Socioeconomic inequalities in burden of disease due to traffic noise in the Nordic countries Anette Kocbach Bølling 1 Norwegian Institute of Public Health Department of Air Quality and Noise, Lovisenberggata 8, 0456 Oslo, Norway Centre for Disease Burden, Zander Kaaes gate 7, 5015 Bergen, Norway Jesse Daniel Thacher 2 Danish Cancer Society Research Center Diet, Genes and Environment, Strandboulevarden 49, 2100 Copenhagen, Denmark Søren Toksvig Klitkou 3 Norwegian Institute of Public Health Centre for Disease Burden, Zander Kaaes gate 7, 5015 Bergen, Norway Carl Michael Baravelli 4 Norwegian Institute of Public Health Centre for Disease Burden, Zander Kaaes gate 7, 5015 Bergen, Norway Eva M. Andersson 5 University of Gothenburg Occupational and Environmental Medicine, School of Public Health and Community Medicine, Institute of Medicine, Sahlgrenska Academy, Medicinaregatan 16A, 405 30, Gothenburg, Sweden

Sahlgrenska University Hospital Department of Occupational and Environmental Medicine, Medicinaregatan 16A, 405 30, Gothenburg, Swe- den Leo Stockfelt 6 University of Gothenburg Occupational and Environmental Medicine, School of Public Health and Community Medicine, Institute of Medicine, Sahlgrenska Academy, Medicinaregatan 16A, 405 30, Gothenburg, Sweden

Sahlgrenska University Hospital Department of Occupational and Environmental Medicine, Medicinaregatan 16A, 405 30, Gothenburg, Swe- den Natalia Vincens 7 University of Gothenburg Occupational and Environmental Medicine, School of Public Health and Community Medicine, Institute of Medicine, Sahlgrenska Academy, Medicinaregatan 18A, 41390 Gothenburg, Sweden

1 Anette.Kocbach@fhi.no 2 jesse@cancer.dk 3 SorenToksvig.Klitkou@fhi.no 4 CarlMichael.Baravelli@fhi.no 5 eva.m.andersson@amm.gu.se 6 leo.stockfelt@amm.gu.se 7 natalia.vincens@amm.gu.se

Mette Sørensen 8 Danish Cancer Society Research Center Diet, Genes and Environment, Strandboulevarden 49, 2100 Copenhagen, Denmark Anu Turunen 9 Finnish Institute for Health and Welfare Environmental Health Unit, P.O.Box 95, FI-70701 Kuopio, Finland Tarja Yli-Tuomi 10 Finnish Institute for Health and Welfare Environmental Health Unit, P.O.Box 95, FI-70701 Kuopio, Finland Mikael Ögren 11 University of Gothenburg Occupational and Environmental Medicine, School of Public Health and Community Medicine, Institute of Medicine, Sahlgrenska Academy, Medicinaregatan 16A, 405 30, Gothenburg, Sweden

Sahlgrenska University Hospital Department of Occupational and Environmental Medicine, Medicinaregatan 16A, 405 30, Gothenburg, Swe- den Virpi Kollanus 12 Finnish Institute for Health and Welfare Environmental Health Unit, P.O.Box 95, FI-70701 Kuopio, Finland Timo Lanki 13 Finnish Institute for Health and Welfare Environmental Health Unit, P.O.Box 95, FI-70701 Kuopio, Finland

University of Eastern Finland Yliopistonranta 1, 70210 Kuopio, Finland. Jenny Selander 14 Karolinska Institutet Institute of Environmental Medicine, Nobels väg 13, 171 77 Stockholm, Sweden Gerhard Sulo 15 Norwegian Institute of Public Health Centre for Disease Burden, Zander Kaaes gate 7, 5015 Bergen, Norway Gunn Marit Aasvang 16 Norwegian Institute of Public Health Department of Air Quality and Noise, Lovisenberggata 8, 0456 Oslo, Norway Centre for Disease Burden, Zander Kaaes gate 7, 5015 Bergen, Norway

8 mettes@CANCER.DK 9 anu.turunen@thl.fi 10 tarja.yli-tuomi@thl.fi 11 mikael.ogren@amm.gu.se 12 virpi.kollanus@thl.fi 13 timo.lanki@thl.fi 14 Jenny.Selander@ki.se 15 Gerhard.Sulo@fhi.no 16 GunnMarit.Aasvang@fhi.no

ABSTRACT Environmental noise is the second largest environmental risk factor in disease burden estimates for Europe. While socioeconomic inequalities in noise exposure have been reported, the impact of soci- oeconomic status (SES) on the disease burden attributable to noise exposure has to our knowledge not been reported previously. The aim of this study is to assess the impact of SES on traffic noise exposure and the associated disease burden in selected Nordic populations. We have employed na- tionwide data on road traffic noise exposure and SES from the Danish Nationwide Model, and will also include the Norwegian Nationwide Model and cohorts from all the Nordic countries in future analyses. The impact of household income, education and type of housing on traffic noise exposure was assessed using linear regression analyses. The most consistent finding was that, compared to medium and low income, high income was associated with lower noise exposure. Based on stratified exposure distributions according to income (low, medium and high) burden of disease estimates were calculated in terms of Years Lived with Disability (YLD) for high degree of noise annoyance. The YLD for high noise annoyance was up to 20% lower in the high compared to the lower income pop- ulations.

1. INTRODUCTION

Exposure to environmental noise is associated with increased risk of several negative health out- comes, including annoyance and sleep disturbance, cognitive impairment in children and cardiovas- cular and metabolic effects [1]. According to the European Environment Agency, environmental noise is the second largest environmental risk factor in Europe in terms of burden of disease estimates, and road traffic noise accounts for a large proportion of this burden [1-2].

Socioeconomic inequalities in exposure to a number of environmental risk factors like noise, air pollution and extreme temperatures have been reported [3]. Several studies have found associations between socioeconomic status (SES) or different indexes of deprivation, combining different measures of SES, and increased noise exposure [4-5]. However, this is not necessarily a universal pattern and there are some notable exceptions; for instance, many city center locations attract wealth- ier residents although the noise exposure is high [4].

The impact of SES on the disease burden attributable to noise exposure has to our knowledge not been reported previously. The aim of this study is to assess the impact of SES on traffic noise expo- sure and the associated disease burden in selected Nordic populations. Only data for Denmark are presented here, and high degree of annoyance was the only health outcome included in these prelim- inary analyses. Further work will include populations and cohort data from the other Nordic countries, as well as additional health outcomes associated with noise.

2. METHODS

2.1. Noise exposure and study areas Road traffic noise exposure (L den ) was estimated at the most exposed façades of all residential build- ings in Denmark for 2014 as described previously [6]. In brief, noise calculations were conducted utilizing the Nordic prediction method for road traffic noise using SoundPLAN. Information on all residential addresses in Denmark was obtained from the Building and Housing Registry.

A selection of study areas were included, reflecting different levels of diversity in socioeconomic status; Copenhagen municipality, Greater Copenhagen (including 9 municipalities adjacent to Co- penhagen; i.e. the same geographical area included for Copenhagen in the noise mapping according

to the Environmental Noise Directive, 2002/49/EC (END)) and entire Denmark. For each geograph- ical area, individuals > 18y were included in the analysis. 2.2. Indicators of SES Data from Statistics Denmark (DST) were used to define individual level socioeconomic variables. For education, three categories were included: low (mandatory), medium (secondary or vocational) and high (medium or long), reflecting the highest attained education. For income, the family equiva- lent disposable income was used as defined by DST: family disposable income/(0.5+(0.5 x no. of persons >14y) + (0.3 x no. of persons <15y)). Three categories were included in the current analysis based on quintiles of family income; low (lowest quintile), medium (three middle quintiles) and high (highest quintile). 2.3. Linear regression analysis In addition to road traffic noise, the linear regression analysis included sex, age, education, income and type of housing. The selection of included variables was based on an a priori DAG (directed acyclic graph). The variable categories were as follows (with reference in parenthesis): male and female (female); 10y age categories (0-10y); low, medium and high education (low); low, medium and high income (low); single house, semi-detached house and apartment (apartment). Road traffic noise exposure (L den ) was included as a continuous variable from 35 dB, and all noise levels below were set to 35 dB. The analysis was performed in SAS 9.4. 2.4. Estimation of burden of disease Burden of disease (BoD) due to road traffic noise was assessed in terms high de gree of annoyance (HA) in these preliminary analyses, using the number of years lost due to ill-health/disability (YLD) . The exposure-response function from the WHO systematic review was used to calculate the percent- age of highly annoyed (%HA i ) for a certain noise level L den,i and exposure category i (equation 1) in 1 dB exposure categories [7]. We calculated the total number of HA in each population, based on the number of exposed in each exposure category (N i ) and the percentage HA (%HA i ) and then summa- rized over all noise categories above 55 dB (equation 2). Finally, the YLD was calculated by multi- plying the total number of HA with the disability weight for HA (DW HA =0.02) recommended by WHO [8].

2 (1)

%HA = 78.972 −3.1162 ∗𝐿 𝑑𝑒𝑛,𝑖 + 0.0342 ∗𝐿 𝑑𝑒𝑛,𝑖

𝑁 𝐻𝐴 = ෍𝑁 𝑖 ∗%𝐻𝐴 𝑖 (2)

𝑌𝐿𝐷 𝐻𝐴 = 𝑁 𝐻𝐴 𝐷𝑊 𝐻𝐴 (3)

Burden of disease was assessed for the entire population and populations stratified according to so- cioeconomic factors for each geographical area. The YLD estimates are reported as rates (YLDs/100 000) since the population sizes differ between the stratified populations and between the different geographical areas. Additional health outcomes, including high degree of sleep disturbance, ischemic heart disease, diabetes and stroke will be included in the final analyses.

3. RESULTS

3.1. Noise exposure and SES In the linear regression analysis of the full model the explanatory power was relatively low in all the geographical areas, with R 2 values below 0.1 (Table 1). Type of housing had the highest impact on noise exposure. The highest estimates for housing were observed for entire Denmark, suggesting up to 5 dB lower exposure for individuals living in single houses and semi-detached houses compared to apartments.

With regard to the socioeconomic indicators, household income showed the most consistent pat- tern for the analyzed Danish populations. High and medium income was associated with lower noise exposure than low household income, with decreasing estimates for increasing levels of income in the analysis of Copenhagen and Greater Copenhagen, but not for Denmark. In contrast, high and medium education was in most instances associated with higher noise exposure, although without a consistent trend across the education levels. Moreover, being male was associated with higher expo- sures, while there was no clear trend in the noise exposure for the different age groups (data not shown). Table 1: Results from linear regression analysis . The table shows the explanatory power of the model (R 2 ), the number of observations the analysis was based on (N), and the estimates for the intercept as well as for a selection of variables; the household income, education and housing categories relative to the respective ref- erence categories (see methods section). Note that age and sex were also included in the model, but estimates are not displayed. In the table * denotes p < 0.05 while ** denotes p < 0.001.

Area R 2 N Intercept Income Education Housing

Med. High Med. High Single Semi

Copenhagen 0.056 427 708 61.17 -0.20 ** -0.53 ** 0.24 ** 0.07 * -5.05 ** -3.80 **

Greater CPH 0.075 803 758 60.61 -0.22 ** -0.56 ** 0.29 ** 0.18 ** -3.53 ** -3.40 **

Denmark 0.096 3 899 580 58.74 -0.30 ** 0.09 ** 0.26 ** 0.37 ** -5.31 ** -5.04 **

3.2. Noise exposure in stratified populations Based on the results of the linear regression analysis, income was chosen as the indicator of SES to generate stratified noise exposure distributions for the different populations. The normalized expo- sure distributions were shifted towards lower noise levels for the populations with higher income compared to the distributions for the low-income populations for Copenhagen (Figure 1). A similar shift was observed for the other geographical areas (data not shown).

Figure 1: Noise exposure distributions for populations stratified according to income for Copenhagen. The distributions are normalized relative to the population size within each stratified population to allow for comparison of the relative differences in the exposure profiles. The dotted line indicates 55 dB, which is the cut-off used in the calculation of YLD. 3.3. Burden of disease in total versus stratified populations The burden of disease estimates, in terms of YLD for high annoyance (HA) per 100 000 inhabitants, were higher for Copenhagen and Greater Copenhagen than for entire Denmark (Figure 2). For the populations stratified according to level of income, the YLD per 100 000 decreased for increasing income for all the geographical areas. The rates for YLD due to high noise annoyance were up to 20% lower in the high compared to the lower household income populations.

Figure 2: YLD for high annoyance (HA) attributable to road traffic noise for total and stratified popu- lations. The figure shows YLD/100 000 for the total populations and the populations stratified with regard to income, for each geographical area. The percentages above the columns reflect the reductions in YLD/100 000 relative to the low-income population.

4. DISCUSSION

Several studies report that reduced socioeconomic status (SES) has been associated with increased exposure to environmental noise [4-5]. In line with this, we observed that exposure to road traffic noise was lower in populations with medium and high income as compared to low income, with a trend for increasing effect estimates (more negative) for higher income in Copenhagen and Greater Copenhagen. In contrast, medium and high education was generally associated with increased noise exposure, suggesting a more complex relationship between noise exposure and these different indi- cators of SES. Inclusion of an interaction term in the linear regression analysis, or use of a deprivation index combining multiple SES factors, will be useful tools in the further analysis of the association between SES and traffic noise exposure.

The highest estimates were observed for type of housing in the linear regression analysis, with estimates for single houses and semi-detached houses being up to 5 dB lower than for apartments [4]. In larger cities, where noise levels are high, the proportion of apartments is also high, and this may contribute to a high noise level for apartments. In contrast, noise levels are lower in the countryside, where single or semi-detached houses are more common. Moreover, all dwelling units in apartment buildings will be assigned the same noise level in the highest exposed façade approach, which could potentially contribute to an over-estimation of the noise exposure estimates assigned to apartments. Type of housing is often pointed out as a factor associated with noise exposure, as more affluent residents may be more likely to afford single or semi-detached houses as well as better-constructed housing. While housing conditions (e.g. need for renovation) or social housing has been included in several studies regarding environmental noise and SES, we could not find any studies assessing as- sociations between different house types on noise exposure [4-5].

To study the impact of SES on the disease burden attributable to noise exposure in the Nordic countries, we analyzed populations stratified according to low, medium and high income in three geographical areas in Denmark. The YLD for high degree of noise annoyance was up to 20% lower in the high compared to the lower income populations. Moreover, the impact of income was very similar across the three geographical areas, showing a trend for lower YLD/100 000 in the high than medium and low-income populations. This was in contrast to the linear regression analysis, where a consistent trend across the income levels was only observed for Copenhagen and Greater Copenha- gen. In light of the complex relationship between noise exposure and the different SES indicators, a more sophisticated analysis is warranted in further assessments of the impact of SES on the disease burden attributable to noise. Possible approaches include using a deprivation index combining mul- tiple SES factors to stratify populations or use of weighted noise exposure distributions based on the linear regression analyses.

In contrast to the considerable differences in YLD estimates between the income groups, the linear regression analysis resulted in estimates smaller than -0.6 dB for both medium and high income com- pared to the low-income population. This discrepancy was probably due to the cut-off at L den 55 dB used in the YLD calculation, since a considerable proportion of the population was shifted from above to below the cut-off in the high and medium compared to the low-income population. Nevertheless, the results demonstrate that a relatively small estimate in the linear regression analysis might result in a considerable impact in the estimated burden for income-stratified populations.

The presented YLD estimates should be interpreted with caution, as they represent preliminary analyses including only high degree of annoyance and a single measure of SES. Further work will include populations and cohort data from the other Nordic countries, additional health outcomes as- sociated with noise, as well as a more sophisticated assessment of the impact of SES on the burden attributable to noise and inclusion of uncertainty estimates.

5. CONCLUSIONS

To our knowledge, this is the first attempt to assess the impact of SES on the burden of disease attributable to environmental noise exposure. Our preliminary results demonstrate that social inequal- ities in exposure to road traffic noise may translate to social differences in the attributable burden in the Danish population. Further analyses are required to assess how different SES factors affect the burden of disease attributable to environmental noise exposure, and to examine possible differences across the Nordic countries. 6. ACKNOWLEDGEMENTS

We gratefully acknowledge NordForsk for funding the NordSOUND study (grant number 83597). 7. REFERENCES

1. World Health Organization, Environmental noise guidelines for the European region , WHO Re-

gional Office for Europe, Copenhagen (2018). 2. European Environment Agency, Healthy environment, healthy lives: how the environment influ-

ences health and well-being in Europe . EEA Report NO 21/2019. Luxembourg: Publications Of- fice of the European Union (2020). 3. European Environment Agency, Unequal exposure and unequal impacts: social vulnerability to

air pollution, noise and extreme temperatures in Europe . EEA Report NO 22/2018. Luxembourg: Publications Office of the European Union (2018). 4. Science for Environment Policy, Links between noise and air pollution and socioeconomic status .

In-depth Report 13 produced for the European Commission, DG Environment by the Science Communication Unit, UWE, Bristol (2016). 5. Dreger S, Schüle SA, Hilz LK, Bolte G. Social Inequalities in Environmental Noise Exposure: A

Review of Evidence in the WHO European Region. International Journal of Environmental Re- search and Public Health, 16(6) , (2019). 6. Thacher, J. D., Poulsen, A. H., Raaschou-Nielsen, O., Jensen, A., Hillig, K., Roswall, N.,

Hvidtfeldt, U., Jensen. S. S., Levin, G., Valenci, V. H., Sorensen, M., High-resolution assessment of road traffic noise exposure in Denmark. Environmental Research 182 (2020). 7. Guski, R., D. Schreckenberg, and R. Schuemer. WHO environmental noise guidelines for the

European region: A systematic review on environmental noise and annoyance. International Journal of Environmental Research and Public Health , 14(12) (2017). 8. WHO, Burden of disease from environmental noise. Quantification of healthy life years lost in

Europe. WHO Regional office for Europe: Copenhagen, Denmark (2011).