Road traffic noise exposure and depressed
mood. A geographically detailed analysis of
Amsterdam, the Netherlands.
Joost Schuurman
Bachelorscriptie Sociale Geografie en Planologie
Begeleider: Els Veldhuizen
Studentnummer: 11296054
Inleverdatum: 1 juli 2019
Introduction
This study aims to uncover the relationship between road traffic noise exposure and
depressed mood in Amsterdam. Extensive research has already been executed on this subject yet produces ambiguous results. The subject of road traffic noise causing depressed mood has been researched for Amsterdam by Leijssen et al. (2018). Their results showed for example, that noise exposure above 70 dB corresponds with higher chances of depressed mood and noise exposure between 60 and 64 dB corresponds with lower chances of depressed mood. Concurrently, they found a clear connection between noise exposure and depressed mood. This study continues where Leijssen et al. (2018) left off by replicating their analysis before adding multiple distinctions.
First of all, Leijssen et al. (2018) used both personal characteristics and neighbourhood characteristics related to socioeconomic status as confounders in their statistical analysis. Their neighbourhood characteristics had the rather large postcode4 (pc4) scale, which is similar to the size of a large neighbourhood. In this study, neighbourhood characteristics are produced for a postcode6 (pc6) scale which contains around 20 addresses per pc6 area meaning that characteristics can be expressed in a much more accurate and detailed manner.
The second distinction made in this study is the addition of more recent road traffic noise exposure. Both the 2011 noise exposure data used in Leijssen et al. (2018) and noise exposure data from 2017 are included in the analysis. At the start of this research, it was intended to produce data on noise exposure through GIS methods. This was done for 2017 as this was the only publicly available map on the subject. Further on in the research process, it became possible to include the noise data used in Leijssen et al., (2018). This was delayed because that noise data belonged to a third party. Interesting differences were found between the two datasets, which can only be partially explained by difference in measuring methods. This meant that something else was causing the variation. This unknown variation was not easily explained, which led the research in the direction of comparing over time, by doing an analysis of developments in Amsterdam concerning road traffic noise between 2011 and 2017. Adding more geographically detailed data and a comparison over time is what gives this study its additional value.
The first part of the analysis in this study consists of a partial replication of the Leijssen et al. (2018) results. The second part of the analysis is dedicated to the addition of more detailed confounders concerning respondents’ living environments and exploring their correlations with both noise exposure and depressed mood. This includes comparing the effects of living environment characteristics based on 100 meter and 300 meter spatial buffers. Finally, a comparison is made between noise exposure in 2011 and 2017. The use of GIS software is required for the data collection of the living environment characteristics and the noise data from 2017. Multilevel logistic regression analysis is performed in STATA. Results are presented in odds ratios with a 95% Confidence Interval.
Theory
The effects of noise pollution, defined as sounds that are either unwanted or deemed
unnecessary (Goines & Hagler, 2007), on physical and mental health have been increasingly researched the last five years (Janssen & Hong, 2017). Concerns are growing because it is believed that current global urbanization patterns will bring about larger urban areas which will lead to more people living close to sources of noise, specifically road traffic noise. This has led the World Health Organization (2011) to call road traffic noise pollution the second most problematic nuisance for the health and well-being of those exposed, following air pollution. Air and noise pollution go hand in hand which doubles impacts on health for those living close to their sources.
Goines and Hagler (2007) have made an extensive overview called ‘Noise Pollution, A Modern Plague. In this piece, the many adverse health effects of noise have been
categorized as following: hearing impairment, interference with spoken communication, sleep disturbances, cardiovascular disturbances, disturbances in mental health, impaired task performance and negative social behavior and annoyance reactions. This study will focus on depression, which is in the mental health category. It is important to note that there is also an indirect link to depression. Noise annoyance by itself can have negative effects on mental health (Ouis, 2001), and specifically on depression (Beutel et al., 2016), making exposure to loud noise a double-edged sword. Additionally, noise pollution can cause stress, which in turn increases the risk of depressive symptoms (Goines & Hagler, 2007; Lupien et al., 2009). Lastly, noise pollution can cause insomnia, which has been proven to increase the chance of someone being depressed (Riemann & Voderholzer, 2003).
This study focuses on road traffic noise. Research results on the relationship between road traffic noise and mental health are not fully consistent but lean towards there being a relation between the two (Ouis, 2001; Sørensen et al., 2011; Bodin et al., 2009; Roswall et al., 2015). The same goes for the relationship between road traffic noise and depression. As far as studies could be found, only one paper contradicted the established connections
between road noise and depression. This is the paper written by Stansfeld et al. (1996) where no significant association was found between noise exposure and depressive symptoms. It stands alone against studies from Frankfurt, the Essen region and Gotheborg, all of them finding connections between noise pollution and either depression or depressive symptoms (Seidler et al., 2017; Öhrström, 1991; Orban et al., 2015).
Leijssen et al. (2018) studied the same subject for the Amsterdam municipality. They included demographic, ethnic and socioeconomic characteristics of the individual as confounders to get a more precise image of the relationship between road traffic noise and depression. Several characteristics of the living environment (income, liveability, green and blue space) were also established for the respondents to explore its impact on the
aforementioned relationship. The personal characteristics interacted much stronger than the living environment characteristics with both depressed mood and traffic noise. Although it is logical for one’s personal situation to be the most important influence on one’s mental health, it was surprising to see the factors from the living environment interact as little as they did. Leijssen et al. (2018) calculated the living environment confounders by using the Dutch 4-digit postcodes (pc4), which can be compared to the scale of a large neighbourhood. Using a smaller scale of measurement might bring about much interaction with depressed mood as the living environment is more realistically represented. 6-digit postcode areas (pc6), containing around 20 addresses usually, are used to establish more accurate values on the following characteristics of the living environment: average property value, percentage of people living on a minimal income, green space and blue space. These are established by calculating the
average values within a spatial buffer around the pc6 areas. This will produce a more accurate image of the relation between road traffic noise pollution and depressed mood as the living environment is represented in a more exact manner.
The focus on the living environment is based on plenty of research producing
evidence for neighbourhood effects and their interaction with health (Pickett & Pearl, 2001). The authors warn, though, that these findings might often be a false representation of
personal characteristics. For example, people with a low socioeconomic status often live in a neighbourhood with a low socioeconomic status (SES). The variables used to represent neighbourhood SES will correlate strongly with health effects, but when the same personal SES variables are added, the neighbourhood effects can become negligible. Pickett & Pearl (2001) do provide multiple studies where neighbourhood effects were still found even after controlling for personal SES. So, to be able to say anything about the effect of either the personal or environmental characteristics, multiple variables on both need to be included as confounders. This study uses multiple variables reflecting a respondents’ personal
socioeconomic status (ethnicity, highest completed education, occupation and marital status) and two variables reflecting the socioeconomic status of the living environment (average property value and percentage of people living on a minimal income).
More specifically, the relationship between the socioeconomic status of the living environment and depression is an ambiguous subject among academics. One study found that residents of a poverty area had twice as big a chance of reporting depressed feelings as residents in a non-poverty area. Yet, when they controlled for income, education, race, chronic conditions, smoking, BMI and alcohol consumption, ‘the poverty effect’ turned out to be insignificant (Yen & Kaplan, 1999). This tells us that the SES of the living environment does have an effect, but that this effect is inherently connected to the SES of the individual as low SES individuals live in low SES neighbourhoods. Another study, however, claims that neighbourhood residential instability significantly increases the likelihood of depression because of what they call neighbourhood chronic stress (Matheson et al., 2006). They claim that the daily stress of living in neighbourhoods with higher material deprivation and residential mobility is directly associated with depression. Furthermore, a study found a correlation between the quality of the built environment and depression (Galea et al., 2005). Besides the SES of the living environment, there is reason to believe that living near green space reduces the risk of depression (Cohen-Cline et al., 2015; Miles et al., 2012). It is believed that its presence plays a protective role on mental health (Gascon et al., 2018). The authors attribute this to the hypothesis that direct contact with green environments increases stress recovery and attention. Besides this, green spaces function as a calm, silent retreat because exposure to environmental nuisances such as noise pollution is generally much lower. Because of its clear connection with depression, green space in the living environment is also considered in this study. Blue space is also added because of Amsterdam’s’ important connection with water. I believe that in this city, blue space functions in the same way as green space, offering people calm areas with a connection to nature.
In short, there is no full consensus on the effects of the living environment. It is, therefore, important to accurately measure characteristics of the living environment. Its confounding effect on the relationship between noise pollution and depression will then be more precise.
Methodology
This cross-sectional study tests the statistical correlation of road traffic noise exposure, personal characteristics and characteristics of the living environment on depressed mood of residents of the Amsterdam municipality. Road traffic noise exposure in 2011 is gathered from the The Netherlands Environmental Assessment Agency and is measured in A-weighted decibels (PBL, 2011). Road traffic noise exposure in 2017 is based on noise maps from the National Institute for Public Health and the Environment (Atlas Leefomgeving, 2017). Noise exposure is expressed in A-weighted decibels (dB (A)). The following personal
characteristics are included: age, sex, marital status, highest completed education, occupation and ethnicity. Characteristics of the living environment include: average property value, percentage of residents living on a minimal income, percentage of green space and percentage of blue space. Multilevel logistic regression analysis is used to establish the relationship between noise pollution and depression accurately. The hypothesis is that a smaller scale calculation of the living environment will decrease the strength of this relationship. This is based on the assumption that the living environment influences depression, and because of the more detailed way the characteristics of the living
environment are measured here, correlations are stronger and therefore the influence of road traffic noise is weaker. The total study population is 21.733 residents of the Amsterdam municipality.
The following research question is answered. Main question
Is there an association between road traffic noise exposure and depressed mood in Amsterdam?
Sub questions
Is there an association between road traffic noise exposure and depressed mood in Amsterdam after confounding for personal characteristics?
Is there an association between road traffic noise exposure and depressed mood in Amsterdam after confounding for personal and living environment characteristics? Does this association change between noise exposure in 2011 and in 2017?
Data collection
Road traffic noise from 2011 is gathered from the The Netherlands Environmental Assessment Agency (PBL, 2011). This agency used the EMPARA noise tool to establish average noise levels for each address through a raster map (cell size: 25m by 25m) (PBL, 2008). Since it is impossible to physically measure noise exposure for each address, this noise tool calculates noise levels by considering, among other things, which kinds of vehicles drive on a road, how intensely a road is used and how the produced noise interacts with barriers such as actual noise barriers, but also buildings. Although this is fairly accurate, full precision can not be expected because the amount of noise will always be influenced by more factors than can be included in such a model. The common Lden (Lday-evening-night) measure is used in this study. Its value is based on noise exposure within a 24-hour period, which
averages the values of the day, evening and night, but adds 5dB to evening noise and 10dB to night noise. This is due to the fact that noise in the evening and night is considered more harmful and disturbing (Atlas Leefomgeving, 2016). The noise level for each pc6 area is calculated by averaging the noise levels of all addresses in each pc6 area. Furthermore, Noise
was categorized as follows: 45–54 dB(A), 55–59 dB(A), 60–64 dB(A), 65–69 dB(A), ≥70 dB(A). The Netherlands Environmental Assessment Agency provided ‘A-weighted’ (A) dB values. As humans are more sensitive to lower frequency noise, it is important to weigh this more heavily than high frequency noise. A-weighting does this and thus provides a sound measure that is more relevant for humans (PBL, 2008).
Road traffic noise from 2017 is based on the Lden road traffic noise map that was published by the National Institute for Public Health and the Environment (cell size: 10m by 10m) (Atlas Leefomgeving, 2017). GIS software QGIS and ArcGIS are used to perform the following actions. The noise exposure levels on a pc6 area scale are calculated by overlaying the Atlas Leefomgeving (2017) road traffic noise map (polygon) and a map of the center of each 6-digit postcode area (point). Each point will extract the decibel value according to its geographical location. The noise maps from Atlas Leefomgeving (2017) are established in accordance to the Dutch national calculative prescriptions on noise (RMV2012). Both the RMV2012 prescriptions and the EMPARA noise tool base their calculations on the Standaard Rekenmethode 1 (SRM1), with the difference being that the EMPARA noise tool includes noise protection from barriers and other objects like buildings. Thus, noise exposure in both years can be compared, keeping in mind the small difference in calculation methods.
HELIUS (Healthy Life in an Urban Setting) provides both the data on personal characteristics and our dependant variable depressed mood. HELIUS is an Amsterdam based collaboration study between the Amsterdam Public Health Service (GGD) and the Academic Medical Center Amsterdam (AMC) doing wide range cross-sectional surveys. Their 2011 to 2015 survey gives us the following data on their approximating 23.000 respondents: age, sex, marital status, highest completed education, ethnicity and depressed mood. The dependant variable ‘depressed mood’ needs to be elaborated here. It is a binary value (depressed mood: yes or no) based on the 9-item Patient Health Questionnaire (PHQ-9). This is a quick survey determining depressive symptoms over the preceding two weeks. 9 questions are asked to which respondents can respond to varying from never (0) to nearly every day (3). A total score of 10 or higher will get a ‘depressed mood: yes’, while those with a score below 10 get a ‘depressed mood: no’. Regarding the population sample of this survey (Leijssen et al., 2018); its participants were randomly sampled from the municipal population register of Amsterdam and were stratified by ethnicity. Of those invited to participate, around 55% responded. Out of this group, 50% agreed to participate in the survey. Participants that were unable to fill in the survey themselves were assisted by ethically matched interviewers. Exclusions were made for participants with missing data regarding depressive symptoms (not related to the current study, but excluded nonetheless), noise exposure, educational level and occupation.
The use of GIS software makes up a big part of the data collection of the living environment characteristics. First the living environment is defined as a spatial buffer of 100 and 300 meter around the center point of each pc6 area. Then average property value and percentage of people living on a minimal income are calculated by extracting the mean value of all pc6 areas situated within each drawn buffer. The average property value and percentage of households with a minimal income for each 6-digit postcode is provided by OIS
Amsterdam.
In the research of Veldhuizen et al. (2015), similar living environment characteristics were calculated with the use of spatial buffers ranging from 50 to 1000 meter. The 100 meter buffer showed most significant results, which is the reason for including the 100 meter buffer here. The inclusion of a 300 meter makes this analysis more solid as random effect errors are prevented. Besides this, there is no universal definition of one’s living environment so it will be interesting to see what variation arises between the two sizes.
Out of the approximately 18.000 pc6 areas in Amsterdam, around 2500 areas are not included in our analysis. This is due to the unavailability of data concerning average property value. When inspected on a map, most of the missing values are located in more affluent districts (Centrum and Zuid). Apparently, it is more difficult to establish property values here, possibly due to lack of disclosure and/or shady business in the high-end Amsterdam housing market. This is a hole in our data that needs to be acknowledged yet should not be
problematic as both districts are still represented by non-missing pc6 areas. This is also the reason that the total study population is 21,733 residents instead of the approximately 23.000 residents provided by HELIUS. They had to be excluded because they lived in an excluded pc6 area.
Average blue and green space are calculated by overlaying BGT (Basisregistratie Grootschalige Topografie) maps and the buffer maps. The BGT maps contain very accurate polygons of green space areas and bodies of water (equalling blue space). Green space includes not only parks, but also roadside green areas, small green plazas and patches of grass. The amount of square meters of blue and green space within each buffer is then calculated. The result is divided by the total area of the buffers and a percentage is
established. Note that the inclusion of these different green areas significantly improves the accuracy of the variable green space. Besides this, in Leijssen (2018) blue and green space was calculated for the much larger 4-digit postcode (pc4) areas. This means that respondents living on the edge of such an area will get the value of their pc4 area, even though they might live very close to blue and green space located just outside their pc4 area.
Data analysis
Multilevel logistic regression analysis is performed in STATA and is used to establish each variables’ influence on depressed mood. The analysis is logistic because the dependant variable ‘depressed mood’ is binary. The analysis is multilevel because multiple respondents might live in the same postcode area, thus need to be included separately. Three models are used in this analysis. Model 1 includes only personal characteristics as confounders on the relationship between noise exposure and depressed mood. This model functions as a
replication of the Leijssen et al. (2018) study and results will be compared. Model 2a and 2b additionally include characteristics of the living environment calculated with a 100 meter buffer and 300 meter buffer respectively. This will show us the effects of the living environment as well as the more relevant scale at which to define the living environment.
Further on, references will be made to the inner and outer city of Amsterdam. The inner city is defined as all the area lying within the A10 ring road, with the exception of Amsterdam Noord. The outer city consists of Amsterdam Noord and all the remaining areas outside the ring road within the municipality of Amsterdam. See figure 1 for a map
Figure 1, Visual representation of the inner and outer city areas of Amsterdam based on this studies definition.
Results
Table 1 represents the prevalence of depressed mood within the socio-economic subgroups in the study population. Out of the total study population, 14.9% reported depressed mood (women = 17.2 and men= 11.7). Concerning age, the 60+ group has the least people with depressed mood (10%). The prevalence of depressed mood within ethnic groups is lowest for the Dutch (7.5%) and Ghanaian (8.9%) respondents, where Turkish, Moroccan and
Surinamese (23.1%, 20.3% and 13.6%) respondents score much higher. The education and occupation groups show logical patterns: those who are highly educated and employed show lowest prevalence of depressed mood, and those who have lower education and are not employed have a much stronger tendency of having depressed mood. On the subject of marital status, people that are married, not married or living together have fairly low scores (13.5%, 14.8% and 8.9%), where those who are divorced or have become widow/widower have much larger prevalence for depressed mood (23% and 18.4%).
Table 1. Individual characteristics of the total study population (n=21,733), and prevalence of depressed mood (PHQ-9
sum-score ≥ 10)
Characteristic Total n Prevalence of depressed
mood, % Total 21773 14.9 Sex Men Women 9227 13373 11.7 17.2 Age 18 – 29 30 – 39 40 – 49 50 – 59 ≥ 60 4090 3897 5345 5691 2750 15.4 14.5 16.1 15.9 10 Ethnic group Dutch Moroccan Turkish Surinamese Ghanaian Other 3936 3916 3757 7783 2334 46 7.5 20.3 23.1 13.6 8.9 23.9 Education Low Medium-low Medium-high High Occupation Unemployed Disabled
Not in labour force
3842 5926 6507 5349 3136 1683 3838 21.8 15.6 15.1 8.7 24.9 36.2 14.2
Employed 12911 9.8
Marital status
Married/civil union
Living together
Not married/never been married
Divorced Widow/widower 8407 2307 7481 3031 424 13.5 8.9 14.8 23.0 18.4
Table 2a and table 2b show the personal socio-economic characteristics interacting with noise exposure. First, noise exposure in 2011 is discussed. In the total study population, 47% encounters 55-59dB of noise exposure and 0.9% lives in the 70dB+ high noise exposure areas. Regarding ethnicity, the Dutch group has a surprising 2,1% of its population
encountering 70dB+ noise exposure, which is much higher than the other ethnic groups. The same goes for the highly educated group scoring 1.6% in the high noise exposure category, where its lower counterparts score only 0.9% or 0.6%. Occupation and marital status show no statistical evidence of having differences within its groups. Comparing this to the results with the 2017 (table 2b) road traffic noise data, some things come to attention. First of all, on average, the noise levels in 2017 are lower than in 2011 because the distribution of the total population shifts to the lower noise levels. Regarding ethnicity, the Dutch group is no longer the most represented group in the 70dB+ noise exposure areas (0.6%) but has been overtaken by both Moroccan and Turkish groups (0.8%, 0.8%). Surinamese and Ghanaian people area exposed to the least amount of noise in both years, consequently exceeding the other groups by around 10% in the lowest noise category. Regarding education, the highly educated group has gone from 1.6% to 0.5% in the 70dB+ category. They do show a higher value in the 65-69 dB category. Occupation and marital status show no interesting patterns in 2017 either.
Table 2a. Road traffic noise exposure in 2011 in the total study sample (n=21,733) by personal characteristics. Traffic noise exposure 2011 (%)
45-54 dB(A) 55-59 dB(A) 60-64 dB(A) 65-69 dB(A) ≥70 dB(A) Total Personal characteristics Ethnicity Dutch Moroccan Turkish Surinamese Ghanaian Other Education Low Medium-low Medium-high High Occupation Unemployed Disabled
Not in labour force
Employed Marital status Married/civil union 32.5 28.8 26.0 29.3 37.0 39.1 58.7 31.0 34.9 32.3 31.0 30.3 31.9 30.3 33.8 33.3 47.1 43.5 48.4 47.6 46.7 52.2 23.9 47.9 48.6 47.9 44.4 50.0 50.2 47.3 45.9 46.0 15.3 19.0 19.3 16.9 13.1 7.2 13.0 16.0 12.7 15.2 17.9 15.0 13.2 16.6 15.3 15.7 4.2 6.6 5.3 4.9 2.9 1.3 4.3 4.2 3.3 3.9 5.5 3.9 3.9 4.7 4.2 4.0 0.9 2.1 1.1 1.3 0.3 0.1 0 0.9 0.6 0.6 1.6 0.8 0.8 1.2 0.9 0.9
Living together
Not married/never been married
Divorced Widow/widower 32.4 30.4 35.8 30.4 45.6 48.4 47.7 50.7 16.1 15.8 12.4 14.3 5.0 4.4 3.6 3.5 0.9 1.1 0.6 1.2
Table 2b. Road traffic noise exposure in 2017 in the total study sample (n=21,733) by personal characteristics Traffic noise exposure 2017 (%)
45-54 dB(A) 55-59 dB(A) 60-64 dB(A) 65-69 dB(A) ≥70 dB(A) Total Personal characteristics Ethnicity Dutch Moroccan Turkish Surinamese Ghanaian Other Education Low Medium-low Medium-high High Occupation Unemployed Disabled
Not in labour force
Employed Marital status 52.7 49.4 45.4 46.4 57.9 62.4 78.3 50.0 55.2 53.2 50.9 50.9 52.5 49.1 54.2 30.4 30.6 34.1 33.6 28.3 26.2 15.2 32.1 29.5 30.2 30.5 33.2 30.6 32.6 28.9 13.8 14.9 16.6 16.3 11.7 10.8 4.3 14.4 13.0 13.9 14.5 13.2 13.9 14.8 13.7 2.6 4.5 3.1 2.9 2.0 0.5 2.2 2.8 2.0 2.3 3.6 2.3 2.6 2.6 2.7 0.5 0.6 0.8 0.8 0.1 0.2 0.0 0.7 0.3 0.4 0.5 0.4 0.4 0.8 0.4
Married/civil union
Living together
Not married/never been married
Divorced Widow/widower 52.4 53.6 51.5 55.3 54.2 30.2 29.2 31.3 29.5 29.0 14.0 13.7 14.0 13.3 14.0 2.8 3.2 2.6 1.6 2.1 0.6 0.3 0.5 0.3 0.7
Table 3 represents the correlation (Pearson r) with both depressed mood and traffic noise. This table allows us to see in which way the living area characteristics interact with depressed mood and traffic noise. Besides this, both data based on 100 and 300 meter buffers are
included to see if they show large differences and to see which buffer size interacts the most. The correlations with depressed mood and 2011 noise exposure are first discussed, after which this is compared to the noise exposure in 2017. First of all, average property values correlate negatively with depressed mood (100m = -.05 and, 300m = -.038), but positively with traffic noise (100m = .08 and 300m = .121) which can be explained by the pattern found in Amsterdam of the more affluent living in the traffic rich inner city and the less affluent living in quieter outer neighbourhoods. The percentage of people living on a minimal income in the living area does not correlate with depressed mood and correlates only slightly with traffic noise (100m = -0.35 and 300m = -.034). Green space shows a small negative
correlation with depressed mood on both buffer sizes (100m = -.03 and, 300m = -.037), yet correlates much stronger with traffic noise (100m = -.292 and 300m = -.315). This indicates that green space does not have its influence on depressed mood directly, but rather that more green space goes together with less traffic noise and therefore less depressed mood. This will be discussed in more detail in the interpretation section. Blue space seems to show the same dynamic as green space, yet to a lesser degree. It correlates negatively with depressed mood on 100 meter buffers (-.043) and shows no correlation on 300 meter buffers. There is a stronger correlation between blue space and traffic noise (100m = -.049 and 300m = -.061).
Road traffic noise data from 2017 produces correlations in the same directions, although they vary in strength. In each case, actually, the correlations with 2017 noise data are weaker, with the exception of percentage of people living on a minimal income based on a 100 meter buffer.
Table 3. Pearson Correlation Coefficient for living area characteristics, depressed mood and traffic noise exposure in 2011
and 2017. Estimates with p value <0.01 are highlighted in bold.
Characteristic Mean SD Correlation with
depressed mood (Pearson r) Correlation with traffic noise (2011) (Pearson r) Correlation with traffic noise (2017) (Pearson r) 100 meter buffers Average property value
Percentage of people living on a minimal income Green space Blue space 201615 9.59 18.01 4.38 80148 21.95 15.79 7.46 -.05 .002 -.03 -.043 .08 -.035 -.292 -.049 .04 -.095 -.168 -.037
300 meter buffers Average property value
Percentage of people living on a minimal income Green space Blue space 201273 6.36 -.21.95 7.57 72380 .54 13.67 7.95 -.038 .005 -.037 -.015 .121 -.034 -.315 -.061 .069 -.017 -.178 -.04
The prevalence of depressed mood within the noise level categories is presented in Table 4. Again, the 2011 noise data is discussed first. The lowest noise level category (45-54 dB) has 14.7% prevalence of depressed mood, where the highest noise level category (≥70dB) scores 22.5%. The last three columns present the results of the multilevel logistic regression
analysis. Model 1 adjusts for all personal characteristics (age, sex, ethnicity, educational level, occupational status, marital status). Model 2a adjusts for characteristics of the living environment (average property value, percentage of people living on a minimal income, green space and blue space) that were calculated by defining the living environment as the area within a 100 meter circular buffer around the pc6 area of respondents. Model 2b is similar to model 2a but is based on 300 meter buffers as a respondents’ living environment. Model 1 shows an odds ratio of 0.83 (95% CI 0.74, 0.95) for the noise exposure level of 60-64 dB and an odds ratio of 1.52 (95% CI 1.04, 2.22) for the ≥70dB noise exposure level. Adjusting for living area characteristics (Model 2a and 2b) gives increasing odds ratios. Model 2a, based on 100 meter buffers, increases odds ratios slightly: an odds ratio of 0.84 (95% CI 0.73, 0.95) for the 60-64dB category and an odds ratio of 1.54 (95% CI 1.06, 2.27) for the highest noise level category of ≥70dB. Model 2b, based on 300 meter buffers, shows a larger variation: an odds ratio of 0.85 (95% CI 0.75, 0.97) for the 60-64dB category and an odds ratio of 1.59 (95% CI 1.09, 2.34) for the ≥70dB category. This indicates that defining the living area characteristics with a 300 meter buffer is more relevant than with 100 meter buffers.
Road traffic noise exposure in 2017 produces different results. Firstly, the prevalence of depressed mood is less spread out, which is related to the fact that the average noise exposure levels are lower in the 2017 data. Besides this, confident odds ratios are only found in the 55-59 dB category, although they barely vary between the three models used (Model 1: 0.89 (95% CI 0.81, 0.97), Model 2: 0.89 (95% CI 0.81, 0.98), Model 3: 0.90 (95% CI 0.81, 0.98)). No other confident odds ratios are produced using the 2017 road traffic noise data. Thus, overall, 2011 noise exposure is a better predictor of depressed mood than the 2017 road traffic noise data.
Table 4. The association of different levels of noise exposure from road traffic with depressed mood (PHQ-9 sum-score ≥10) in the total study sample (n=21,773). Exposure Prevalence of depressed mood (%) Model 11 OR [95%-CI] Model 2a2 OR [95%-CI] Model 2b3 OR [95%-CI] Noise exposure 2011 45-54 dB(A) 55-59 dB(A) 60-64 dB(A) 65-69 dB(A) ≥70dB(A) 14.7 15.0 13.9 16.4 22.5 1.00 0.94 [0.86, 1.03] 0.83 [0.74, 0.95] 1.03 [0.85, 1.26] 1.52 [1.04, 2.22] 1.00 0.94 [0.86, 1.02] 0.84 [0.73, 0.95] 1.05 [0.85, 1.28] 1.54 [1.06, 2.27] 1.00 0.95 [0.87, 1.04] 0.85 [0.75, 0.97] 1.07 [0.87, 1.31] 1.59 [1.09, 2.34] Noise exposure 2017 45-54 dB(A) 55-59 dB(A) 60-64 dB(A) 65-69 dB(A) ≥70dB(A) 14.6 14.5 16.5 15.9 19.6 1.00 0.89 [0.81, 0.97] 1.06 [0.94, 1.19] 0.99 [0.77, 1.28] 1.11 [0.65, 1.87] 1.00 0.89 [0.81, 0.98] 1.06 [0.94, 1.19] 1.01 [0.78, 1.30] 1.12 [0.66, 1.91] 1.00 0.90 [0.81, 0.98] 1.06 [0.94, 1.20] 1.01 [0.78, 1.30] 1.15 [0.68, 1.95]
1 Model 1: adjusted for individual characteristics (age, sex, ethnicity, educational level, occupational status, marital status). 2 Model 2a:
model 1 plus additional adjustments for living area characteristics based on a 100 meter buffer (average property value, percentage of people living on a minimal income, green space and blue space). 3Model 2b: model 1 plus additional adjustments for living area characteristics
Discussion
Summary of results
This study is in accordance with the literature on road traffic noise and depressed mood, showing that there is an association between the two. Exposure to high noise levels was associated with depressed mood despite confounding for personal and living environment characteristics. The personal characteristics show many more and stronger connections with depressed mood than the living environment characteristics. Results of the 300 meter buffer living environments show slightly more correlation than their 100 meter counterparts. Road traffic noise exposure in 2011 correlates more strongly with depressed mood than road traffic noise exposure in 2017.
Evaluation of study limitations
Firstly, as was mentioned before, around 2500 pc6 areas were not included since data on average housing prices was missing. These pc6 areas were located mostly in more affluent parts of the city. I speculate that this is caused by the lack of transparency in the high-end Amsterdam housing market. Although this is a large amount of missing data, I do not think it is problematic since there are still plenty of non-missing pc6 areas representing the more affluent neighbourhoods.
Secondly, only road traffic noise is included in this study, even though noise pollution has many more sources. Noise from nightlife, for example, can be quite persistent in
Amsterdam, but is hard to be mapped. Noise from construction is also causing a lot of nuisance as many homes in the inner city are renovating and expanding. Besides all this, air traffic noise originating from the near Schiphol airport is a heavily controversial subject in the Amsterdam metropolitan area. To include these sources into this study would be too ambitious due to the unavailability time and data.
Thirdly, it was not possible to account for the difference in sensitivity to noise exposure between respondents. There are many ways in which this can differ such as differences in hearing ability, time spent at home and the tendency to get annoyed by noise. Besides this, housing conditions such as location of the bedroom (roadside or not), ability to ‘flee’ from the noise in more isolated rooms (typically at the back of the house) and the general insulation of the house are all factors that can be considered (de Kluizenaar et al., 2013). These factors will influence the amount of noise exposure, and the way respondents can deal with it. They can not, however, be included in our analysis.
Consistent with the structure of this study, initially road traffic noise exposure in 2011 is discussed, followed by the interpretation of the differences with the road traffic noise exposure in 2017. Furthermore, this section has the following structure. Firstly, the replication of the Leijssen et al. (2018) study is discussed. Secondly, the effects of the
inclusion of more detailed geographic data on the living environment are interpreted. Thirdly, the findings are placed in the context of Amsterdam. Finally, changes over time are
interpreted by examining the road traffic noise exposure in 2017.
The comparison of results with Leijssen et al. (2018) is not shown in detail here. Our findings are similar and show the same patterns. To keep this study ordered and not to be repetitive, only the most relevant findings are discussed. Although the outcomes differ slightly, mostly due to the loss of some respondents as a result of missing data, the main findings are the same. Regarding ethnicity, the Dutch group is exposed to higher noise levels, yet report less depressed mood than the other ethnic groups. Highly educated individuals also live in noisier areas on average yet report less depressed mood. In short, this is the case for the more affluent groups in the city. This can be explained by the fact that the inner city is the most affluent area, but also the most traffic-intensive area (Savini et al., 2016). Both
cars/trucks and trams are more abundant the closer one gets to the city center. Furthermore, this study provides similar evidence for a positive association between road traffic noise exposure and depressed mood. People living in a 60 to 64 dB noise exposure area have 17% less chance of having depressed mood, while those living in a 70dB+ noise exposure area have a 52% larger chance of having depressed mood.
The main goal of this study is to explore the way the living environment interacts with the relationship between road traffic noise exposure and depressed mood. This is based on the expectation that gathering more geographically detailed data on the living environment reflects its effects in a more precise manner. Based on the assumption that one's' living environment influences depressive tendencies, the expectation was that this addition would act as another confounder, effectively decreasing the statistical evidence for noise exposure predicting depressed mood. It turns out that this is not the case. Although this addition did produce variation in the odds ratios, the variation went into the other direction. The 60-64 dB category predicts 2% less depressed mood and the 70dB+ category predicts 5% more
depressed mood. This is explained by the fact that the living environment characteristics correlate stronger with noise exposure than with depressed mood. The increase of the odds ratios in model 2a and model 2b, thus, comes from the interaction with noise exposure, rather than depressed mood. Nevertheless, this interaction with noise exposure is relevant here. The aforementioned inner-outer city dichotomy is confirmed by the fact that areas with higher average property values have larger noise exposure. The same can be said about the fact that areas with more green space have less noise exposure, as the outer city areas support much more green space than the dense inner city. A minor thing to consider here is that a small part of the negative correlation between green space and noise exposure is explained by the fact that trees found in green space absorb some of the noise produced by roads (Islam et al., 2012). It is surprising, however, that both green and blue space show such weak correlations with depression because multiple other studies found strong connections (Cohen-Cline et al., 2015; Miles et al., 2012; Gascon et al., 2018). This most likely has to do with the distribution of both green and blue space in the city. It is more prevalent in the outer city areas, as well as depressed mood. Location specific urban geography functions as an exception to the rule here. While urban green and blue space are most likely still beneficial for those living around it, the socioeconomic situation is of more importance.
None of the findings confirmed the existence of the neighbourhood poverty effect or neighbourhood chronic stress in Amsterdam (Yen & Kaplan, 1999; Matheson et al., 2006). The absence of correlation between depression and percentage of people living on a minimal
income indicates this, because it would be expected that living among this group would amplify the chance of depressed mood. Reflecting on Pickett & Pearl (2001), Amsterdam can be added to the list of places where personal characteristics, rather than those of the living environment, are much more important predictors of health impacts.
In short, the living environment was expected to weaken the relationship between noise exposure and depressed mood, yet it does the exact opposite because it correlates more strongly with noise exposure and not with depressed mood. Furthermore, living environment characteristics based on 300 meter buffers show stronger correlations overall, which implies that the living environment is not an important factor to consider in relation to
depression. The socioeconomic status of the individual is of much larger influence, at least in the case of Amsterdam.
All of this points us to an interesting dynamic in Amsterdam. Lower SES groups experience higher noise exposure and lower SES living environments have more green space and less traffic noise exposure, yet still contain on average more people with depressed mood. This exposes dynamics of a thriving and gentrifying European global city. The more affluent return to the inner city, and the less affluent move/are pushed out to the outer areas and suburbs. This is opposite to the conventional image of the low SES groups ending up in the noisiest, least green and least attractive parts of the city (in-between infrastructure corridors or industries). It means that in Amsterdam, despite living in the quietest and greenest neighbourhoods, the less affluent suffer more from depression. Although road traffic noise exposure has been proven to make a difference, one’s position on the societal ladder seems to be a more important predictor.
One might say that this reversed dynamic sort of evens everything out, the affluent having to deal with more environmental nuisance than the less affluent. But one might also say that noise pollution needs to be reduced in any case, especially when inner cities are densifying, and more and more people get exposed to disturbing noise. At least, that is the current frame. But when the road traffic noise exposure in 2017 is assessed, it seems that Amsterdam is an exception to this frame. Overall, the city has become quieter as the average dB levels are lower. Besides, the inner-outer city dichotomy appears less strong as the Dutch and highly educated groups are no longer the most represented in the higher noise exposure categories. Before attempting to explain this variation, a couple of side notes have to be provided. Firstly, the noise data from 2011 naturally has a better fit with the HELIUS survey data collected between 2011 and 2015, which partially explains the smaller correlations with the 2017 noise exposure data. Secondly, slightly different measuring methods have been used, which can explain some of the variation, especially on the lower average dB level in 2017. Thirdly, the cell sizes differ between the two noise maps (2011: 25m by 25m; 2017: 10m by 10m) which results in higher values for some pc6 areas and lower values for others.
In this section I provide alternative explanations to the fact that the 2017 noise data indicates a decrease in noise overall, and a decrease in the differences between the inner and outer city.
Amsterdam boasts the title of ‘bicycle capital of the world’ as it has extensive (and still improving) cycling infrastructure (van der Zee, 2015). 36% of total movements in the city were made by bicycle in 2017. Even though Amsterdam is growing, car usage is
declining, which can mostly be appointed to the increasing popularity of the bicycle (Verkeer in Beeld, 2017).
In 2008, Amsterdam implemented an environmental zone for the inner city (roughly overlapping the area defined as inner city in this study) (Expertise Centrum Milieuzones, n.d.). This environmental zone is restricted for the older and dirtier lorries transporting cargo into the city. Although this measure was mainly taken to improve air quality, older vehicles
tend to produce more noise pollution. Thus, forcing importers of cargo to switch to
alternatives decreases overall noise production in the city. Since 2008 the city has expanded restrictions to this zone multiple times to include e.g. delivery vans.
Amsterdam has also been a frontrunner in Electric Vehicle usage, with its largest advancements happening within our period of interest (2011-2017). In 2017, Amsterdam has the smartest and most used network of public charging stations in the world, consisting of an exceptional 3,834 stations (RVO, 2017:1; RVO, 2017:2). This was accomplished by the municipalities’ multiple implementations of policy to reduce carbon-based motor vehicle usage. For instance, electric vehicle users get priority for parking permits, of which the demand is even larger in the inner city due to the lack of space (RVO, 2017:2). Residents of the more affluent inner city are also more capable of switching to electric driving, as this is still a rather expensive step to be taken. Besides this, the municipality is switching to electric vehicles for almost all the services it provides (RVO, 2017:2). As electric vehicles are quieter than fuel powered vehicles, this development adds to the reduction of noise pollution.
Public transit in Amsterdam is also quite extensive (both within the city as regionally), and its usage is still growing. De Koster (2017) coincidentally mentions its increase in usage in Amsterdam between 2011 and 2017: an 11% increase, adding up to 227 million passengers per year. Availability of public transit discourages people from taking a car into the busy inner city as it will often be quicker and as there is no worry about having to park. The only distinction that has to be made is that the inner city is largely covered by trams. An increase in public transit usage would therefore decrease car usage, but increase tram usage, which might make the thoroughfares louder rather than quieter.
The aforementioned trends of decreasing road traffic noise exposure and decreasing differences between the inner and outer city are supported by these recent developments. All of which at least roughly overlap with the period between 2011 and 2017. But to make noise exposure data properly comparable over time, the concerned national agencies need to adopt the exact same measuring prescription. Besides this, it is more relevant to compare this subject over period longer than six years.
Conclusion
This study adds to the growing evidence for road traffic noise affecting depressed mood. After confounding for personal and neighbourhood characteristics, the association still stands. Yet when road traffic noise exposure from 2017 is considered, the association becomes weaker. High road traffic noise exposure resulting in depressed mood depends on the personal and environmental situation of any resident. The personal situation is a much stronger confounder The main goal of this study, to include more geographically detailed information on the living environment, was fulfilled and yielded no staggering results. The inclusion of an extra set of noise exposure data did raise interesting questions. The
developments in Amsterdam show that the city is dealing reasonably well with noise
pollution. A lot of effort is put in policy to reduce noise and air pollution in the busiest parts of town. Awareness of environmental nuisances is constantly growing, as well as electric driving and usage of other modes of transportation. At the same time, Amsterdam is still in the middle of an exceptional growth boom which will keep attracting more and more people to Amsterdam. Therefore, countering noise pollution needs to be done as a continuous effort, constantly reevaluating and improving.
Concerning research on this topic, it would be good to have qualitative research to support and/or question the findings of the large amount of quantitative studies. It could provide an insight in the exact way that road traffic noise contributes to depressed mood, specifically if this happens on a more conscious or subconscious level. In other words, the question is if the strength of noise inducing depressed mood is dependent on mindset or mood, or if has a more direct universal effect. Additionally, qualitative research could answer a persistent question that has arisen during the research process. Might the lack of negative correlation between depressed mood and blue and green space be at least partially attributed to the idea that people with depressed mood leave the house less often, thus making less use of blue and green spaces?
Acknowledgements
I want to thank HELIUS for their cooperation in providing the necessary data. Anton Kunst, Head of Department of Public Health, Amsterdam UMC, University of Amsterdam, provided me with many insights and directions, which I am thankful for. Lastly and mostly I want to thank Els Veldhuizen, my thesis counselor at the University of Amsterdam, for putting so much effort in helping me with requesting data in the academic world.
References
Atlas Leefomgeving (2016) Uitleg geluid. Retrieved from:
https://www.atlasleefomgeving.nl/meer-weten/geluid/uitleg-geluid Accessed date: 25 June
2019
Atlas Leefomgeving (2017) map: “Geluid wegverkeer alle wegen (Lden)”. Retrieved
from:
https://www.atlasleefomgeving.nl/kaarten?config=3ef897de-127f-471a-959b- 93b7597de188&gm-x=150000&gm-y=455000&gm-z=3&gm-b=1544180834512,true,1;1553244467282,true,0.8;
Accessed date: 19 april 2019.
Bodin, T., Albin, M., Ardö, J., Stroh, E., Östergren, P.O. and Björk, J. (2009) Road traffic noise and hypertension: results from a cross-sectional public health survey in southern Sweden. Environmental Health, 8(1), p.38.
Cohen-Cline, H., Turkheimer, E. and Duncan, G.E. (2015) Access to green space, physical activity and mental health: a twin study. J Epidemiol Community Health, 69(6), pp.523-529.
Expertise Centrum Milieuzones (n.d.) Amsterdam. Retrieved from:
https://www.milieuzones.nl/amsterdam Accessed date: 27 June 2019.
Galea, S., Ahern, J., Rudenstine, S., Wallace, Z., & Vlahov, D. (2005) Urban built
environment and depression: a multilevel analysis. Journal of Epidemiology & Community
Health, 59(10), 822-827.
Gascon, M., Sánchez-Benavides, G., Dadvand, P., Martínez, D., Gramunt, N., Gotsens, X., Cirach, M., Vert, C., Molinuevo, J.L., Crous-Bou, M. and Nieuwenhuijsen, M. (2018) Long-term exposure to residential green and blue spaces and anxiety and depression in adults: a cross-sectional study. Environmental research, 162, pp.231-239.
Goines, L. and Hagler, L. (2007) Noise pollution: a modern plague. Southern medical
journal, 100(3), pp.287-295.
Islam, M. N., Rahman, K. S., Bahar, M. M., Habib, M. A., Ando, K., & Hattori, N. (2012) Pollution attenuation by roadside greenbelt in and around urban areas. Urban Forestry &
Janssen, S. and Hong, J. (2017) Recent progress in the field of community response to noise.
12th ICBEN Congress on Noise as a Public Health Problem, 18-22 June, 2017, Zurich, 1-9.
de Kluizenaar, Y., Janssen, S., Vos, H., Salomons, E., Zhou, H. and van den Berg, F. (2013) Road traffic noise and annoyance: A quantification of the effect of quiet side exposure at dwellings. International journal of environmental research and public health, 10(6), pp.2258-2270.
de Koster, Y. (2017) Ondanks toename fiets, auto op nummer één. Binnenlands Bestuur, 6 January 2017
Leijssen, J.B., Snijder, M.B., Timmermans, E.J., Generaal, E., Stronks, K. and Kunst, A.E. (2018) The association between road traffic noise and depressed mood among different ethnic and socioeconomic groups. The HELIUS study. International journal of hygiene and
environmental health. 222(2), pp.221-229.
Lupien, S.J., McEwen, B.S., Gunnar, M.R. and Heim, C. (2009) Effects of stress throughout the lifespan on the brain, behaviour and cognition. Nature reviews neuroscience, 10(6), p.434.
Matheson, F. I., Moineddin, R., Dunn, J. R., Creatore, M. I., Gozdyra, P., & Glazier, R. H. (2006) Urban neighborhoods, chronic stress, gender and depression. Social science & medicine, 63(10), 2604-2616.
Miles, R., Coutts, C. and Mohamadi, A. (2012) Neighborhood urban form, social environment, and depression. Journal of Urban Health, 89(1), pp.1-18.
Öhrström, E. (1991) Psycho-social effects of traffic noise exposure. Journal of Sound and
Vibration, 151(3), pp.513-517.
Orban, E., McDonald, K., Sutcliffe, R., Hoffmann, B., Fuks, K.B., Dragano, N., Viehmann, A., Erbel, R., Jöckel, K.H., Pundt, N. and Moebus, S. (2015) Residential road traffic noise and high depressive symptoms after five years of follow-up: results from the Heinz Nixdorf recall study. Environmental health perspectives, 124(5), pp.578-585.
Ouis, D. (2001) Annoyance from road traffic noise: a review. Journal of environmental
psychology, 21(1), pp.101-120.
Pickett, K.E. and Pearl, M. (2001) Multilevel analyses of neighbourhood socioeconomic context and health outcomes: a critical review. Journal of Epidemiology & Community
Health, 55(2), pp.111-122.
PBL (2011) Geluidsbelasting wegverkeer in Nederland 2011. Retrieved from
https://geoservice.pbl.nl/arcgis/rest/services/projecten/Geluidbelasting_wegverkeer_in_Neder
land_20002011/MapServer , Accessed date: 25 june 2019.
PBL (2008) Modelling local environmental quality and its impact on health. Retrieved from
https://www.pbl.nl/sites/default/files/cms/publicaties/550034001.pdf, Accessed date: 27 june
Riemann, D. and Voderholzer, U. (2003) Primary insomnia: a risk factor to develop depression?. Journal of affective disorders, 76(1-3), pp.255-259.
RVO (2017:1) Verzilvering Verdienpotentieel Elektrisch Vervoer, pp. 22. Rijksdienst voor Ondernemend Nederland.
RVO (2017:2) Electric transport in the Netherlands, Highlights 2017. pp. 16-31. Rijksdienst voor Ondernemend Nederland
Roswall, N., Høgh, V., Envold-Bidstrup, P., Raaschou-Nielsen, O., Ketzel, M., Overvad, K., Olsen, A. and Sørensen, M. (2015) Residential exposure to traffic noise and health-related quality of life—a population-based study. PLoS One, 10(3), p.p 120-199.
Savini, F., Boterman, W. R., Van Gent, W. P., & Majoor, S. (2016) Amsterdam in the 21st century: Geography, housing, spatial development and politics. Cities, 52, 103-113.
Seidler, A., Hegewald, J., Seidler, A.L., Schubert, M., Wagner, M., Dröge, P., Haufe, E., Schmitt, J., Swart, E. and Zeeb, H. (2017) Association between aircraft, road and railway traffic noise and depression in a large case-control study based on secondary data.
Environmental research, 152, pp.263-271.
Sørensen, M., Hvidberg, M., Andersen, Z.J., Nordsborg, R.B., Lillelund, K.G., Jakobsen, J., Tjønneland, A., Overvad, K. and Raaschou-Nielsen, O. (2011) Road traffic noise and stroke: a prospective cohort study. European heart journal, 32(6), pp.737-744.
Stansfeld, S., Gallacher, J., Babisch, W. and Shipley, M. (1996) Road traffic noise and psychiatric disorder: prospective findings from the Caerphilly Study. Bmj, 313(7052), pp.266-267.
Veldhuizen, E., Musterd, S., Dijkshoorn, H., & Kunst, A. (2015) Association between self-rated health and the ethnic composition of the residential environment of six ethnic groups in Amsterdam. International journal of environmental research and public health, 12(11), 14382-14399.
Verkeer in Beeld (2017) Auto steeds minder populair in Amsterdam. Retrieved from:
https://www.verkeerinbeeld.nl/nieuws/060117/auto-steeds-minder-populair-in-amsterdam
Accessed date: 27 June 2019.
World Health Organization (2011) European Commission: Burden of Disease from
Environmental Noise: Quantification of Healthy Life Years Lost in Europe. The WHO
European Center for Environment and Health.
Yen, I. H., & Kaplan, G. A. (1999) Poverty area residence and changes in depression and perceived health status: evidence from the Alameda County Study. International journal of epidemiology, 28(1), 90-94.
van der Zee, R. (2015) How Amsterdam became the bicycle capital of the world. The
