Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
UNEMPLOYMENT, GOVERNMENT HEALTHCARE SPENDING AND CEREBROVASCULAR MORTALITY, WORLDWIDE 1981-2009: AN ECOLOGICAL STUDY Dr Mahiben Maruthappu, MA BM BCh1,2 Dr Joseph Shalhoub, BSc MBBS MRCS FHEA PhD 1 Ms Zoon Tariq, BA 3 Mr Callum Williams, BA 3,4 Professor Thomas Zeltner, MD LLM 5,6 Professor Alun H Davies, MA DM FRCS FHEA FACPh FEBVS 1 Professor Rifat Atun, FRCP MBA, FFPH 1,2 1 Imperial College London, London SW7 2AZ, UK 2 3 4 5 Harvard University, MA 02138, USA University of Oxford, Oxford OX1 2JD, UK The Economist, 25 St James’s Street, London SW1A 1HG, UK Special Envoy for Financing to the Director General of the World Health Organization (WHO), Avenue Appia 20, 1211 Geneva 27, Switzerland 6 University of Bern, Gerechtigkeitsgasse 31, Bern, CH 3011, Switzerland 1 Please address correspondence to: Dr Mahiben Maruthappu Senior Fellow Chair & Chief Executive’s Office, NHS England Skipton House, 80 London Road London SE1 6LH UK [email protected] Financial Disclosure and Products Statement None of the participating authors has a conflicting financial interest related to the work detailed in this manuscript; nor do any of the authors maintain a financial stake in any product, device or drug cited in this report. We declare no support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous three years, and no other relationships or activities that could appear to have influenced the submitted work. Information on data used in this report is available upon request. 2 ABSTRACT Background: The global economic downturn has been associated with unemployment rises, reduced health spending and worsened population health. This has raised the question of how economic variations affect health outcomes. We sought to determine the effect of changes in unemployment and government healthcare expenditure on cerebrovascular mortality globally. Methods: Data were obtained from the World Bank and World Health Organization. Multivariate regression analysis was used to assess the effect of changes in unemployment and government healthcare expenditure on cerebrovascular mortality. Country-specific differences in infrastructure and demographics were controlled for. One to five year lag analyses and robustness checks were conducted. Findings: Across 99 countries worldwide, between 1981-2009, every 1% increase in unemployment was associated with a significant increase in cerebrovascular mortality (coefficient 187 deaths/100,000, CI: 86.6-288, p=0.0003). Every 1% rise in government healthcare expenditure, across both genders, was associated with significant decreases in cerebrovascular deaths (coefficient 869 deaths/100,000, CI: 383-1354, p=0.0005). The association between unemployment and cerebrovascular mortality remained statistically significant for at least five years subsequent to the 1% unemployment rise, while the association between government healthcare expenditure and cerebrovascular mortality remained significant for two years. These relationships were both shown to be independent of changes in gross 3 domestic product per capita, inflation, interest rates, urbanization, nutrition, education and outof-pocket spending. Interpretation: Rises in unemployment and reductions in government healthcare expenditure are associated with significant increases in cerebrovascular mortality globally. Clinicians may also need to consider unemployment as a possible risk factor for cerebrovascular disease mortality. Funding: None. Key words: cerebrovascular disease; stroke; mortality; unemployment; economic recession; austerity; government spending; public health expenditure. 4 INTRODUCTION Stroke is the second leading cause of global mortality, accounting for 6.7 million deaths in 2012(1). In upper-middle income countries, cerebrovascular events are the leading killer, responsible for 126 deaths per 100,000 population(1). Such statistics have been the impetus for public health strategies to reduce cerebrovascular mortality. A number of studies have reported an association between socioeconomic status and the development of cerebrovascular disease in addition to increased stroke mortality(2-7). In recent years, the role of economic changes in healthcare has been accentuated. The global population has faced an economic crisis, presenting unexpected socioeconomic challenges to healthcare. A number of governments have adopted policies aimed to reduce budget deficits through a series of tax rises and spending cuts, often directly leading to rising unemployment rates, and in the context of healthcare, reduced health system spending(8, 9). In countries where particularly aggressive fiscal austerity measures have been implemented – including Greece, Cyprus, Portugal and Spain – unemployment has risen sharply(10). Studies have attempted to evaluate the impact of economic changes on population health outcomes(11, 12), however these have yet to be translated to policy development. Further, these studies have also almost exclusively focused on all-cause mortality or suicide, rather than the mortality attributed to specific conditions. This raises the question of how economic variations, both within and outside of economic crises, affect disease-specific health outcomes, including cerebrovascular mortality. 5 We sought to determine whether there is an association between two key economic variables: unemployment and government healthcare expenditure, and cerebrovascular mortality, worldwide, between the years 1981 and 2009. We hypothesized that increased unemployment, and reduced government healthcare expenditure, would be associated with rises in cerebrovascular deaths, attributable to a variety of factors including reduced access to care. 6 METHODS Cerebrovascular mortality data for the years 1981 to 2009 were obtained from the WHO mortality database(13), which is updated annually from civil registration systems of member states. The quality of the data has been evaluated by the WHO(14). Cerebrovascular mortality was defined as annual deaths from cerebrovascular diseases (under the International Classification of Diseases (ICD) codes: ICD-9 430-438; ICD-10 I60-I69, G45) per 100,000 population(15). Age standardized death rates (ASDR) were employed as the basis of the analysis. The ASDR, as defined by the WHO, is the weighted average of age-specific mortality rate per 100,000 population, where the weights are the proportions of persons in the corresponding age groups of the WHO standard population(16). ASDR was selected as it controls for differences in age distribution within populations by incorporating age-specific mortality rates(16). Socioeconomic data were obtained from the World Bank’s Development Indicators and Global Development Finance 2013 edition for the period 1981 to 2009(10). Government debt, as a percentage of gross domestic product (GDP), data were obtained from the International Monetary Fund (IMF) Historical Public Debt Database(17). Data were available for 99 countries for each of the two analyses, namely unemployment and government healthcare expenditure (Table A, see Appendix). There was some variation in the 99 states included in each analysis. This was due to different data availability for unemployment and government healthcare expenditure respectively. Unemployment, as defined by the World Bank(18), was taken to be the share of the labour force that is without work but available and seeking employment. 7 Government healthcare expenditure was measured as a percentage of GDP; it was taken to include recurrent and capital spending from government budgets, external borrowings and grants, and social insurance funds(19). Multivariate regression analysis was used to separately assess the relationship between cerebrovascular mortality (dependent variable), and either unemployment (independent variable – Table B, see Appendix) or government healthcare expenditure (independent variable – Table C, see Appendix). To ensure that results were not driven by extreme observations for certain countries, a fixed-effects approach was used in the regression models, including 99 dummy variables for the 99 countries in each dataset. Doing this meant that models evaluated mortality changes within individual countries while holding constant time-invariant differences between countries, including higher predispositions to cerebrovascular mortality as well as political, healthcare, cultural, and structural differences. The demographic structure of the selected countries was controlled for by incorporating total population size, in addition to the percentage of the population over 65 years of age and less than 15 years old into the model. This is because country demographics can affect both cerebrovascular incidence and mortality rates. A CookWeisberg test(20) was used to assess for and to confirm heteroskedasticity (where sub-samples have different distributions) in the data used. Therefore, robust standard errors were included in the regression models; accounting for the heterogeneity in the unemployment or government expenditure datasets due to, for example, differences in the way that countries measured unemployment rates or government healthcare expenditure. Due to the inclusion of over 100 control variables (in turn losing degrees of freedom and reducing sample size), the approach was conservative. This methodology has been widely used in similar health-economic studies, and is 8 regarded as a statistically robust and conservative approach(12, 21-23). The basic linear fixed effects statistical model was therefore: ΔHi,t–ΔHi=α+β×(ΔIi,t–ΔUi) +η×t+εi,t Where i is country and t year; H is the dependent variable health metric (cerebrovascular mortality); I is the independent variable (either unemployment or public health spending); α represents the population structure of the country being analyzed, η is a dummy variable for each country included in the regression model, and ɛ is the error term. One, 2, 3, 4 and 5-year time-lag multivariate analyses were conducted to quantify the long-term effects of changes in unemployment on cerebrovascular mortality. Robustness checks were also conducted to control for economic factors, urbanisation and nutrition, education and out of pocket expenses. Stata SE version 12 was used for the analysis (Stata Corporation, Texas, USA). 9 RESULTS The results of the regression models are displayed in Tables 2 and 3 and are adjusted for population size, demographic structure and inter-country differences including higher predispositions to cerebrovascular mortality as well as political, healthcare, cultural, and structural differences. The results (Table 1) show that a 1% rise in unemployment was associated with a statistically significant rise in cerebrovascular mortality in both men and women (coefficient for men 76.8 deaths/100,000, CI: 28.8-125, p=0.0017; coefficient for women 111 deaths/100,000, CI: 55.2166, p=0.0001). Between genders, the effect was greater for women than for men. The results (Table 2) demonstrate that a 1% rise in government healthcare expenditure (measured as a percentage of GDP) was associated with a statistically significant decrease in cerebrovascular mortality in both men and women (coefficient for men -466 deaths/100,000, CI: -708 to -224, p=0.0002; coefficient for women -402 deaths/100,000, CI: -671 to -133, p=0.0034). Between genders, the effect was more pronounced for men than for women. Lag analysis Further analysis was performed to investigate whether this association lasted in the long-term. The results displayed in Table 1 show that following a 1% rise in unemployment, associated cerebrovascular mortality continued to increase, in both men and women, to a peak at 4 years (coefficient for both genders 208 deaths/100,000, CI: 98.4-317,p=0.0002, Figure 1). The 10 increased cerebrovascular mortality remained significantly raised at year five (coefficient for both genders 175 deaths/100,000, CI: 72.2-278,p=0.0009, Figure 1). The results displayed in Table 2 show that following a 1% rise in government healthcare expenditure, associated cerebrovascular mortality continued to fall significantly for two years in males, but for women the association with cerebrovascular mortality was limited to the year of the rise in government healthcare expenditure. Overall, across both genders, increases in government healthcare expenditure were associated with significantly decreased cerebrovascular mortality for two years after the expenditure change (coefficient -479 deaths/100,000, CI: -948 to -9.67, p=0.0455, Figure 2). Robustness checks In order to adjust for confounding factors, the analysis was re-run with additional economic and infrastructure controls included. Firstly, to control for economic factors, variables to account for changes to GDP per capita, inflation and interest rates were introduced. Secondly urbanisation and nutrition (mean calorie intake) were controlled for. Next, education (percentage of children with progression to secondary school), and finally out-of-pocket expenses were controlled for. Results are displayed in Table 3 and show that broadly, across both sexes, the association between unemployment and increased cerebrovascular mortality remained statistically significant when each of these variables were controlled for. This was also the case for government healthcare expenditure; the relationship between reduced spending and increased cerebrovascular mortality remained statistically significant when each of the economic, infrastructure, nutritional and educational variables were controlled for, with males, females, and both genders combined. 11 The significant associations described for both unemployment and cerebrovascular mortality, and government healthcare expenditure and cerebrovascular mortality, were maintained when analyses were re-run, controlling for changes in crude death rate. This suggests that the associations seen are specific for cerebrovascular deaths, beyond the increased mortality risk inherent to the unemployed and those in countries with reduced government healthcare spending. To ensure that the analysis had not been impacted upon by suboptimal data quality, a further robustness check was performed whereby the analysis was repeated using only high quality data (WHO level 1 and 2 surveillance data quality(24)). Again, the findings remained significant for both the unemployment and government healthcare expenditure analyses, and for both genders. 12 DISCUSSION Principal findings This study has shown that across 99 countries worldwide between 1981 and 2009, there is an association between unemployment and cerebrovascular mortality, and government healthcare expenditure and cerebrovascular mortality. The association between unemployment and cerebrovascular mortality remained statistically significant for at least five years subsequent to the 1% unemployment rise, whilst the overall link between government healthcare expenditure and cerebrovascular mortality remained significant for two years. These relationships were shown to be independent of GDP per capita changes, inflation, interest rates, urbanisation, nutrition and education. Causal mechanisms Although our analyses only demonstrate associations, there are a number of potential intermediate mechanisms by which unemployment may result in increased risk of stroke and, subsequently, a higher cerebrovascular mortality. These mechanisms include, but are not limited to, traditional cardiovascular risk factors: smoking(25), diet, psychological disturbances(26) – including depression(27)– hypertension, and poor diet(28) resulting in the metabolic syndrome(29). Recently, in addition to mean blood pressure, variability in blood pressure has been highlighted as an independent risk factor for stroke(30). There is consequently biological plausibility for links between unemployment, periods of stress, variations in blood pressure and cerebrovascular mortality. Such changes however occur in the long-term, possibly explaining the 13 results of the lag analysis, however they do not account for the short-term associations between unemployment and cerebrovascular mortality. In the short-term, unemployment and government healthcare expenditure are both likely to impair individuals’ access to care, specifically in terms of treatment of acute stroke with its respective sequelae. Consistent with this, higher income has been associated with higher rates of stroke unit admission, neurology consultations, referrals to secondary prevention clinics, and physician visits after hospital discharge(31). Congruently, low socioeconomic status has been associated with a lower chance of receiving optimal acute stroke care(32). In this study, unemployment had a greater effect on women, whilst government healthcare expenditure was more relevant to males. The cause of this is likely to be a complex interplay between biological, social and economic factors. Gender differences have previously been identified in the context of cerebrovascular mortality. In a US study investigating the association between employment status and mortality among 7361 middle-aged women, after adjusting for sociodemographic factors and selected risk factors for mortality, employed women had a lower risk of mortality than homemakers, including circulatory system-related deaths(33). Policy implications Policy implications can be considered at a system level, or at the level of individuals.This work furthers existing evidence of the association between government healthcare expenditure and health outcomes, reinforcing the importance of government spending in minimizing cerebrovascular mortality. This is of particular significance given the recent economic crisis 14 where many governments have implemented austerity measures; either due to government policy, such as in the UK, or under compulsion as a result of accepting financial assistance packages following problems with bound monetary policy, such as in Greece and Ireland(34). Austerity measures have in turn been associated with reduced public spending(34). The findings of this study therefore suggest that such policy interventions may be exacerbating the adverse health effects of the global economic downturn, specifically with regard to cerebrovascular mortality, rather than ameliorating them. Outside of economic crises, the findings of this study have relevance in debates over cost-control in healthcare and times when governments attempt to reduce healthcare spending in favour of alternative initiatives. The results suggest that caution must be taken with regard to cost-control and budget restrictions of healthcare; if cost reductions are not achieved as a result of efficiency improvements, they may entail worse quality care and in turn greater cerebrovascular mortality. From the perspective of unemployment, the association between increased unemployment and worsened cerebrovascular mortality reinforces arguments against austerity measures from a public health perspective, which have been associated with markedly increased unemployment rates(10). Further, this work supports policies promoting return-to-work, social welfare and those that prevent further job losses. This is of particular importance as forecasts predict that in some countries unemployment may only return to pre-recession levels in 2017(35). For clinicians, it is important to recognise that macro-level multinational policy can affect mortality for specific conditions at the individual patient level, affecting day-to-day clinical practice. Clinicians should be aware that unemployment could be a risk factor for 15 cerebrovascular mortality. Furthermore, unemployed patients should receive particular attention to their traditional cardiovascular risk factors. Limitations Our study has several limitations that should be considered: firstly, public health outcomes and economic trends were examined at a national and international level, without consideration of local discrepancies; such trends however, are likely of relevance to multinational policy development by agencies such as the IMF, World Bank, and WHO. Second, information regarding the distribution of unemployment throughout different social classes could not be incorporated; this is notable as economic changes have been shown to impact those of lower socioeconomic status the most. Thirdly, whilst a number of controls were included in robustness checks, healthcare factors – such as access to primary care – were not incorporated into the regression models. Additionally, although this study highlights important associations, it does not demonstrate causality. Potential confounding factors linking socioeconomic indices and cardiovascular diseases, including stroke, are stress, depression (especially when untreated)(27), sleep patterns, cigarette smoking(25) and alcohol consumption(36), environmental factors, and obesity(29). Furthermore, low socioeconomic status and psychological distress have been demonstrated as synergistic predictors of stroke mortality(26). Sufficiently high-resolutiondata for these variables were not available for integration into the multivariate statistical models. This may have resulted in individual-level residual confounding. Our fixed-effects approach, although removing time-invariant differences between countries, did not control for time-varying confounding at the country level. Moreover, there are limitations regarding the quality of the multinational cerebrovascular mortality data acquired from the WHO, as previous studies have 16 shown both over- and under-diagnosis of acute stroke as a cause of death in some countries(37). Finally, with regard to government healthcare spending, it is worth noting that we did not account for changes in efficiency; indeed, it is feasible that a country spends less on healthcare but achieves greater outcomes due to the efficiency of its system. Nevertheless, our study used worldwide data, taken from high-quality, objective, and centralized databases, avoiding selection and recall bias. The volume of data used allowed for high statistical powering and multiple robustness checks. By focusing on countries at a global level over a 30year period, this study permitted consideration of macroscopic trends. Notably, our study used a conservative, fixed-effects multivariate regression analysis model. This model along with the multiple robustness checks addresses many criticisms levelled at other studies looking at the relationship between health outcomes and economic changes, including the omission of potential confounders. By using this panel-data approach, we also controlled for inter-country timeinvariant heterogeneity; something that aggregate, time-series analyses fail to do. Further all data used are publicly accessible, supporting the reproducibility of our study. Conclusions This study has shown that unemployment and public spending on healthcare are both associated with significant changes in cerebrovascular mortality worldwide. Current policy responses to the economic crisis, which may increase unemployment and decrease public spending, are of concern, and may present additional barriers to the management of cerebrovascular disease, possibly increasing cerebrovascular mortality. At a policy level, initiatives that bolster employment and maintain public expenditure may decreaselikelihood of cerebrovascular 17 mortality. Clinicians should also consider unemployment as a possible risk factor when assessing patients in the context of cerebrovascular disease. 18 REFERENCES 1. World Health Organization. The top 10 causes of death. World Health Organization; 2014 [updated May 2014; Fact sheet N°310; cited 19/09/2014]; Available from: http://who.int/mediacentre/factsheets/fs310/en/. 2. Hanchate AD, Schwamm LH, Huang W, Hylek EM. Comparison of ischemic stroke outcomes and patient and hospital characteristics by race/ethnicity and socioeconomic status. Stroke2013;44(2):469-76. 3. Chen R, McKevitt C, Rudd AG, Wolfe CD. Socioeconomic deprivation and survival after stroke: findings from the prospective South London stroke register of 1995 to 2011. Stroke 2014;45(1):217-23. 4. Brown AF, Liang LJ, Vassar SD, et al. Neighborhood socioeconomic disadvantage and mortality after stroke. Neurology 2013;80(6):520-7. 5. Lindmark A, Glader EL, Asplund K, Norrving B, Eriksson M. Socioeconomic disparities in stroke case fatality - Observations from Riks-Stroke, the Swedish stroke register. Int J Stroke 2013. 6. Sergeev AV. Stroke mortality disparities in the population of the Appalachian Mountain region. Ethn Dis 2013;23(3):286-91. 7. Jaja BN, Saposnik G, Nisenbaum R, et al. Effect of socioeconomic status on inpatient mortality and use of postacute care after subarachnoid hemorrhage. Stroke; a journal of cerebral circulation 2013;44(10):2842-7. 8. Stuckler D, Basu S. The Body Economic: Why Austerity Kills. First ed. New York: Basic Books; 2013. 19 9. Martin A, Lassman D, Whittle L, Catlin A. Recession contributes to slowest annual rate of increase in health spending in five decades. Health Aff (Millwood) 2011;30(1):11-22. 10. World Bank. Employment. 2014 [updated 2014; cited 19/09/2014]; Available from: http://datatopics.worldbank.org/jobs/topic/employment. 11. Stuckler D, Meissner CM, King LP. Can a bank crisis break your heart? Global Health 2008;4:1. 12. Stuckler D, Basu S, Suhrcke M, Coutts A, McKee M. The public health effect of economic crises and alternative policy responses in Europe: an empirical analysis. Lancet 2009;374(9686):315-23. 13. World Health Organization. WHO Mortality Database. 2012 [updated 2012; cited 19/09/2014]; Available from: http://www.who.int/healthinfo/mortality_data/en/. 14. Mathers CD, Fat DM, Inoue M, Rao C, Lopez AD. Counting the dead and what they died from: an assessment of the global status of cause of death data. Bulletin of the World Health Organization 2005;83(3):171-7. 15. World Health Organization. International Classification of Diseases (ICD) [cited 19/09/2014]; Available from: http://www.who.int/classifications/icd/en/. 16. World Health Organisation. Indicator and Measurement Registry (version 1.7.0) [cited 19/09/2014]; Available from: http://www.who.int/gho/indicator_registry/en/. 17. International Monetary Fund. Historical Public Debt Database [cited 19/09/2014]; Available from: http://www.worldbank.org/qpsd. 18. World Bank. Umemployment, total [cited 19/09/2014]; Available from: http://data.worldbank.org/indicator/SL.UEM.TOTL.ZS. 19. World Bank [cited 19/09/2014]; Available from: http://data.worldbank.org. 20 20. Cook RD, Weisberg S. Diagnostics for heteroscedasticity in regression. Biometrika 1983;70(1):1-10. 21. Davidson R, Mackinnon JG. Estimation and inference in econometrics. New York ; Oxford: Oxford University Press; 1993. 22. Stuckler D, King L, McKee M. Mass privatisation and the post-communist mortality crisis: a cross-national analysis. Lancet 2009;373(9661):399-407. 23. Gerdtham UG, Ruhm CJ. Deaths rise in good economic times: evidence from the OECD. Econ Hum Biol 2006;4(3):298-316. 24. World Health Organization. Mortality and burden of disease estimates for WHO Member States in 2004. Geneva; 2009 [updated 2009; cited 19/09/2014]; Available from: http://www.who.int/healthinfo/global_burden_disease/estimates_country/en/. 25. Shinton R, Beevers G. Meta-analysis of relation between cigarette smoking and stroke. BMJ 1989;298(6676):789-94. 26. Lazzarino AI, Hamer M, Stamatakis E, Steptoe A. Low socioeconomic status and psychological distress as synergistic predictors of mortality from stroke and coronary heart disease. Psychosom Med 2013;75(3):311-6. 27. Rahman I, Humphreys K, Bennet AM, Ingelsson E, Pedersen NL, Magnusson PK. Clinical depression, antidepressant use and risk of future cardiovascular disease. Eur J Epidemiol 2013;28(7):589-95. 28. Mejean C, Droomers M, van der Schouw YT, et al. The contribution of diet and lifestyle to socioeconomic inequalities in cardiovascular morbidity and mortality. Int J Cardiol 2013;168(6):5190-5. 21 29. Xu G, Ma M, Liu X, Hankey GJ. Is there a stroke belt in China and why? Stroke; a journal of cerebral circulation 2013;44(7):1775-83. 30. Rothwell PM, Howard SC, Dolan E, et al. Prognostic significance of visit-to-visit variability, maximum systolic blood pressure, and episodic hypertension. Lancet 2010;375(9718):895-905. 31. Huang K, Khan N, Kwan A, Fang J, Yun L, Kapral MK. Socioeconomic status and care after stroke: results from the Registry of the Canadian Stroke Network. Stroke; a journal of cerebral circulation 2013;44(2):477-82. 32. Langagergaard V, Palnum KH, Mehnert F, et al. Socioeconomic differences in quality of care and clinical outcome after stroke: a nationwide population-based study. Stroke; a journal of cerebral circulation 2011;42(10):2896-902. 33. Rose KM, Carson AP, Catellier D, et al. Women's employment status and mortality: the atherosclerosis risk in communities study. J Womens Health (Larchmt) 2004;13(10):1108-18. 34. Karanikolos M, Mladovsky P, Cylus J, et al. Financial crisis, austerity, and health in Europe. Lancet 2013;381(9874):1323-31. 35. Office for Budget Responsibility. Economic and fiscal outlook. 2011 [updated 2011; cited 19/09/2014]; Available from: http://budgetresponsibility.org.uk/economic-and-fiscaloutlook-march-2011/. 36. Brenner MH. Relation of economic change to Swedish health and social well-being, 1950-1980. Soc Sci Med 1987;25(2):183-95. 37. Asplund K, Bonita R, Kuulasmaa K, et al. Multinational comparisons of stroke epidemiology. Evaluation of case ascertainment in the WHO MONICA Stroke Study. World 22 Health Organization Monitoring Trends and Determinants in Cardiovascular Disease. Stroke 1995;26(3):355-60. 23 TABLES TABLE 1: Multiple regression and lag analysis examining the association between unemployment and cerebrovascular mortality The association between a 1% annual increase in unemployment and cerebrovascular mortality, controlling for population size, population structure (proportion of population below 14 years of age, proportion of population above 65 years of age) and controlling for inter-country differences in healthcare infrastructure in addition to political, cultural and structural differences (by introducing dummy variables for each of the 99 countries). Number of years after Cerebrovascular Mortality Cerebrovascular Mortality Cerebrovascular Mortality per 100,000 Population – Males per 100,000 Population – Females per 100,000 Population – Total 1% rise in unemployment Co-efficient P value Lower Upper confidence confidence interval interval Co-efficient P value Lower Upper confidence confidence interval interval Co-efficient P value Lower Upper confidence confidence interval interval Year 0 76.8017 0.0017 28.7677 124.8357 110.6915 0.0001 55.2027 166.1804 187.4933 0.0003 86.6132 288.3733 Year 1 73.5142 0.0000 38.8693 108.1590 113.0801 0.0000 65.6415 160.5188 186.5943 0.0000 107.3886 265.8000 Year 2 76.3933 0.0000 42.9474 109.8391 114.9787 0.0000 63.0931 166.8644 191.3720 0.0000 107.9746 274.7694 Year 3 83.2844 0.0001 42.2016 124.3671 119.3565 0.0001 61.6646 177.0484 202.6409 0.0000 105.7971 299.4846 Year 4 89.5167 0.0003 41.5960 137.4374 118.3769 0.0003 54.4908 182.2629 207.8936 0.0002 98.3657 317.4214 Year 5 77.4850 0.0013 30.3998 124.5702 97.5798 0.0012 38.7557 156.4040 175.0648 0.0009 72.2281 277.9015 (year of change in unemployment) 24 TABLE 2: Multiple regression and lag analysis examining the association between government healthcare expenditure and cerebrovascular mortality The association between a 1% annual increase in government healthcare expenditure (measured as a percentage of gross domestic product) and cerebrovascular mortality, controlling for population size, population structure (proportion of population below 14 years of age, proportion of population above 65 years of age) and controlling for inter-country differences in healthcare infrastructure in addition to political, cultural and structural differences (by introducing dummy variables for each of the 99 countries). Number of years after Cerebrovascular Mortality Cerebrovascular Mortality Cerebrovascular Mortality per 100,000 Population – Males per 100,000 Population – Females per 100,000 Population – Total 1% rise in public health spending Co-efficient P value Lower Upper confidence confidence interval interval Co-efficient P value Lower Upper confidence confidence interval interval Co-efficient P value Lower Upper confidence confidence interval interval Year 0 -466.3814 0.0002 -708.3203 -224.4424 -402.1509 0.0034 -670.8491 -133.4527 -868.5323 0.0005 -1,354.0851 -382.9795 Year 1 -424.4079 0.0009 -673.7293 -175.0865 -276.8732 0.0579 -563.0360 9.2896 -701.2811 0.0055 -1,195.6212 -206.9410 Year 2 -338.0197 0.0014 -544.8691 -131.1704 -141.2928 0.3638 -446.4759 163.8902 -479.3126 0.0455 -948.9538 -9.6714 Year 3 -191.2689 0.0985 -418.2631 35.7254 102.3362 0.6186 -300.9594 505.6317 -88.9327 0.7755 -700.8505 522.9851 Year 4 -47.5893 0.7587 -351.5746 256.3960 348.7724 0.2280 -218.7609 916.3058 301.1831 0.4913 -557.5444 1,159.9107 Year 5 176.1068 0.4212 -253.5085 605.7221 633.9594 0.1011 -124.1933 1,392.1121 810.0662 0.1789 -372.1113 1,992.2438 (year of change in unemployment) 25 TABLE 3: Robustness checks Multiple regression analyses were re-run using the controls in the original analysis (population size, proportion of population above 65 years of age, proportion below 15 years of age, and 99 country controls), in addition to those mentioned in the tables below, for males (A), females (B), and both males and females (C). The data shows the association between a 1% rise in unemployment (above) and a 1% rise in government healthcare expenditure (below) with cerebrovascular mortality, using the mentioned controls. A. MALES Unemployment Analysis Total number of Robustness check Controls used in multiple regression controls in Coefficient P value regression Economic controls Original analysis controls and: changes in GDP per capita, inflation, interest rates Infrastructure Original analysis controls and: urbanisation, nutrition (mean calorie controls intake) Lower Upper confidence confidence interval interval 105 88.2482 0.0014 34.0193 142.4772 104 46.8630 0.0110 10.7440 82.9820 107 47.4382 0.0334 3.7398 91.1367 Economic and Original analysis controls and: GDP per capita, inflation, interest rates, infrastructure controls urbanisation, nutrition (mean calorie intake) Crude death rate Original analysis controls and: crude death rate 103 80.8182 0.0009 33.1875 128.4489 Original analysis controls 102 80.2045 0.0021 29.1685 131.2404 Lower Upper P value confidence confidence interval interval WHO data quality check Public Health Spending Analysis Total number of Robustness check Controls used in multiple regression controls in Coefficient regression Economic controls Original analysis controls and: changes in GDP per capita, inflation, interest rates 105 -343.8167 0.0155 -622.0187 -65.6147 104 -559.1992 0.0003 -858.9808 -259.4177 107 -489.3470 0.0083 -852.5544 -126.1395 Infrastructure Original analysis controls and: urbanisation, nutrition (mean calorie controls intake) Economic and Original analysis controls and: GDP per capita, inflation, interest rates, infrastructure controls urbanisation, nutrition (mean calorie intake) Crude death rate Original analysis controls and: crude death rate 103 -303.3863 0.0038 -508.4560 -98.3165 Original analysis controls 102 -493.3691 0.0002 -752.8275 -233.9106 WHO data quality check 26 B. FEMALES Unemployment Analysis Total number of Robustness check Controls used in multiple regression controls in Coefficient P value regression Economic controls Original analysis controls and: changes in GDP per capita, inflation, interest rates Lower Upper confidence confidence interval interval 105 125.1431 0.0001 62.5004 187.7857 104 75.4905 0.0016 28.6402 122.3408 107 76.7294 0.0034 25.3724 128.0864 Infrastructure Original analysis controls and: urbanisation, nutrition (mean calorie controls intake) Economic and Original analysis controls and: GDP per capita, inflation, interest rates, infrastructure controls urbanisation, nutrition (mean calorie intake) Crude death rate Original analysis controls and: crude death rate 103 116.2260 0.0000 61.6872 170.7648 Original analysis controls 102 118.4280 0.0001 59.3504 177.5056 Lower Upper P value confidence confidence interval interval WHO data quality check Public Health Spending Analysis Total number of Robustness check Controls used in multiple regression controls in Coefficient regression Economic controls Original analysis controls and: changes in GDP per capita, inflation, interest rates Infrastructure Original analysis controls and: urbanisation, nutrition (mean calorie controls intake) 105 -278.5776 0.0367 -539.9429 -17.2124 104 -565.4826 0.0007 -890.1044 -240.8608 107 -517.6799 0.0010 -825.9609 -209.3989 Economic and Original analysis controls and: GDP per capita, inflation, interest rates, infrastructure controls urbanisation, nutrition (mean calorie intake) Crude death rate Original analysis controls and: crude death rate 103 -191.7352 0.1381 -445.2797 61.8093 Original analysis controls 102 -421.3254 0.0044 -710.7370 -131.9137 WHO data quality check 27 C. BOTH MALES & FEMALES Unemployment Analysis Total number of Robustness check Controls used in multiple regression controls in Coefficient P value regression Economic controls Original analysis controls and: changes in GDP per capita, inflation, interest rates Lower Upper confidence confidence interval interval 105 213.3913 0.0003 98.703 328.0791 104 122.3535 0.0021 44.5702 200.1368 107 124.1677 0.0073 33.5950 214.7403 Infrastructure Original analysis controls and: urbanisation, nutrition (mean calorie controls intake) Economic and Original analysis controls and: GDP per capita, inflation, interest rates, infrastructure controls urbanisation, nutrition (mean calorie intake) Crude death rate Original analysis controls and: crude death rate 103 197.0442 0.0001 97.5528 296.5355 Original analysis controls 102 198.6325 0.0003 91.4299 305.8351 Lower Upper P value confidence confidence interval interval WHO data quality check Public Health Spending Analysis Total number of Robustness check Controls used in multiple regression controls in Coefficient regression Economic controls Original analysis controls and: changes in GDP per capita, inflation, interest rates Infrastructure Original analysis controls and: urbanisation, nutrition (mean calorie controls intake) 105 -622.3943 0.0193 -1,143.3612 -101.4275 104 -1,124.6818 0.0002 -1,712.9558 -536.4078 107 -1,007.0269 0.0025 -1,658.9321 -355.1217 Economic and Original analysis controls and: GDP per capita, inflation, interest rates, infrastructure controls urbanisation, nutrition (mean calorie intake) Crude death rate Original analysis controls and: crude death rate 103 -495.1215 0.0257 -930.0318 -60.2111 Original analysis controls 102 -914.6944 0.0006 -1,436.7352 -392.6537 WHO data quality check 28 FIGURE LEGENDS Additional Cerebrovascular Deaths per 100,000 Population 220 200 **** *** **** *** **** *** 180 160 140 120 100 80 60 40 20 0 ar Ye 0 r1 a Ye r2 a Ye r3 a Ye ar Ye 4 r5 a Ye Year Since 1% Increase in Unemployment FIGURE 1: Multiple regression and lag analysis examining the association between unemployment and cerebrovascular mortality Results for both genders for change in cerebrovascular mortality (per 100,000 population) for 5 years following a 1% increase in unemployment (at year 0). ***p<0.001; ****p<0.0001. 29 NS NS NS * ** Ye ar 5 Ye ar 4 Ye ar 3 Ye ar 2 Ye ar 1 *** Ye ar 0 Additional Cerebrovascular Deaths per 100,000 Population 900 800 700 600 500 400 300 200 100 0 -100 -200 -300 -400 -500 -600 -700 -800 -900 -1000 Year Since 1% Increase in Public Health Spending FIGURE 2: Multiple regression and lag analysis examining the association between government healthcare expenditure and cerebrovascular mortality Results for both genders for change in cerebrovascular mortality (per 100,000 population) for 5 years following a 1% increase in public health spending (at year 0). *p<0.05; **p<0.01; ***p<0.001; NS, non-significant. 30