Hal7 on

psychological well-being), or on particular groups (e.g., Rajmil et al.8 on children). The aim of this

paper is to systematically identify, critically appraise and synthesise the empirical evidence on the

impact of the European financial crisis on health outcomes.

METHODS

Search strategy and selection process

Our steps for identifying and reviewing the evidence were based on the recommendations of the

Centre for Reviews and Dissemination.9 First, we conducted structured searches of online databases:

PUBMED, ISI Web of Knowledge, EBSCOhost, Scopus, and Google Scholar. Second, we hand-

searched key healthcare journals: The Lancet, the European Journal of Public Health, Health Policy,

Social Science and Medicine, the British Medical Journal, PLoS One, and BMJ Open. Finally, we

searched the websites of relevant organizations that have published reports and data on population

health and healthcare, including the World Health Organisation, OECD, European Observatory on

Health Systems and Policies, and UNICEF.

The search terms we employed included ‘financial crisis’; ‘economic crisis’; ‘recession’;

‘austerity’; in conjunction with ‘health’; ‘health outcomes’; ‘healthcare’; and ‘Europe’.

We included studies published from January 2008, when the European financial crisis

intensified, to December 2015. Only studies published in English were included. We limited our

Page 3 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

4

search to empirical, quantitative papers and excluded qualitative studies, opinion papers,

commentaries and systematic literature reviews. We excluded a) empirical cross-sectional studies that

do not have a control group—i.e., a country that was not hit by the crisis; b) longitudinal studies that

do not follow the same group before and after the crisis; c) studies that present only descriptive

evidence; d) studies that examine the impact of the crisis on health behaviours, such as smoking and

drinking; and e) studies that look at the impact of macroeconomic indicators on health outcomes in

general, unless they specifically investigate this in the context of the current crisis—i.e., by clearly

defining the period of crisis. We also excluded conference proceedings and abstracts if the full text

was not available. Multi-country studies were included if their results on European countries could be

identified separately.

Our research assistant (RA) searched the above websites and, along with one of the authors,

initially screened titles for relevance to the topic. The RA and one author then screened the abstracts

of the selected papers independently, and when disagreements occurred, the second author became

involved. The two authors then read and discussed the full text of the papers and excluded those that

did not meet the aims of the study. In this way, we decided on the final list of papers to be included

for analysis. The RA searched the references and citations of the selected papers. Abstracts and the

full text of additional papers identified were reviewed in the same manner. This review process

followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)

guidelines.10 The PRISMA checklist is provided in the Appendix (Table A).

Data extraction and analysis

Data were extracted using a standardised data extraction form. The extracted data include: country,

period examined, definition of the crisis period, crisis indicator, data and population, study design and

methods, main findings, and health outcome examined. Due to the heterogeneity of the studies in

terms of study design and analysis, we decided to thematically analyse them per outcome. Formal

quantitative synthesis (meta-analysis) was further impeded by the fact that some studies covered

overlapping questions and settings and/or used different analyses on overlapping datasets; whenever

such overlap exists, we point this out. The data extraction table is provided as an Appendix (Table B).

Risk of bias assessment

Given the nature of our research question and the lack of appropriate, widely-used quality assessment

tools for the type of studies we reviewed, we developed our own tool and assessed the quality of the

included studies over seven key domains of bias: 1) selection bias; 2) ecologic fallacy; 3) confounding

bias; 4) reporting bias; 5) time bias; 6) measurement error in exposure indicator; and 7) measurement

error in health outcome. The tool is presented in Table 1, along with the definition of each domain.

Studies were given a rating for each domain, with each one scored as 1 (Strong - low risk of bias), 2

(Moderate – moderate risk of bias), or 3 (Weak - high risk of bias). An overall rating for each study

Page 4 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

5

was based on the rating of each domain. A study was given an overall rating of 1 (strong) if none of

its domains was rated as weak; 2 (moderate) if up to two domains were rated as weak; and 3 (weak) if

three or more domains were rated as weak. Each study was assessed independently by two authors.

We compared the rating for each domain, as well as the overall rating, and reached a consensus on the

final rating for each included study. The weighted kappa for each domain, as well as for the overall

rating, was calculated.

[Insert Table 1 here]

Patient involvement

No patients were involved in the study.

RESULTS

The results of the review process are shown in Figure 1. We screened 4,801 studies by title and

abstract for possible inclusion. The full text of 107 studies was assessed for eligibility. In total, 39

studies met our selection criteria and were included in the systematic review.

[Insert Figure 1 here]

Table B in the Appendix presents the data extraction of the papers that were included in the

review. The vast majority of the papers focus on Spain (n=10) and Greece (n=9). Other countries

included the UK and its four constituencies (n=6), Ireland (n=1), Italy (n=1), Iceland (n=2) and France

(n=1). Eight papers use data from more than one country, either to collectively analyse the data or to

compare outcomes across countries. The health outcomes that the papers study included suicide

(n=14), mental health (n=14), self-rated health (n=12), mortality (n=2) and other health outcomes

(n=9). Some papers study more than one health outcome. The studies are summarised in Table 2.


Risk of bias assessment

Of the 39 studies we reviewed, 29 (74%) were rated as weak, showing a high risk for bias in at least

three domains. Eight studies (21%) were rated as moderate, showing a high risk for bias in up to two

domains (Table 2). Only two studies were rated strong in overall risk assessment. The weighted kappa

for each domain and the overall rating are also presented in Table 3.

Most of the studies suffered from a high risk of ecological fallacy, as they make inferences

about individual characteristics based on aggregate-level data. Time bias is also a common concern.

Most studies use data covering periods shorter than ten years, consider fewer than three years post-

Page 5 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

6

crisis, and do not account for potential lag effects. Many studies also are at high risk for reporting

bias, measurement bias in exposures, and measurement bias in outcome assessment.


Suicides

Most of the 14 studies report a significant increase in suicides during the financial crisis and find that

men, particularly those of working age and the unemployed, were more significantly affected, while

suicide rates among women were largely unaffected.

All 4 Greek studies use data from the Hellenic Statistical Authority (ELSTAT) but apply

different analyses. Using interrupted time-series analysis from 1983 to 2012, Branas et al.11 find that

total and male suicides increased in June 2011 by 35.7% (p<0.001) and 18% (p<0.01), respectively,

and female suicides increased by 35.8% (p<0.05) in May 2011. Kontaxakis et al.12 use data from

2001-2011 and compare the specific suicide rate in 2008-2011 versus 2001-2007. They find that

overall SSR increased (from a negative rate of -3.9% to a positive 27.2%), and the rate for men

increased (from a negative rate of -8.4% to a positive +26.9%, p=0.047), especially for men aged 30-

54 years, and decrease for men aged 60-64 years. For women, no statistically significant change was

observed. Madianos et al.13 using 1990-2011 data, find that age-adjusted suicide rates increased by

19% between 2005 and 2011. Rachiotis et al.14 using ELSTAT data for 2003-2012 and correlation and

regression analyses, find that overall suicide rates increased 35% between 2010-2012; rates increased

for both genders. Comparing 2003-2010 and 2011-2012, suicides increased by 29% for men (p<0.01)

and 33% for women (p=0.03). Unemployment in 2003-2012 was associated with suicides for

working-aged men but not for women.

In Spain, Cordoba-Dona et al.15 find an increase in suicide attempt rates for both genders

during 2008-2012 in Andalucía. They find a significant association between suicide attempts and

unemployment for men but not for women. Another Spanish study,16 using monthly suicide rates by

region and age group and population data from Spain’s National Statistics Institute for 2005-2010,

finds 8% increase in the suicide rate (p=0.03) during the crisis. Suicides increased by 9% in the

Mediterranean region and by 19% among men. Results for women and fother regions are non-

significant.

In Ireland, an interrupted series analysis using data from the Irish Central Statistics Office

finds that, by the end of 2012, the male suicide rate was 57% higher (i.e., 476 more suicides) than if

the pre-crisis trend had continued.17 Those affected the most were men aged 25-44. Female suicide

rates were unchanged.

In England and Wales, Coope et al.18 find no change in overall male suicides. They document

a halt in the previous downward trend in suicide rates for men aged 16-34 in 2006 and an upward

trend in early 2010 for those aged 35-44. For women, there was no change in suicide rates. Another

Page 6 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

7

study finds no change in suicides rates in England after 2008 for both men and women, with potential

heterogeneity across regions.19

Six studies provide comparative results across several countries. Laanani et al.20 using data

for 2000-2010 from 8 countries find a 3.3% increase in the overall suicide rate during the crisis.

Associations between unemployment rate and suicide rate was significant in three countries: 0.7%

(95% CI 0.0% to 1.4%) in the Netherlands, 1.0% (95% CI 0.2% to1.8%) in the UK and 1.9% (95% CI

0.8% to 2.9%) in France. They estimate that unemployment accounted for 564 additional suicides in

France, 57 in the Netherlands and 456 in the UK. They find no associations in Austria, Finland, Spain

and Sweden. Comparing suicide rates for 2003-2007 and 2008-2012 in Greece, Iceland and Finland,

Tapia Granados and Rodriguez21 find an increase in suicides in 2008-2012 for Greece only, although

they report that even at its peak in 2011, this rate remained low—a third of the mean suicide rate in

the EU.

Using 2000-2009 WHO mortality data for 27 European countries, Chang et al.22 find 2937

additional male suicides in 2009 over the previous trend, with the highest increase among males aged

15-24, while suicide rates for women remained stable. Another study on 29 European countries during

2000-2011 finds a strong correlation between male suicide rates and all economic indices except for

GDP per capita, and a correlation between female suicides and unemployment only.23 They find that

the temporal relationship does not support causality (suicide rates increased several months before the

crisis emerged). Reeves et al.24 use data on male suicides from 20 EU countries (1981-2011) and

estimate 6,998 excess suicides post-2008, based on pre-crisis suicide trends. Of these, 1,077 (15%) are

attributed to increased unemployment. Further, they estimate that spending on active labour market

programmes prevented 540 (50%) suicides, and high levels of social capital prevented another 210

(19%).

Mental health

Most, but not all, of the 14 studies find an association between deteriorating economic indicators and

poor mental health, particularly among men.

In Italy, De Vogli et al.25 find that the crisis resulted in an additional 548 deaths due to mental

and behavioural disorders (i.e., 0.303 per 100,000 deaths per year). Further, they estimate that 22.4%

of these deaths can be attributed to income loss and 16.4% to unemployment.

In Greece, using longitudinal telephone surveys, Drydakis26 finds that unemployment was

associated more with poor mental health in 2010-2013 than in 2008-2009. Among women,

unemployment increased poor mental health from 4.3% to 7.3% and, among men, from 3% to 4.9%

(p<0.001). Another Greek study that finds the odds of suffering from major depression was 2.6 times

greater in 2011 than in 2008.27

In Spain, four studies use data from the Spanish National Health Surveys (2006 and 2011/12).

Among young people (16-24 years), Aguilar-Palacio et al.28 find that the prevalence of poor mental

Page 7 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

8

health decreased for women in 2012 versus 2006, with no change for men. However, they identify

unemployment (>1 year) as a risk factor for poor mental health in 2012 for men. Bartoll et al.29 also

find reduced prevalence of poor mental health among young women (prevalence ratio 0.89, p<0.001)

with no change for men. For the older population, they find that the prevalence of poor mental health

increased among men with no change among women over age 25. Comparing immigrants with

natives, Gotsens et al.30 report a new onset of inequalities in poor mental health between immigrants

and native men and an equalisation of the previously lower use of psychotropic drugs by male

immigrants, with no change for female immigrants. Finally, Zapata Moya et al.31 report that

depression increased by 12% in 2009 (p<0.05) and by 23% in 2011 (p<0.001) among women, and by

13% in 2011 (p<0.10) among men compared to 2003. These effects disappear after controlling for

changes in GDP growth rate.

Of the two remaining studies on Spain, Gili et al.32 find that unemployment was associated

with major depressive disorders in both 2006 and 2010 (odds ratio=1.54 and 1.72, p<0.001) among

primary healthcare attendees. In 2010, the association between unemployment and minor depressive

disorders increased (OR=1.20, p<0.001), while for dysthymia, there was no association. One third of

major depression in 2012 is attributed to individual and family unemployment and mortgage payment

difficulties. Lastly, Rajmil et al.33, studying children in Catalonia, report no changes in 2010-2012

versus 2006 and no association between children’s mental health and parental employment status.

In France, using a prospective national survey, Malard et al.34 find no change in the

prevalence of major depressive episodes and generalized anxiety disorder (GAD) for both genders in

the working population— except for an increase of 7.4% (p=0.007) in GAD among women working

in the public sector. Astell-Burt and Feng,35 using population data from the Quarterly Labour Force

Survey in the UK, find a 0.2% increase in depression and 0.1% increase in mental illness during the

crisis. Curl and Kearns36 examine the effect of financial difficulties on mental health in deprived areas

of Glasgow and find that decreased affordability was associated with declining mental health.

Katikireddi et al.,37 using repeated cross-sectional surveys in England, find that prevalence of poor

mental health in men increased after 2008 (by 5% in 2009 and by 3% in 2010), but the men’s

changing employment status does not explain this increase. The same study shows no change in poor

mental health prevalence among women. Barr et al.38 find that, although mental health problems

increased post-crisis (+0.08% prevalence per quarter), only 36% of the additional problems could be

explained by rising unemployment and declining wages.

Self-rated health

Twelve studies focus on the impact of the crisis on self-rated health (SRH), finding mixed results

depending on the country and group analysed.

Three studies on Spain, analysing data from the Spanish National Health Survey (SNHS), find

an improvement in SRH during the crisis. Aguilar-Palacio et al.28 find that people aged 16-24 had

Page 8 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

9

better SRH in 2012 than in 2006 and this is more evident among women. They find no association

between unemployment and SRH, but students of both genders had lower risk of poor SRH than

workers did. Bartoll et al.39, show that the probability of good SRH increased for by 7.6% (p<0.01) for

men and by 9.6% (p<0.01) for women in 2011/2 versus 2001. Regidor et al.,40 finds that the

prevalence of poor SRH declined during the financial crisis by 5.7%. A fourth Spanish study using

data from the same survey30 finds that immigrants who arrived before 2006 had worse SRH than

natives. For women, inequalities in poor SRH between immigrants and natives increased.

A study on Catalonia33 observes an improvement in health-related quality of life (HRQOL),

but children whose mothers had only a primary education had poorer HRQOL scores in 2010-12 than

in 2006.

In Greece, Zavras et al.41 find that the overall prevalence of good and very good SRH in 2006

was 71.0%, whereas in 2011, the prevalence declined to 68.8% (p< 0.05), with the unemployed less

likely to report good health. Another study using a different cross-sectional survey finds that

unemployment was associated with poor SRH, and women were more affected than men.26 A

difference-in-difference study comparing Greece to Ireland, using Poland as a control, finds an

increase in the prevalence of poor SRH in Greece but not in Ireland in 2010.42 A similar study

comparing Greece to Poland finds an increase in poor SRH in Greece after the crisis.43

Three comparative studies find evidence of a negative impact of the crisis on SRH.

Comparing Estonia, Lithuania and Finland, Reile et al.44 find that during 2008-2010, the prevalence of

poor SRH increased to 52% (from 50%) in Estonia and to 48% (from 47%) in Lithuania. The increase

was not statistically significant, but marked the end of the previous positive trend of improving health

status. Studying 23 European countries, Ferrarini et al.45 find that unemployment insurance reduced

the transition to deteriorating SRH during the crisis. Huijts et al.46 examine the impact of job loss and

recovery on SRH in 27 European countries. Job loss during the crisis was negatively associated with

SRH for both women and men. Job recovery within a year led to health recovery for women, while

men experienced an enduring health disadvantage.

Mortality

Of the two studies on mortality rates, Regidor et al.,40 using 1995-2011 national registry data from

Spain, find that premature mortality rates from several causes fell during the crisis. Overall mortality

in 2008-2011 dropped by 2.5% (cardiovascular diseases 5.6%; respiratory diseases 2.8%; digestive

diseases 2.8%; genitourinary diseases 7%; HIV 6.8%; other infectious diseases 8.7%; road accidents

14.2%; other unintentional injuries 3.8%; suicides 1.3%; homicides 6.9%; cancer mortality remained

constant). Tapia-Granados and Rodriguez21 analyse WHO data on Iceland, Finland and Greece for

1990-2012 and conclude that the crisis had no impact on mortality, as mortality dropped faster or

continued falling after 2007 as quickly as in previous years for almost all age groups. Comparisons

Page 9 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

10

between Greece and the other two countries show no significant differences in most mortality

indicators.

Other health outcomes

Nine studies examining other health outcomes find mixed results. Eiriksdottir et al.,47 studying infant

health in Iceland, find increased odds of low-birth-weight deliveries after the crisis (from 2.5% to

3%), particularly among mothers younger than 25 and the unemployed. After controlling for age,

parity and seasonality, the increase was significant (adjusted OR=1.25 95% CI 1.02-1.53), but the

association disappeared after controlling for other variables (sex, diabetes, hypertension, relationship

status, place of residency and employment status). There was no significant change in small-for-

gestational age and preterm births. When looking at pregnancy-induced hypertensive disorders,

another Icelandic study finds that the prevalence increased in the first year following the economic

collapse (2.4% vs 3.9%; adjusted OR 1.47; 95% CI 1.13-1.91) but not in subsequent years.48 The

association disappeared after controlling for unemployment. Comparing Greece, Finland and Iceland,

a study finds no effect on the incidence of TB and HIV during the crisis, which continued to drop.21

Regidor et al.40 find the same for HIV incidence in Spain (1% decrease, not significant). Gotsens et

al.30 and Rajmil,33 when controlling for unemployment, also find no significant changes in Spain

regarding chronic activity limitation among immigrant women and child obesity. Zapata Moya et al.31

present mixed results in Spain, showing that, compared to 2006, in 2011, diabetes increased among in

both genders, but the effect was non-significant when controlling for real GDP growth rate and low

work intensity indicator. The study also finds a marginal decrease in myocardial infarction for men

and malignant tumours in women during the same period, but this effect is non-significant after

controlling for the macroeconomic context. Astell-Burt and Feng35 report an increased prevalence of

cardiovascular (0.6%) and respiratory problems (1%) in the UK in 2010 versus 2008. Reeves et al.49

combine data for 21 EU countries from 1991-2012 and conclude that during 2007-2012, there was no

significant association between social protection spending and TB case detection (-0.59 for a US$100

increase in social protection spending, 95% CI -1.31 to 0.14).

DISCUSSION

The amount of evidence on the impact of the European financial crisis on health outcomes is growing.

Thirty-nine studies met our criteria and were analysed, the vast majority of them focusing on two

countries in the South: Spain and Greece. The main health outcomes that these studies explore are

suicides and mental health.

Although they find differences across and within countries, most studies show a significant

increase in suicides during the financial crisis, in particular among men. Studies looking at mental

health find similar increases, but these results are more mixed. Studies focusing on mortality show a

Page 10 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

11

different picture, with overall mortality not being affected or even declining during the crisis years. It

has been argued that this was probably due to fewer working hours50 and healthier lifestyles during

years of economic difficulties.28 Some claim that factors other than the crisis, such as improvements

in road safety policies and a declining prevalence of injectable-drug use,50 explain these findings.

Our main finding of mixed effects on health outcomes squares with most previous literature

on financial crises and health.51 An empirical study conducted prior to the crisis52 on data from 26

European countries (1970-2007) shows that unemployment was positively associated with suicide

rates and homicides and negatively with deaths due to road accidents. Data from the US also suggest

that overall mortality is procyclical and decreases during financial crises, while suicides are

countercyclical and increase when the economy worsens.53

Our findings shed some light on the groups affected the most during the financial crisis. The

results on gender and age are somehow contradictory, but, overall, men of working age seem to have

been more severely affected, as reflected mainly in suicide trends and self-rated health. In terms of

mental health, though, women seem to have performed worse than men. There is also some evidence

that the health of immigrants, especially those who had illegal status and lacked social security,

deteriorated much more during the crisis than that of natives. This is consistent with previous studies

that show a worse impact on groups that lack social protection.52 Finally, there is some evidence

suggesting that the crisis increased social inequalities in health, disproportionately affecting

immigrants 30, those who were less educated, and those living in certain regions.19

The studies reviewed present significant methodological limitations. Earlier studies were

inevitably constrained by data availability; many mortality and suicide studies use data from only one

or two years into the crisis. In addition, there are issues with the design of some studies, such as those

on self-reported health that used cross-sectional data and, therefore, had difficulties establishing

temporality, let alone causality. Most importantly, from a policy perspective, most studies fail to

capture the mechanisms that affect health outcomes: studies that use year dummies to capture the

impact of the financial crisis or split the data into pre- and post-crisis periods are not informative as to

whether the effect was due to reductions in government spending, increased household financial

constraints, or both. Finally, observed age and gender differences should be interpreted cautiously as

all subgroup analyses that are difficult to know whether they were pre-specified.

A major limitation of this review is that it inevitably explored relatively short-term effects of

the crisis on health outcomes. It may take some years before the full consequences of the financial

crisis are observed.54 Moreover, reporting bias can be an issue not only for single studies, but also for

the field at large. Publication bias may result in significant associations being preferentially published,

and this may be an even greater issue for secondary, exploratory analyses of factors or subgroups

associated with different health outcomes. Furthermore, depending on what investigators believe

about topics that are socially and politically sensitive, the published evidence may be affected by

allegiance and confirmation biases. We also focused only on English language papers and those that

Page 11 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

12

have been published in full text. As a robustness check, we looked at studies that were published in

Greek, Spanish and Italian that were excluded from our analysis on the basis of language, only to

confirm that these papers did not meet our other selection criteria. Proceedings abstracts and local

language publications may also contain some useful information, but their quality is less certain, and

it is possible that allegiance biases may be even stronger in some of them. Finally, measurement error

can be substantial for some of the examined crisis indicators and health outcomes. In particular,

suicides may be under-recorded and misreported in death certificates for social or religious reasons in

some countries, such as Greece. It is unknown whether the crisis affected the extent of this potential

bias.

CONCLUSIONS

The evidence on the impact of the European financial crisis on health outcomes is mixed, and

the data and methodologies employed by many papers are susceptible to substantial bias. We need

more empirical studies that explore the impact of the crisis on health and, more importantly,

investigate the mechanisms that affect health outcomes. As more data become available, it is also

possible to consider the potential lag effects of the crisis, a clear limitation of most studies to-date.

Better data are needed and governments should make accurate data on health outcomes available as

quickly as they seem to do for economic indicators.

Page 12 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

13

FOOTNOTES

We would like to thank Maria Raisa Jessica (Ryc) Aquino for helping with the initial literature search.

Funding declaration: The study was financially supported by a pump-priming scheme from the

School of Health Sciences at City University London.

Contributors: DP and CS were involved in the initial conception and design of the study. DP secured

funding. DP and CS developed the search strategy and extracted data from included studies. DP and

CS were involved in the data analysis. DP, CS and JI developed the assessment of risk of bias tool.

DP, CS and JI were involved in the interpretation and discussion of results. CS developed the first

draft of the manuscript and all authors critically revised it and approved the final version. DP and CS

are co-first authors and study guarantors.

Copyright/Licence: The Corresponding Author has the right to grant on behalf of all authors and

does grant on behalf of all authors, a worldwide licence to the Publishers and its licensees in

perpetuity, in all forms, formats and media (whether known now or created in the future), to i)

publish, reproduce, distribute, display and store the Contribution, ii) translate the Contribution into

other languages, create adaptations, reprints, include within collections and create summaries, extracts

and/or, abstracts of the Contribution, iii) create any other derivative work(s) based on the

Contribution, iv) to exploit all subsidiary rights in the Contribution, v) the inclusion of electronic links

from the Contribution to third party material where-ever it may be located; and, vi) licence any third

party to do any or all of the above.

Competing interests: All authors have completed the ICMJE uniform disclosure form at

http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and

declare: the study was financially supported by a pump-priming scheme from the School of Health

Sciences at City University London; no financial relationships with any organisations that might have

an interest in the submitted work in the previous three years; no other relationships or activities that

could appear to have influenced the submitted work.

Ethical approval: Not required.

Transparency: The study guarantors (DP and CS) affirm that the manuscript is an honest, accurate, and transparent account of the study being reported. No important aspect of the study has been omitted. No discrepancies are withheld.

Page 13 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

14

REFERENCES

1. Cylus J, Pearson M. The crisis and its implications for household financial security, government resources and health expenditure. In: Thomson S, Figueras J, Evetovits T, et al., eds. Economic Crisis, Health Systems and Health in Europe: Impact and implications for policy. Maidenhead: Open University Press, 2015.

2. Quaglio G, Karapiperis T, Van Woensel L, et al. Austerity and health in Europe. Health Policy (Amsterdam, Netherlands) 2013;113(1-2):13-19.

3. Kentikelenis A, Karanikolos M, Papanicolas I, et al. Health effects of financial crisis: omens of a Greek tragedy. Lancet 2011;378(9801):1457-58.

4. Karanikolos M, Mladovsky P, Cylus J, et al. Financial crisis, austerity, and health in Europe. Lancet 2013;381(9874):1323-31.

5. Stuckler D, Basu S, Suhrcke M, et al. The public health effect of economic crises and alternative policy responses in Europe: an empirical analysis. The Lancet 2009;374(9686):315-23.

6. Simou E, Koutsogeorgou E. Effects of the economic crisis on health and healthcare in Greece in the literature from 2009 to 2013: A systematic review. Health Policy 2014;115(2-3):111-19.

7. Van Hal G. The true cost of the economic crisis on psychological well-being: A review. Psychology Research and Behavior Management 2015;8.

8. Rajmil L, Fernandez de Sanmamed MJ, Choonara I, et al. Impact of the 2008 economic and financial crisis on child health: a systematic review. Int J Environ Res Public Health 2014;11(6):6528-46.

9. CRD. Systematic Reviews: York Publishing Services Ltd, 2009. 10. Moher D, Liberati A, Tetzlaff J, et al. Preferred Reporting Items for Systematic Reviews and

Meta-Analyses: The PRISMA Statement. PLoS Med 2009;6(7):e1000097. 11. Branas CC, Kastanaki AE, Michalodimitrakis M, et al. The impact of economic austerity and

prosperity events on suicide in Greece: a 30-year interrupted time-series analysis. BMJ Open 2015;5(1).

12. Kontaxakis V, Papaslanis T, Havaki-Kontaxak B, et al. Suicide in Greece: 2001-2011. Psychiatriki 2013;24(3):170-74.

13. Madianos MG, Alexiou T, Patelakis A, et al. Suicide, unemployment and other socioeconomic factors: Evidence from the economic crisis in Greece. European Journal of Psychiatry 2014;28(1):39-49.

14. Rachiotis G, Stuckler D, McKee M, et al. What has happened to suicides during the Greek economic crisis? Findings from an ecological study of suicides and their determinants (2003-2012). BMJ Open 2015;5(3):e007295.

15. Córdoba-Doña JA, San Sebastián M, Escolar-Pujolar A, et al. Economic crisis and suicidal behaviour: the role of unemployment, sex and age in Andalusia, southern Spain. International Journal For Equity In Health 2014;13:55-55.

16. Lopez Bernal JA, Gasparrini A, Artundo CM, et al. The effect of the late 2000s financial crisis on suicides in Spain: an interrupted time-series analysis. Eur J Public Health 2013;23(5):732-6.

17. Corcoran P, Griffin E, Arensman E, et al. Impact of the economic recession and subsequent austerity on suicide and self-harm in Ireland: An interrupted time series analysis. Int J Epidemiol 2015;44(3):969-77.

18. Coope C, Gunnell D, Hollingworth W, et al. Suicide and the 2008 economic recession: who is most at risk? Trends in suicide rates in England and Wales 2001-2011. Soc Sci Med 2014;117:76-85.

19. Saurina C, Bragulat B, Saez M, et al. A conditional model for estimating the increase in suicides associated with the 2008-2010 economic recession in England. J Epidemiol Community Health 2013;67(9):779-87.

20. Laanani M, Ghosn W, Jougla E, et al. Impact of unemployment variations on suicide mortality in Western European countries (2000-2010). J Epidemiol Community Health 2015;69(2):103-9.

21. Tapia Granados JA, Rodriguez JM. Health, economic crisis, and austerity: A comparison of Greece, Finland and Iceland. Health Policy 2015.

22. Chang SS, Stuckler D, Yip P, et al. Impact of 2008 global economic crisis on suicide: time trend study in 54 countries. Bmj 2013;347:f5239.

Page 14 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

15

23. Fountoulakis KN, Kawohl W, Theodorakis PN, et al. Relationship of suicide rates to economic variables in Europe: 2000-2011. Br J Psychiatry 2014;205(6):486-96.

24. Reeves A, McKee M, Stuckler D. Economic suicides in the Great Recession in Europe and North America. British Journal of Psychiatry 2014;205(3):246-47.

25. De Vogli R, Vieno A, Lenzi M. Mortality due to mental and behavioral disorders associated with the Great Recession (2008-10) in Italy: a time trend analysis. Eur J Public Health 2014;24(3):419-21.

26. Drydakis N. The effect of unemployment on self-reported health and mental health in Greece from 2008 to 2013: A longitudinal study before and during the financial crisis. Social Science & Medicine 2015;128:43-51.

27. Economou M, Madianos M, Peppou LE, et al. Major depression in the era of economic crisis: a replication of a cross-sectional study across Greece. Journal Of Affective Disorders 2013;145(3):308-14.

28. Aguilar-Palacio I, Carrera-Lasfuentes P, Rabanaque MJ. Youth unemployment and economic recession in Spain: influence on health and lifestyles in young people (16-24 years old). Int J Public Health 2015;60(4):427-35.

29. Bartoll X, Palencia L, Malmusi D, et al. The evolution of mental health in Spain during the economic crisis. Eur J Public Health 2014;24(3):415-8.

30. Gotsens M, Malmusi D, Villarroel N, et al. Health inequality between immigrants and natives in Spain: the loss of the healthy immigrant effect in times of economic crisis. Eur J Public Health 2015.

31. Zapata Moya AR, Buffel V, Navarro Yanez CJ, et al. Social inequality in morbidity, framed within the current economic crisis in Spain. International Journal for Equity in Health 2015;14:131.

32. Gili M, Roca M, Basu S, et al. The mental health risks of economic crisis in Spain: Evidence from primary care centres, 2006 and 2010. European Journal of Public Health 2013;23(1):103-08.

33. Rajmil L, Medina-Bustos A, Fernandez de Sanmamed MJ, et al. Impact of the economic crisis on children's health in Catalonia: a before-after approach. BMJ Open 2013;3(8):e003286.

34. Malard L, Chastang JF, Niedhammer I. Changes in major depressive and generalized anxiety disorders in the national French working population between 2006 and 2010. J Affect Disord 2015;178:52-9.

35. Astell-Burt T, Feng X. Health and the 2008 economic recession: evidence from the United Kingdom. Plos One 2013;8(2):e56674-e74.

36. Curl A, Kearns A. Financial Difficulty and Mental Wellbring in an Age of Austerity: The Experience in Deprived Communities. Social Policy and Society 2015;14(2):217-40.

37. Katikireddi SV, Niedzwiedz CL, Popham F. Trends in population mental health before and after the 2008 recession: a repeat cross-sectional analysis of the 1991-2010 Health Surveys of England. BMJ Open 2012;2(5).

38. Barr B, Kinderman P, Whitehead M. Trends in mental health inequalities in England during a period of recession, austerity and welfare reform 2004 to 2013. Soc Sci Med 2015;147:324-31.

39. Bartoll X, Toffolutti V, Malmusi D, et al. Health and health behaviours before and during the Great Recession, overall and by socioeconomic status, using data from four repeated cross-sectional health surveys in Spain (2001-2012). BMC Public Health 2015;15:865.

40. Regidor E, Barrio G, Bravo MJ, et al. Has health in Spain been declining since the economic crisis? Journal Of Epidemiology And Community Health 2014;68(3):280-82.

41. Zavras D, Tsiantou V, Pavi E, et al. Impact of economic crisis and other demographic and socio-economic factors on self-rated health in Greece. The European Journal of Public Health 2013;23(2):206-10.

42. Hessel P, Vandoros S, Avendano M. The differential impact of the financial crisis on health in Ireland and Greece: a quasi-experimental approach. Public health 2014;128(10):911-19.

43. Vandoros S, Hessel P, Leone T, et al. Have health trends worsened in Greece as a result of the financial crisis? A quasi-experimental approach. European Journal of Public Health 2013;23(5):727-31.

Page 15 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

16

44. Reile R, Helakorpi S, Klumbiene J, et al. The recent economic recession and self-rated health in Estonia, Lithuania and Finland: a comparative cross-sectional study in 2004-2010. J Epidemiol Community Health 2014;68(11):1072-9.

45. Ferrarini T, Nelson K, Sjoberg O. Unemployment insurance and deteriorating self-rated health in 23 European countries. J Epidemiol Community Health 2014;68(7):657-62.

46. Huijts T, Reeves A, McKee M, et al. The impacts of job loss and job recovery on self-rated health: testing the mediating role of financial strain and income. Eur J Public Health 2015;25(5):801-6.

47. Eiriksdottir V, Asgeirsdottir T, Bjarnadottir R, et al. Low Birth Weight, Small for Gestational Age and Preterm Births before and after the Economic Collapse in Iceland: A Population Based Cohort Study. PLoS One 2013;8(12):e80499.

48. Eiriksdottir VH, Valdimarsdottir UA, Asgeirsdottir TL, et al. Pregnancy-Induced Hypertensive Disorders before and after a National Economic Collapse: A Population Based Cohort Study. PLoS One 2015;10(9):e0138534.

49. Reeves A, Basu S, McKee M, et al. Social protection and tuberculosis control in 21 European countries, 1995-2012: a cross-national statistical modelling analysis. The Lancet Infectious Diseases 2014;14(11):1105-12.

50. Regidor E, Barrio G, Bravo M, et al. Has health in Spain been declining since the economic crisis? Journal of Epidemiology and Community Health 2014;68:280-82.

51. Catalano R, Goldman-Mellor S, Saxton K, et al. The health effects of economic decline. Annu Rev Public Health 2011;32:431-50.

52. Stuckler D, Basu S, Suhrcke M, et al. The public health effect of economic crises and alternative policy responses in Europe: an empirical analysis. Lancet 2009;374(9686):315-23.

53. Ruhm C. Health effects of economic crisis. NBER 2015;Working Paper 21604. 54. Fountoulakis KN, Gonda X, Dome P, et al. Possible delayed effect of unemployment on suicidal

rates: the case of Hungary. Ann Gen Psychiatry 2014;13:12.

Page 16 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

17

TABLES AND FIGURES

Figure 1. PRISMA flow diagram

Records identified through database searching

(n = 5670)

Scr

ee

nin

g

Incl

ud

ed

E

lig

ibil

ity

Ide

nti

fica

tio

n

Additional records identified through other sources

(n = 17)

Records after duplicates removed (n = 4801)

Records screened

(n = 4801)

Records excluded

(n = 4694)

Full-text articles assessed for eligibility

(n = 107)

Full-text articles excluded, with reasons (n = 68)

• Conference proceedings: 5

• Not empirical/review: 13

• Cross-sectional with no

control country: 16

• Not on health outcomes: 7

• Not on European countries: 1

• 2008 crisis not specifically analysed: 25

• Qualitative:1 Studies included in

quantitative synthesis (n = 39)

Page 17 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

18

Table 1: Assessment of risk of bias tool

Risk of bias Definition Scoring (1-low risk; 2-medium risk; 3-high risk)

Selection Bias The sample is not representative of the population intended to be analysed.

Are the individuals selected to participate in the study likely to be representative of the target population? For survey studies in particular, a response rate of >80% is low risk, 60-79% is moderate risk and <60% or not reported response

rate is high risk. 1. Very likely 2. Somewhat likely 3. Not likely

Ecological

Fallacy

Logical fallacy inherent in

making causal inferences from group data to individual behaviors

The study…

1. …is non-ecological 3. … is ecological

Confounding Bias

The analysis does not account for those confounders that are

expected to have an effect on the outcome and that have not been accounted for by the inclusion of other

Were analyses appropriately adjusted for confounders? 1. For most confounders

2. For some confounders 3. No or can’t tell

Reporting Bias Inappropriate or inaccurate

reporting of aspects of the study

Are all aspects of the study (including aims, methods and

results) clearly described and reported? 1. All aspects of the study are clearly described and

clear models are provided to support the findings 2. Some aspects of the study are not clear 3. Certain aspects of the study are missing and it is

hard to interpret the results provided (full models are not provided)

Time Bias Bias introduced due to the timeframe considered

Were analyses appropriately considering a) more than 10 years of data b) at least 3 years post crisis and c) potential

lag effects? 1. All three factors are considered 2. Two of the three factors are considered 3. One or none of the three are considered

Measurement

Error in Exposure Variable

Errors related to the exposure

variable

When considering the economic crisis, are analyses

appropriately using… 1. More than one macro-economic variable 2. One macro-economic variable 3. No macro-economic variable, just the year of the

crisis

Measurement Error in Health Outcome

Errors related to the health outcome measurement

The health outcome is… 1. A clinical indicator or officially recorded data that

is not likely to be mis-reported or mis-stated 2. Self-reported validated tool

3. Self-reported tool but not clear if it was validated or official recorded data that are likely to be mis-stated (e.g. suicide in death certificate)

Overall Rating:

1. Strong: No “high risk” score

2. Moderate: up to two “high risk” score

3. Weak: two or more “high risk” score.

Page 18 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Only19

Table 2: Summary of characteristics for the 39 eligible studies

Authors Year Country Time period

considered Methods

Main

outcome

considered*

Aguilar-Palacio et al. 2015 Spain 2006, 2011/12 Chi square tests and multivariate logistic regressions SRH, MH

Astell-Burt and Feng 2013 UK 2006-2010 Multivariate logistic regressions MH

Barr et al. 2015 England 2004-2013 Segmented linear regression MH

Bartoll et al. 2014 Spain 2006, 2011/2012 Multivariate Poisson regressions MH

Bartoll et al. 2015 Spain 2001-2012 Multivariate linear regression SRH

Branas et al. 2015 Greece 1983 - 2012 Interrupted time-series analysis S

Chang et al. 2013 54 countries (27 European) 2000-2009 Time trend analysis S

Corcoran et al. 2015 Ireland 1980-2012 Interrupted time series analyses S

Coope et al. 2014 UK (England&Wales) 2001-2011 Joinpoint regression analysis S

Córdoba-Doña et al. 2014 Spain (Andalucía) 2003-2012 Negative binomial and fixed effects linear regressions S

Curl and Kearns 2015 Scotland (Glasgow) 2006, 2008, 2011 Binary logistic regressions and correlation analysis MH

De Vogli et al. 2014 Italy 2000-2010 Fixed effects linear regressions MH

Drydakis 2015 Greece 2008-2013 Fixed effect logit regressions SRH, MH

Economou et al. 2013 Greece 2008 & 2011 Correlation analysis and logistic regression MH

Eiríksdóttir et al. 2013 Iceland 2006-2009 Logistic regression analysis InfH

Eiríksdóttir et al. 2015 Iceland 2005-2012 Logistic regression analysis MatH

Ferrarini et al. 2014 23 European countries 2006 & 2009 Hierarchical logistic conditional change models SRH

Fountoulakis et al. 2014 29 European countries 2000-2011 Correlation analysis and random effects regressions S

Gili et al. 2013 Spain 2006/7 & 2010/11 Multivariate linear probability regressions and Levin’s formula

MH

Gotsens et al. 2015 Spain 2006 and 2012 Poisson regressions were used to obtain prevalence ratios SRH, MH

Hessel et al. 2014 Greece and Ireland 2006-2010 Logistic regressions with DID, using Poland as control SRH

Huijts et al. 2015 27 European countries 2007-2009 Linear regression models SRH

Katikireddi et al. 2012 UK (England) 1991-2010 Logistic regressions MH

Kontaxakis et al. 2013 Greece 2001-2011 Correlation analysis S

Laanani et al. 2015 8 West European countries 2000-2010 Quasi-Poisson regression and sensitivity analysis S

Lopez Bernal et al. 2013 Spain 2005-2010 Interrupted time-series analysis S

Madianos et al. 2014 Greece 1990-2011 Correlation analysis S

Malard et al. 2015 France 2006 & 2010 GEE for logistic regression MH

Rachiotis et al. 2015 Greece 2003-2012 Correlation and regression analysis S

Page 19 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Authors Year Country Time period

considered Methods

Main

outcome

considered*

Rajmil et al. 2013 Spain (Catalonia) 2006 & 2010/12 Before-after comparisons and multivariate regression analysis. MH, SRH,

Reeves et al. 2014 20 EU countries 1981-2011 Multivariate regressions S

Reeves et al. 2014 21 EU countries 1991-2012 Multivariate regressions and simulations TB

Regidor et al. 2014 Spain 1995-2011 Joinpoint regression and average annual percent change analysis

M, SRH

Reile et al. 2014 Estonia, Lithuania & Finland 2004-2010 Logistic regressions using Finland as a control. SRH

Saurina et al. 2013 England 1993-2010 Hierarchical mixed models S

Tapia Granados and Rodriguez 2015 Greece, Finland, Iceland 1990-2012 Multivariate regressions M, S

Vandoros et al. 2013 Greece 2006-2009 DID with Poland as control SRH

Zapata Moya et al. 2015 Spain 2003-2012 Logistic three-level analyses MH

Zavras et al. 2012 Greece 2006 & 2011 Correlation analysis and logistic regressions SRH

* (InfH: infant health; M: Mortality; MatH: maternal health; MH: mental health; S: suicides; SRH: self-rated health; TB: tuberculosis

Page 20 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table 3: Assessment of risk of bias for the 39 eligible studies

Paper (author, year) Selection

Bias

Ecological

Fallacy

Confounding

bias

Reporting

bias

Time

bias

Measurement

error in

exposure

variable

Measurement

error in

health

outcome

Overall

rating

Aguilar-Palacio et al., 2015 3 1 2 3 3 2 3 3

Astell-Burt and Feng, 2013 3 1 2 3 3 3 3 3

Barr et al., 2015 3 1 3 3 3 2 3 3

Bartoll et al., 2014 3 1 2 2 3 3 2 3

Bartoll et al., 2015 2 1 2 3 2 2 2 2

Branas et al., 2015 1 3 3 2 2 3 3 3

Chang et al., 2013 2 3 2 1 3 3 3 3

Coope et al., 2014 2 3 2 3 2 1 3 3

Corcoran et al., 2015 1 3 2 2 2 3 3 3

Córdoba-Doña et al., 2014 3 3 2 2 2 2 2 2

Curl and Kearns, 2015 3 1 3 3 2 2 2 3

De Vogli et al., 2014 1 3 3 3 2 1 3 3

Drydakis 2015 2 1 2 2 2 2 2 1

Economou et al., 2013 2 1 1 3 3 3 2 3

Eiríksdóttir et al., 2013 1 1 2 3 3 3 1 3

Eiríksdóttir et al., 2015 1 1 2 2 2 2 2 1

Ferrarini et al., 2014 2 3 2 3 3 3 3 3

Fountoulakis et al., 2014 2 3 3 3 2 1 3 3

Gili et al., 2013 2 1 2 3 3 1 2 2

Gotsens et al., 2015 3 1 2 3 3 3 2 3

Hessel et al., 2014 3 1 1 2 3 3 2 3

Huijts et al., 2015 1 1 1 2 3 2 2 2

Katikireddi et al., 2012 2 1 1 2 2 2 2 2

Page 21 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Kontaxakis et al., 2013 1 3 3 3 2 3 3 3

Laanani et al., 2015 3 3 1 3 2 2 3 3

Lopez Bernal et al., 2013 2 3 2 2 3 3 3 3

Madianos et al., 2014 1 3 3 3 2 1 3 3

Malard et al., 2015 2 1 2 1 3 3 2 3

Rachiotis et al., 2015 2 3 2 3 2 2 3 3

Rajmil et al., 2013 3 1 2 3 3 2 2 3

Reeves et al. (20 EU), 2014a 2 3 2 3 2 2 3 3

Reeves et al., 2014b 2 3 2 2 2 1 3 2

Regidor et al., 2014 2 3 3 3 2 3 1 3

Reile et al., 2014 3 1 2 2 3 3 3 3

Saurina et al., 2013 1 3 3 3 2 2 3 3

Tapia Granados &

Rodriguez, 2015 1 3 3 3 2 3 3 3

Vandoros et al., 2013 3 1 1 1 3 3 3 3

Zapata Moya et al., 2015 3 1 2 2 2 3 2 2

Zavras et al., 2012 3 1 1 2 3 2 2 2

Weighted kappa 0.67 0.82 0.64 0.59 0.69 0.59 0.74 0.75

Page 22 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


APPENDIX

Table A. PRISMA Checklist

Section/topic # Checklist item Reported on page #

TITLE

Title 1 Identify the report as a systematic review, meta-analysis, or both. 1

ABSTRACT

Structured summary 2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number.

2

INTRODUCTION

Rationale 3 Describe the rationale for the review in the context of what is already known. 3

Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS).

3

METHODS

Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.

N/A

Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered,

language, publication status) used as criteria for eligibility, giving rationale.

3-4

Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.

3

Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

Appendix C

Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable,

included in the meta-analysis).

4

Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

4

Page 23 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

3-4

Risk of bias in individual studies

12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

4-5

Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means). N/A

Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency

(e.g., I2

) for each meta-analysis.

N/A

Risk of bias across studies 15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies).

4-5

Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating

which were pre-specified.

N/A

RESULTS

Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

5

Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.

5 and Appendix B

Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12). 5

Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

6-11

Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency. N/A

Risk of bias across studies 22 Present results of any assessment of risk of bias across studies (see Item 15). 5

Additional analysis 23 Give results of additional analyses, if done (e.g., sensitivity or subgroup analyses, meta-regression [see Item 16]). N/A

DISCUSSION

Page 24 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

11-13

Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

12

Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research. 11

FUNDING

Funding 27 Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review.

14

From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097

Page 25 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table B: Summary of studies included in the review, by health outcome (N=39)

Authors Country Time

period

Crisis

Period

Exposure Data & population Methods Health

outcomes

Results on the impact of the

2008 financial crisis (notes

in italics)

Aguilar-Palacio et al.,

2015

Spain 2006 and 2011/12

2011/12 Time, employment

Cross-sectional Spanish National Health Surveys

(2006 and 2011/12) for 16–24 years old.

Chi square tests and

multivariate logistic regressions

SRH, diagnosed morbidity and

mental health.

After adjusting for age and stratified by gender in 2012

as compared to 2006: - Poor SRH reduced significant for women

(OR=0.52, CI 95% 0.38-0.71, p statistically significant but no actual level given). For men the results

were not significant. - Diagnosed morbidity for women reduced significant

(OR=0.55, CI 95% 0.45-0.67, p statistically significant but no actual level given). For men the results were not significant. - Mental health (GHQ-12) for women improved

(OR=0.61, CI 95% 0.47-0.79, p statistically significant but no actual level given). For men the results

were not significant. When the time of unemployment was considered in men a higher risk of mental illness was observed with long time of

Page 26 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

unemployment (OR=2.33; CI 95% 1.09-4.99) with respect to workers. Poor SRH

reduced in 2012 with respect to 2006 in men (OR 0.61, CI 95% 0.39-0.94) and in women (OR=0.51, CI 95%

0.37-0.70). Diagnosed morbidity improved for women (OR=0.54, CI 95% 0.44-0.67) and so did mental

health (OR=0.60, CI 95% 0.46-0.78). Note: these

results were significant but

the exact p value is not

reported.

Astell-Burt and Feng, 2013

UK 2006-2010

Post 2008 Unemployment Data from the UK Quarterly Labour Force Survey for 16-64 year olds.

Multivariate logistic regressions

Self-reported mental health, cardiovascular and respiratory problems.

As compared to the pre-crisis period (2006), Jan-Mar 2008 reported increased prevalence of cardiovascular conditions by 0.2%, depression by 0.2% and mental illness by 0.1%.

Fully-adjusted association between unemployment and each type of health problem was as follows (unclear if

these refer to post 2008

period only or they cover the

whole period): i) respiratory health problems OR: 1.20 (95% CI 1.16, 1.24);

ii) cardiovascular health

Page 27 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

problems OR: 1.05 (95% CI 1.01, 1.09);

iii) depression OR: 2.98

(95%CI 2.85, 3.10); and

iv) mental illness OR: 3.18

(95% CI 2.98, 3.38).

Barr et al.,

2015

England 2004-

2013

Post 2009 Unemployment Quarterly Labour Force

surveys (QLFS) that includes a rolling panel sample for 18-59 years old.

Segmented

linear regression

Mental health During 2004-2013, for each

1% increase in unemployment, the prevalence of mental health problems increases by 0.15% [95% CI 0.08 to 0.23] and for each £10 decline in median weekly wages, the prevalence of mental health

problems increased by 0.03% [95% CI 0.004 to 0.06]. However, only 36% of the increase in mental health

problems post 2009 could be explained by rising unemployment or declining wages.

Bartoll et al., 2014

Spain 2006 and 2011/2012

2011/2012 Time Cross-sectional Spanish National Health Surveys (2006 and 2011/12).

Multivariate Poisson regressions

Mental health The prevalence of poor mental health among men showed a 15% increase in 2011–2012 compared with that in 2006–2007 (PR =

1.15, 95%CI 1.04–1.26). This increase was relatively larger in men in the 35–44 (PR=1.24, 95%CI 1.04–1.47)

and 45–54 years age groups

Page 28 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

(PR=1.29, 95%CI 1.07-1.55), those in social class IV (PR=1.23, 95%CI 1.07-1.40),

those with primary or secondary education (PR=1.29, 95 %CI 1.02–1.62; PR=1.27, 95%CI 1.07–

1.49, respectively), foreigners (PR = 1.33, 95%CI 1.00–1.77) and breadwinners (PR=1.15,

95%CI 1.03–1.28). However, these associations disappeared after adjusting for age and working status.

Among women, the adjusted prevalence of poor mental health decreased overall (PR = 0.89, 95%CI 0.84–0.95, p<0.001), and was associated with Spain as the country of birth (p<0.01) and non-

breadwinner (p<0.001).

Bartoll et al., 2015

Spain 2001-2012

Post 2011/12

Time, employment

Cross-sectional Spanish National Health Surveys (2001, 2003/04, 2006/07

and 2011/12).

Multivariate linear regression

Self rated health The probability of good self-reported health increased for men by 7.6% (p<0.01) in

2011/12 and by 9.6% (p<0.01) for women. Employed men did better than unemployed but there was no effect of unemployment on women.

Branas et al., Greece 1983 - Post 2008 Time National suicide rates Interrupted Suicide rates In June 2011, overall suicide

Page 29 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

2015 2012 (monthly by sex) from ELSTAT.

time-series analysis.

rates increased by 35.7% (p<0.001) and among men by 18.5% (p<0.01). In May

2011, suicides increased for women by 35.8% (p<0.05).

Chang et al., 2013

54 countries (Including 27 European

countries)

2000-2009.

Data for 2010 available for some

countries.

Post 2008 Time Suicide rates for over 15 years old from the WHO Mortality Database. Data

for 2010 available for only 20 European countries.

Time trend analysis (comparing

actual number of suicides in 2009 with

expected number based on pre-

recession period).

Suicide rates For men, suicides increased in 2009 by 4.2% i.e. 2937 excess suicides (p<0.001),

with the highest increase in the 15-24 age group (11.7%, p<0.001). Out of 27 European countries, 24

showed an increase in male suicides (p<0.05), with Poland showing the largest increase in absolute number

of excess suicides (763, p<0.001). For women, there was no change.

Corcoran et al., 2015

Ireland Suicide deaths (1980-2012) and

self-harm suicide (2004-2012).

Post 2008 Time Suicide deaths from the Irish Central Statistics Office and self-harm presentation to hospital

from the Irish National Registry of Deliberate Self Harm for over 15 years old

Interrupted time series analyses (comparing

actual incidence in 2009 with expected

incidence based on pre-recession period).

Suicides rates and self-harm

The downward trend of declining suicide rates for men during 2000 to 2007 (-0.2 per 100,000 per quarter,

p<0.001) reversed in 2008 and suicide rates started increasing (0.3 per 100,000 per quarter, p=0.006). By end

of 2012, male suicide rate was 57% higher than if the pre-recession trend had continued (i.e. 476 excess suicide, p<0.001). . Men aged 25–64 years were

Page 30 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

affected by higher suicide with the greatest effect observed in 25–44 year-olds.

Rate for male self-harm was also 31% more (i.e. 5029 excess incidences of self-harm, p<0.05), mainly

among the 25-64 age groups. For women, there was no change in suicide rates but

self-harm rates were 22% higher than if the pre-recession trend had continued (i.e. 3833 excess

incidences of self-harm, p<0.05). The increase in self-harm by women was in 15–24 year-olds.

Coope et al., 2014

UK (England and Wales)

2001-2011

Post March 2008

Insolvencies, house repossessions, unemployment, redundancy.

Time

Suicide rates from the Office for National Statistics and economic data from the UK Insolvency Service,

Ministry of Justice and ONS Labour Force Survey for 16-64 years old.

Joinpoint regression analysis

Suicide rates For men, overall no change in the quarterly age-standardised suicide rates after the crisis. Declining suicide rates halted in 2006

for 16-24 and 25-34 years-old (p<0.05). Suicide rates increased for 35-44 age groups until 2008 (not

significant), declined during 2008-2010 (-2.1% quarterly change, p<0.05) and then increased after 2010 (+2.6%, p<0.05). For 45-54 and 55-64 age groups suicide rates steadily increased throughout 2001-2011(+ 0.5% and +

Page 31 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

0.4%, p<0.05) with no marked change after/during the recession.

For women, overall no change in suicide rates (results not shown).

Results for other exposure

variables are not clearly

reported.

Córdoba-

Doña et al., 2014

Spain

(Andalucía)

2003-

2012

Post 2008 Time

Unemployment

Data on suicide attempts

attended by 15-64 year old at emergency services from the Health Emergencies Public

Enterprise Information System (SIEPES)

Negative

binomial and fixed effects linear regressions.

Suicide attempts Assuming that the pre-

recession trends continued, during 2008-2012, there was an excess of 4989 (95%CI 1985-8013) suicide attempts,

2017 (95%CI 87-3987) in men and 2972 (95%CI 1878-4075) in women. Further, the increase in suicide attempts during 2008-2012 was associated with increased unemployment rates for men but not for

women.

Curl and Kearns 2015

Scotland (Glasgow)

2006, 2008 and 2011

Post 2008 Time Cross-sectional household surveys with a nested longitudinal cohort.

Binary logistic regressions and

correlation analysis.

Mental health Poor mental health associated with increased affordability difficulties throughout 2006-2011.

Overall, affordability difficulties static or declined overtime, but increased for at-risk groups. Mental health also worsened during the recession.

Page 32 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

(Results not reported clearly

to extract effect sizes)

De Vogli et al., 2014

Italy 2000-2010

(excluding 2004 and 2005)

Post 2008 Unemployment, GDP

Data on standardized death rates from Italian

Health for All database. Annual data on GDP and unemployment from the Italian Institute of

National Statistics.

Fixed effects linear

regressions (comparing actual rates with

expected based on pre-recession

period).

Mortality from mental health

disorders

Crisis resulted in an additional 0.303 per 100,000

deaths per year (95% CI: 0.192 – 0.0.478, p=0.001) i.e. 548 excess deaths due to mental and behavioural

disorders. Further, income losses during the recession was associated with 0.2% excess deaths (i.e. 123 deaths

and 22.4% of the excess deaths) and unemployment was associated with 0.15 per 100,000 deaths (i.e. 90

deaths and 16.4% of the excess deaths). .

Drydakis 2015

Greece 2008-2013

Post 2010 Unemployment Data from the Longitudinal Labour Market Study (LLMS) conducted via telephone with 18-65 years old in the workforce.

Fixed effect logit regressions.

SRH (possible

range of scores is

1-5, with higher

scores (5)

indicating poor

health) and

mental health

For the 2008-2013 period, results show that unemployed people face more impaired health than do employed people (3.21 versus 2.48, t=8.34,

p<0.001). In addition, unemployed people face more negative mental health symptoms than do employed

people (12.67 versus 9.39, t=12.28, p<0.001). Importantly, the health difference between unemployed and employed individuals is smaller

Page 33 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

(2.97-2.28=0.69) in the 2008- 2009 period than for the same group of individuals

during the 2010-2013 period (3.48-2.52=0.96). The assigned difference between the periods is statistically

significant (t=10.14, p=0.00).

There is a negative

unemployment effect on

health on the order of 0.53

percentage points (or 0.18%).

Women are affected more

than men.

Mental health: During 2008-

2013, unemployment was

associated with 3.2

percentage points (p<0.001)

increase in poor mental

health among men. This

association increased in the

post-crisis period from 3 to

4.9 percentage points

(p<0.001). Moreover, post-

crisis, unemployment due to

firm closures was associated

with poorer mental health

than compared to in 2008-

2009 (2.76 vs 4.16

percentage points, p<0.001).

Page 34 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

Among women,

unemployment increased

poor mental health from 4.3

to 7.3 percentage points

(p<0.001), and for

unemployment due to firm

closures, from 4.1 to 6.7

percentage points (p<0.001).

Hence, women’s mental

health was more adversely

affected by unemployment

than men’s.

Economou et

al., 2013

Greece 2008 and

2011

2011 Time Cross-sectional telephone

surveys with 18-69 years old.

Correlation

analysis and logistic regression.

Mental health One-month prevalence of

depression was higher in 2011 than in 2008: 8.2% vs 3.3% (p<0.0001). In terms of risk estimates, the odds of suffering from major depression was 2.6 times greater in 2011 than in 2008 (OR=2.6, 95%CI: 1.97-3.43).

Eiríksdóttir et al., 2013

Iceland 2006-2009

Post October 2008

Time Data from the National Icelandic Birth Registry on 16,271 women who

had live, singleton births during 2006-2009.

Logistic regression analysis.

Infant health Rates of infants born with low birth weight increased from 2.5% before the crisis

to 3% after the crisis. When controlling for age, parity and seasonality the increase is significant (aOR=1.25 95% CI [1.02, 1.53]. The increase is no longer significant when controlling for more mediating variables.

Page 35 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

There is no significant change in small gestational age and preterm births.

Eiríksdóttir

et al., 2015

Iceland 2005-

2012

Post

October 2008

Time,

unemployment rate

35,211 women whose

pregnancies resulted in live singleton births. Data on pregnancy-induced hypertensive disorders

from the Icelandic Birth Registry.

Logistic

regression analysis.

Maternity health Increased prevalence of

gestational hypertension in the first year following the economic collapse (2.4% vs 3.9%; aOR 1.47; 95% CI

1.13-1.91) but not in the subsequent years. The association disappeared when the authors adjusted for

aggregate unemployment rate. Similarly, there was an increase in prescription fills

of β-blockers in the first year following the collapse (1.9% vs.3.1%; aOR 1.43; 95% CI 1.07–1.90), which disappeared after adjusting for aggregate unemployment rate (aOR 1.05; 95% CI 0.72–1.54). No changes were

observed for preeclampsia or use of calcium channel blockers between the pre- and post-collapse periods.

Ferrarini et al., 2014

23 European countries

2006 and 2009

2009 Time Longitudinal panel SRH on 18-64 years old from EU Statistics on Income & Living Conditions and unemployment insurance data from Social Policy

Hierarchical logistic conditional change models

SRH Unemployment insurance reduced transitions into deteriorating SRH during recession.

Page 36 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

Indicator database.

Fountoulakis, Kawohl et al., 2014b


2000-2011

Varies per country and depends on

the exposure variable

Unemployment, GDP per capita, growth rate

Data from official national statistics agencies of countries.

Correlation analysis and random

effects regressions

Suicide rates Although correlation is evident, temporal relationships do not support a

direct link between the crisis and suicide rates. Overall, higher suicides

associated with higher unemployment and lower growth rate, not with GDP per capita.

For men, higher suicides associated with higher unemployment and lower

growth rate, not with GDP per capita. For women, suicides associated with only unemployment.

Gili et al., 2013

Spain 2006/7 and 2010/11

2010/11 Unemployment, mortgage, repayment

difficulties

Repeated cross-sectional surveys of patients attending primary care

centres.

Multivariate linear probability

regressions and Levin’s formula to calculate

population attributable risks.

Mental health (major and minor depression,

alcohol abuse and dysthymia)

Unemployment associated with major depressive disorders in both 2006

(OR=1.54, p<0.001) and 2010 (OR=1.72, p<0.001), with minor depressive disorders in 2010 (OR=1.20,

p<0.01) and dysthymia in 2006 (1.84, p<0.001). In 2010, family-level unemployment was strongly correlated to all mental

Page 37 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

health disorders and mortgage repayment difficulties with major

depression.

Gotsens et al., 2015

Spain 2006 and 2012

2012 Time Repeated cross-sectional Spanish National Health Surveys for 15-64 years old.

In 2006 23760 individuals were interviewed, while in 2012 there were 16 616.

Poisson regressions were used to obtain

prevalence ratios (PR) for each year.

SRH, mental health, chronic activity limitation and

psychotropic drugs

In both survey years, the immigrant population having arrived before 2006 presents worse self-rated health than

natives (men: PR2006 = 1.32, PR2012 = 1.28 and women: PR2006 = 1.39, PR2012 = 1.56). After adjustment for the

other variables, this probability attenuates in men but remains significant in women (No p values

reported). The probability of poor self-rated health in immigrant women, with respect to native women, was greater in 2012 than in 2006 (in all models), the interaction being significant when overcrowding (P value

= 0.02) and social support (P value = 0.021) are introduced.

Not significant for chronic activity limitation (PR2006 = 0.55; 95% CI: 0.39–0.77, PR2012 = 0.73; 95% CI: 0.49–1.10). For mental health, prevalence of poor mental health more

Page 38 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

among migrant men (PR2006 = 1.10; 95% CI: 0.86–1.40, PR2012 = 1.34; 95% CI: 1.06–

1.69). Not significant for women.

Hessel et al., 2014

Greece and Ireland

2006-2010

Post 2008 for Ireland Post 2009

for Greece

Time Five rounds of European Union Statistics of Income and Living

Conditions.

Logistic regressions with DID,

using Poland as control.

SRH While Poland witnessed a continuing decline in the odds of poor health after the

financial crisis, trends were significantly less favourable in Greece as indicted by the DID estimate that compared

health between 2006-2008 and 2009-2010, (OR=1.22, 95%CI 1.11, 1.33). In contrast, there was no

evidence that the financial crisis influenced health trends in Ireland. The DID estimate was 0.98 (95% CI 0.82, 1.17) for a comparison of health between 2006-2008 and 2009-2010 with respect to

the control population.

Huiits et al., 2015


2007-2009

Post 2008 Employment status

EU Survey on Income and Living Conditions (EU SILC)

Linear regression models

Self-reported health

Job loss during the crisis was associated with worse SRH in both men (b=0.12, 95%

CI: 0.09-0.15) and women (b=0.13, 95% CI: 0.10-0.16). Women who regained employment within a 1 year had similar health to those who did not lose jobs.

Page 39 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

Instead men who regained employment had an enduring health disadvantage

compared to those who had not lost jobs (b=0.11, 85% CI: 0.05-0.16)

Katikireddi et al., 2012

UK (England)

1991-2010

Post 2008 Unemployment Repeated cross-sectional Health Surveys for

England for working age (25–64 years) population.

Logistic regressions

Mental health For women, no statistical change in the prevalence of

poor mental health after 2008. For men, prevalence of poor

mental health increased by 5% in 2009 (p<0.001) and by 3% in 2010 (p=0.001) as compared to 2008, after

adjusting for age, employment & education status.

Kontaxakis et al., 2013

Greece 2001-2011

Post 2008 Time National suicide rates from ELSTAT for over 16-years old.

Correlation analysis

Suicide rates As compared to 2001-2007, overall suicide rate increased in the crisis period (from -3.9% to +27.2%). This increase was found significant for men (from -

8.4% to +26.9%, p=0.047) but not for women.

For men, suicide rates

increased during the crisis for age groups 30-34 (p=0.02), 45-49 (p=0.02) and 50-54 (p=0.006), and decreased for age group 60-64 (p=0.03).

Page 40 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

Results of statistical tests not

clearly reported.

Laanani et

al., 2015

8 Western

European countries (Austria, Finland,

France, Germany, Netherlands, Spain,

Sweden & UK)

2000-

2010

Post 2008 Time

Unemployment

Unemployment and

suicide rates by sex, age group, country and year from Eurostat dataset.

Quasi-

Poisson regression and sensitivity

analysis

Suicide rates Suicides increased in the

crisis period by 3.3% (p<0.001). This increase was significant in Germany (7.4% increase, p<0.001),

Netherlands (0.7%, p<0.01) and UK (7.4%, p<0.01). Assuming pre-crisis trend in

unemployment continued, unemployment variation during crisis accounted for 564 excess suicides in France

(CI 277-845), 57 in Netherlands (CI 9-104) and 456 in UK (CI 126-763). During 2000-2010, suicides increased by 0.3% for a 10% increase in unemployment (p<0.05). The increase was

significant in France (2%, p<0.001), Netherlands (0.7%, p<0.05) and UK (1%, p<0.05)..

Lopez Bernal et al., 2013

Spain 2005-2010

Post April 2008

Time Suicide rates from Spain's national statistics institute (INE)

Interrupted time-series analysis with stratification

Suicide rates Overall, suicide rates increased by 8% after the crisis (p=0,030).

Page 41 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

. Evidence of increase found in the Mediterranean area (Rate Ratio (RR) =1.09,

p=0.037) and among men (RR=1, p=0.007.

Madianos et al., 2014

Greece 1990-2011

Post 2007 Public debt, unemployment rates

Suicide and population data from the Hellenic Statistical Authority

(HSA) and Eurostat.

Correlation analysis

Suicide rates Unemployment associated with increased suicide rates.

Not correlation if recession variables is public debt

Malard et al., 2015

France 2006 and 2010

2010 Time Panel data from a prospective national survey of 20-74 year-old

French workers.

GEE for logistic regression

Mental health Overall no change in prevalence of Major Depressive Episode (MDE)

and Generalized Anxiety Disorder (GAD) in 2010. For women, GAD increased

by 7.4% (OR=1.74, 93%CI: 1.2-2.6) for only those working in the public sector. For men, overall no change. MDE increased by 2.3% (p=0.04) in 2010 for men in general population and

doubled (p=0.03) for men who became unemployed (excluding retirement).

Rachiotis et al., 2015

Greece 2003-2012

Post 2011 Time Unemployment

Suicide data from the Hellenic Statistical

Authority (HSA) and unemployment and GDP data from OECD.

Correlation and

regression analysis

Suicide rates Overall, suicide rate increased after 2010 by 35%

from 3.35 to 4.42 per 100,000 population (p<0.01). An estimated 2.23 rise (95% CI 1.37-3.10) was attributable to job loss linked

Page 42 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

to austerity, which largely accounted for the overall suicide increase in working-

age men. Among men suicides increased by 29% (from 5.75

to 7.43, p<0.01) and among women by 33% (from 1.17 to 1.55, p=0.03), with the highest increase among

working men aged 20-59 (from 6.58 to 8.81 (p<0.01). For women, the highest increase was among 20-59

year-old (from 1.37 to 1.84).

Rajmil et al., 2013

Spain (Catalonia)

2006 and 2010-12

2010-12 Time, employment status

Repeated cross-sectional Catalan Health Survey of children (<15 years). 2200 children participated in 2006 and 1967 in

2010-12.

Before-after comparisons and multivariate regression analysis.

Mental health, SRH, chronic condition, obesity, HRQOL

KS-10 (measuring HRQL) showed higher mean scores (better) in 2010–2012 (85.4; 84.4 to 86.0) compared with 2006 (81.0; 80.7 to 81.7), but lower scores in children with

a maternal primary education (82.4; 80.6 to 84.1) and unemployed families (83.34; 81.89 to 84.9).

Scores on TDS-SDQ (measuring mental health) were slightly lower (better) in 2010–2012, but differences have remained in relation to maternal education and employment

Page 43 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

status.

Reeves et al., 2014

20 EU countries

1981-2011

Post 2008 Unemployment Age-standardized suicide rates for men from WHO

Human Mortality Database and economic data from OECD and EuroStat.

Multivariate regressions

adjusting for pre-existing time trends and country

FEs.

Male suicide rates

Assuming the pre-recession linear trends had continued,

there were 6998 excess male suicides overall. Of these, 1077 (15%) were attributable to the rising unemployment.

Further, spending on active labour market programmes (ALMP) prevented 540 (50%) suicides and high

levels of social capital prevented an additional 210 (19%) suicides. (CIs, p-

values etc. not reported)

During 1981-2011, each percentage point rise in male unemployment was associated with a 0.94% rise (95% CI: 0.51-1.36, p<0.001) in suicide rate. This

association was primarily concentrated in the working-age men (25-64years), where each percentage point

increase in unemployment was associated with 1.39% (95%CI: 0.53-2.24) rise in suicides. No effect was found among over 65+ and under 16 years old. (It was reported

whether these trends

changed post 2008)

Page 44 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

Reeves et al., 2015

21 EU countries

1991-2012

Post 2007 Public health expenditure, GDP

Data on tuberculosis detection and treatment success rates from WHO

TB database and economic data from EuroStat.

Multivariate regressions and

simulations

Tuberculosis During 2007-2012 there was no significant association between social protection

spending and case detection of TB (-0.59 for a US$100 increase in social protection spending (-1.31 to 0.14,

p=0.1066).

Regidor et al., 2014

Spain 1995-2011

Post 2008 Time. GDP adjusted for purchasing power parity

(PPP) is presented but not accounted for in the analysis

Data on mortality for 0-74 years old, SRH for 16-74 years old, and incidence of diagnosed HIV

infections from national registries and National Health Survey.

Joinpoint regression and average annual

percent change analysis

Mortality rate, SRH and diagnosed HIV infections

Mortality from several causes fell during recession with the greatest decline in traffic injuries and illicit

drug-induced deaths. Prevalence of poor-SRH declined during recession by 5.7% (significance level not reported) HIV incidents were reduced

(-1%) but the effect was not significant.

Reile et al., 2014

Estonia, Lithuania and Finland

2004-2010

Post 2008 Time Data on 20-64 years-old from repeated cross-sections postal surveys.

Logistic regressions using

Finland as a control.

SRH During the period of economic crisis in 2008-2010, the prevalence of poor

health increased to 52% in Estonia and to 48% in Lithuania. Although the increase was not statistically significant, it marked the end

Page 45 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

of the previous positive trend of improving health status. A small and statistically

insignificant increase occurred in 2008-2010 in Finland.

For the 2008-2010 recession period, a slight and statistically non-significant increase was observed in the

overall prevalence of less-than-good health among men in all countries.

In 2008-2010, a small and non-significant increase in the prevalence of less-than-good health was found among women in Estonia and Lithuania, whereas the prevalence slightly decreased among Finnish women.

.

Saurina et al., 2013

England 1993-2010

Post 2008 Year Annual suicide data for ≥15 year olds from UK ONS and employment

data (2000-2010) for 16-74 year olds from Eurostat.

Hierarchical mixed models

Suicide rates County-level: no change in suicide rates after 2008 for both men and women.

Regional-level: For men, suicide rate (per 100,000 population) in North East (+1.8, CI 0.3-3.2) and North West (+2.0, CI 0.7-3.3) and decreased in East of England (-1.6, CI -2.7 to -0.6).. For

Page 46 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

women, suicide rate increased in North West (+0.5, CI 0.1-0.8) and

decreased in West Midlands (-0.3, CI -0.7to -0.02) and East of England (-0.5, CI -0.9 to -0.2)

Tapia

Granados and Rodriguez, 2015

Greece,

Finland, Iceland

1990-

2012

Post 2008 Time Latest data available on

life expectancy, mortality, incidence of infectious diseases, vaccination rates, SRH from WHO or

other reputed source.

Multivariate

regressions

All cause and

cause-specific mortality, suicide rate, HIV incidence and

other health indicators.

Most indicators of population

health continued improving during 2008-2012 as compared to 2003-2007, in the three countries including

life expectancy, mortality rates, TB incidence and mortality, cause-specific mortality (CVD, respiratory,

HIV, cancers, transport injuries and infectious diseases.. Suicide rates increased in Greece (p=0.044) but there was no change in Finland and Iceland. Mortality due to mental

diseases worsened in Finland (p=0.047) but remained unchanged in Greece and Iceland.

No effect sizes reported.

Vandoros et al., 2013

Greece (Poland as control)

2006-2009

Post 2009 Time Data from the EU Statistics on Income and Living Conditions Surveys

DID with Poland as control

SRH While Poland continued to experience declines in the odds of self-rated health after the financial crisis, Greece experienced significantly less favourable trends than

Page 47 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

Poland. For example, while the odds of poor self-rated health declined by 10% (OR,

0.91; 95% CI, 0.86–0.95) after the financial crisis in Poland, there was a significant positive

interaction between trend and country (OR, 1.16; 95% CI, 1.04–1.29), indicating less favourable health trends in

Greece as compared with Poland after the financial collapse. .

Page 48 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

Zapata Moya

et al., 2015

Spain 2003-

2012

2009-2011 Time

Mean real GDP growth rate and

change in low work intensity indicator

Three waves (2003-2004,

2006-2007, 2011-2012)

of the Spanish National

Health Survey (SNHS)

and the 2009-2010 wave

of the European Health

Survey in Spain (EHS-S).

Logistic

three-level

analyses

Depression,

diabetes,

myocardial

infarction and

malignant tumors

Among women, depression increased by 12% in 2009

(p<0.05) and by 23% in 2011 (p<0.001) as compared to 2003. Among men there was a 13% increase in depression

in 2011 (p<0.10). However, these effects disappeared after controlling for confounders.

Second, in 2011 women and men are more likely to have diabetes than in 2006

(respectively ORwomen = 1.14. p<0.01; ORmen =1.13, p<0.05) (no effect size reported). When introducing the macroeconomic context and change variables, these period effects are also no longer significant.

The probability of being diagnosed with a myocardial infarction decreases for men

from 2006 to 2011 (OR = 0.88, p<0.10) (no effect size reported) but the effect is no longer significant when introducing the macroeconomic context and change variables. The results are not significant for

women.

Page 49 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

The probability of being diagnosed with a malignant

tumor decreases for women from 2003 to 2011 (OR = 0.87, p<0.10) but the effect is no longer significant when

introducing the macroeconomic context and change variables. The results are not significant for men,

Zavras et al.,

2012

Greece 2006 and

2011

2011 Unemployment Repeated national cross-

sectional surveys conducted in 2006 (personal interviews) and 2011 (phone interviews).

Correlation

analysis and logistic regressions

SRH The overall prevalence of

good and very good SRH in 2006 was 71.0%, whereas in 2011 people with good and very good SRH accounted

for 68.8% (P< 0.05). Individuals with higher income [odds ratio (OR) 1.18], higher education (OR 1.48) and men (OR 1.31) have a higher probability of rating their health as good or

very good. On the other hand, findings for age (OR 0.87) and existence of chronic disease (OR

0.18) indicate that older individuals and those suffering from a chronic disease have a lower probability of rating their health as good or very good.

Page 50 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



period

Crisis

Period


outcomes



in italics)

Although unemployment

(OR 0.79) was marginally

statistically significant

(P=0.05, CI 0.63–0.99), this

finding indicates that the

unemployed were less likely

to report good health.

Page 51 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table C: Search String - PUBMED

Platform: PUBMED

Database: PUBMED

# STRING

1 Search (((europ*[Title/Abstract]) OR "health outcomes"[Title/Abstract]) OR health?care[Title/Abstract]) OR "public

health"[Title/Abstract]

2 Search ((("economic crisis"[Title/Abstract]) OR "financial crisis"[Title/Abstract]) OR austerity[Title/Abstract]) OR

recession[Title/Abstract]

3 Search (impact[Title/Abstract]) OR effect*[Title/Abstract]

4

Search ((((((europ*[Title/Abstract]) OR "health outcomes"[Title/Abstract]) OR health?care[Title/Abstract]) OR "public

health"[Title/Abstract])) AND (((("economic crisis"[Title/Abstract]) OR "financial crisis"[Title/Abstract]) OR

austerity[Title/Abstract]) OR recession[Title/Abstract])) AND ((impact[Title/Abstract]) OR effect*[Title/Abstract])

5

Search ((((((europ*[Title/Abstract]) OR "health outcomes"[Title/Abstract]) OR health?care[Title/Abstract]) OR "public

health"[Title/Abstract])) AND (((("economic crisis"[Title/Abstract]) OR "financial crisis"[Title/Abstract]) OR

austerity[Title/Abstract]) OR recession[Title/Abstract])) AND ((impact[Title/Abstract]) OR effect*[Title/Abstract]) Filters:

Publication date from 2008/01/01 to 2015/12/31

Page 52 of 52


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

impact of the european financial crisis on health outcomes: a … · 2016-09-07 · risk for bias,...

Documents