Empirical spatial air pollution models have been applied extensively to assess exposure in epidemiological studies with increasingly sophisticated and complex statistical algorithms beyond ordinary ...linear regression. However, different algorithms have rarely been compared in terms of their predictive ability.
This study compared 16 algorithms to predict annual average fine particle (PM2.5) and nitrogen dioxide (NO2) concentrations across Europe. The evaluated algorithms included linear stepwise regression, regularization techniques and machine learning methods. Air pollution models were developed based on the 2010 routine monitoring data from the AIRBASE dataset maintained by the European Environmental Agency (543 sites for PM2.5 and 2399 sites for NO2), using satellite observations, dispersion model estimates and land use variables as predictors. We compared the models by performing five-fold cross-validation (CV) and by external validation (EV) using annual average concentrations measured at 416 (PM2.5) and 1396 sites (NO2) from the ESCAPE study. We further assessed the correlations between predictions by each pair of algorithms at the ESCAPE sites.
For PM2.5, the models performed similarly across algorithms with a mean CV R2 of 0.59 and a mean EV R2 of 0.53. Generalized boosted machine, random forest and bagging performed best (CV R2~0.63; EV R2 0.58–0.61), while backward stepwise linear regression, support vector regression and artificial neural network performed less well (CV R2 0.48–0.57; EV R2 0.39–0.46). Most of the PM2.5 model predictions at ESCAPE sites were highly correlated (R2 > 0.85, with the exception of predictions from the artificial neural network). For NO2, the models performed even more similarly across different algorithms, with CV R2s ranging from 0.57 to 0.62, and EV R2s ranging from 0.49 to 0.51. The predicted concentrations from all algorithms at ESCAPE sites were highly correlated (R2 > 0.9). For both pollutants, biases were low for all models except the artificial neural network. Dispersion model estimates and satellite observations were two of the most important predictors for PM2.5 models whilst dispersion model estimates and traffic variables were most important for NO2 models in all algorithms that allow assessment of the importance of variables.
Different statistical algorithms performed similarly when modelling spatial variation in annual average air pollution concentrations using a large number of training sites.
•Multiple statistical algorithms with very different assumptions were compared.•Despite the difference in modeling frameworks, predictions among the models exhibit generally good agreement.•The use of an external evaluation dataset strengthens evaluation by cross-validation.
•Exposure to PM2.5 was associated with higher risk of lung cancer.•Elevated risks persisted even at levels lower than the EU limit value of 25 µg/m3.•No association between NO2, BC or O3 and lung ...cancer incidence was observed.
Ambient air pollution has been associated with lung cancer, but the shape of the exposure-response function - especially at low exposure levels - is not well described. The aim of this study was to address the relationship between long-term low-level air pollution exposure and lung cancer incidence.
The “Effects of Low-level Air Pollution: a Study in Europe” (ELAPSE) collaboration pools seven cohorts from across Europe. We developed hybrid models combining air pollution monitoring, land use data, satellite observations, and dispersion model estimates for nitrogen dioxide (NO2), fine particulate matter (PM2.5), black carbon (BC), and ozone (O3) to assign exposure to cohort participants’ residential addresses in 100 m by 100 m grids. We applied stratified Cox proportional hazards models, adjusting for potential confounders (age, sex, calendar year, marital status, smoking, body mass index, employment status, and neighborhood-level socio-economic status). We fitted linear models, linear models in subsets, Shape-Constrained Health Impact Functions (SCHIF), and natural cubic spline models to assess the shape of the association between air pollution and lung cancer at concentrations below existing standards and guidelines.
The analyses included 307,550 cohort participants. During a mean follow-up of 18.1 years, 3956 incident lung cancer cases occurred. Median (Q1, Q3) annual (2010) exposure levels of NO2, PM2.5, BC and O3 (warm season) were 24.2 µg/m3 (19.5, 29.7), 15.4 µg/m3 (12.8, 17.3), 1.6 10−5m−1 (1.3, 1.8), and 86.6 µg/m3 (78.5, 92.9), respectively. We observed a higher risk for lung cancer with higher exposure to PM2.5 (HR: 1.13, 95% CI: 1.05, 1.23 per 5 µg/m3). This association was robust to adjustment for other pollutants. The SCHIF, spline and subset analyses suggested a linear or supra-linear association with no evidence of a threshold. In subset analyses, risk estimates were clearly elevated for the subset of subjects with exposure below the EU limit value of 25 µg/m3. We did not observe associations between NO2, BC or O3 and lung cancer incidence.
Long-term ambient PM2.5 exposure is associated with lung cancer incidence even at concentrations below current EU limit values and possibly WHO Air Quality Guidelines.
Both social and physical neighbourhood factors may affect residents' health, but few studies have considered the combination of several exposures in relation to individual health status.
To assess a ...range of different potentially relevant physical and social environmental characteristics in a sample of small neighbourhoods in the Netherlands, to study their mutual correlations and to explore associations with morbidity of residents using routinely collected general practitioners' (GPs') data.
For 135 neighbourhoods in 43 Dutch municipalities, we could assess area-level social cohesion and collective efficacy using external questionnaire data, urbanisation, amount of greenspace and water areas, land use diversity, air pollution (particulate matter (PM) with a diameter <10 μm (PM10), PM <2.5 μm (PM2.5) and nitrogen dioxide (NO2), and noise (from road traffic and from railways). Health data of the year 2013 from GPs were available for 4450 residents living in these 135 neighbourhoods, that were representative for the entire country. Morbidity of 10 relevant physical or mental health groupings was considered. Individual-level socio-economic information was obtained from Statistics Netherlands. Associations between neighbourhood exposures and individual morbidity were quantified using multilevel mixed effects logistic regression analyses, adjusted for sex, age (continuous), household income and socio-economic status (individual level) and municipality and neighbourhood (group level).
Most physical exposures were strongly correlated with degree of urbanisation. Social cohesion and collective efficacy tended to be higher in less urbanised municipalities. Degree of urbanisation was associated with higher morbidity of all disease groupings. A higher social cohesion at the municipal level coincided with a lower prevalence of depression, migraine/severe headache and Medically Unexplained Physical Symptoms (MUPS). An increase in both natural and agricultural greenspace in the neighbourhood was weakly associated with less morbidity for all conditions. A high land use diversity was consistently associated with lower morbidities, in particular among non-occupationally active individuals.
A high diversity in land use of neighbourhoods may be beneficial for physical and mental health of the inhabitants. If confirmed, this may be incorporated into urban planning, in particular regarding the diversity of greenspace.
•Less crowded neighbourhoods have both social and environmental benefits.•Greenspace near the home may be beneficial for physical and mental health.•A diverse land use in the neighbourhood may have health benefits for the residents.
Health implications of long-term exposure to ubiquitously present ultrafine particles (UFP) are uncertain. The aim of this study was to investigate the associations between long-term UFP exposure and ...natural and cause-specific mortality (including cardiovascular disease (CVD), respiratory disease, and lung cancer) in the Netherlands.
A Dutch national cohort of 10.8 million adults aged ≥ 30 years was followed from 2013 until 2019. Annual average UFP concentrations were estimated at the home address at baseline, using land-use regression models based on a nationwide mobile monitoring campaign performed at the midpoint of the follow-up period. Cox proportional hazard models were applied, adjusting for individual and area-level socio-economic status covariates. Two-pollutant models with the major regulated pollutants nitrogen dioxide (NO2) and fine particles (PM2.5 and PM10), and the health relevant combustion aerosol pollutant (elemental carbon (EC)) were assessed based on dispersion modelling.
A total of 945,615 natural deaths occurred during 71,008,209 person-years of follow-up. The correlation of UFP concentration with other pollutants ranged from moderate (0.59 (PM2.5)) to high (0.81 (NO2)). We found a significant association between annual average UFP exposure and natural mortality HR 1.012 (95 % CI 1.010–1.015), per interquartile range (IQR) (2723 particles/cm3) increment. Associations were stronger for respiratory disease mortality HR 1.022 (1.013–1.032) and lung cancer mortality HR 1.038 (1.028–1.048) and weaker for CVD mortality HR 1.005 (1.000–1.011). The associations of UFP with natural and lung cancer mortality attenuated but remained significant in all two-pollutant models, whereas the associations with CVD and respiratory mortality attenuated to the null.
Long-term UFP exposure was associated with natural and lung cancer mortality among adults independently from other regulated air pollutants.
Display omitted
•Exposure to PM2.5, NO2, and BC was associated with premature natural mortality in 3.1 million Danes.•Associations were found with cardiorespiratory, diabetes, and lung cancer ...mortality.•We present novel associations with dementia and psychiatric disorders mortality.•Associations persisted below EU limit values of PM2.5 and NO2.•The associations were robust after the indirect adjustment of smoking and obesity.
The association between long-term exposure to air pollution and mortality from cardiorespiratory diseases is well established, yet the evidence for other diseases remains limited.
To examine the associations of long-term exposure to air pollution with mortality from diabetes, dementia, psychiatric disorders, chronic kidney disease (CKD), asthma, acute lower respiratory infection (ALRI), as well as mortality from all-natural and cardiorespiratory causes in the Danish nationwide administrative cohort.
We followed all residents aged ≥ 30 years (3,083,227) in Denmark from 1 January 2000 until 31 December 2017. Annual mean concentrations of fine particulate matter (PM2.5), nitrogen dioxide (NO2), black carbon (BC), and ozone (warm season) were estimated using European-wide hybrid land-use regression models (100 m × 100 m) and assigned to baseline residential addresses. We used Cox proportional hazard models to evaluate the association between air pollution and mortality, accounting for demographic and socioeconomic factors. We additionally applied indirect adjustment for smoking and body mass index (BMI).
During 47,023,454 person-years of follow-up, 803,881 people died from natural causes. Long-term exposure to PM2.5 (mean: 12.4 µg/m3), NO2 (20.3 µg/m3), and/or BC (1.0 × 10-5/m) was statistically significantly associated with all studied mortality outcomes except CKD. A 5 µg/m3 increase in PM2.5 was associated with higher mortality from all-natural causes (hazard ratio 1.11; 95% confidence interval 1.09–1.13), cardiovascular disease (1.09; 1.07–1.12), respiratory disease (1.11; 1.07–1.15), lung cancer (1.19; 1.15–1.24), diabetes (1.10; 1.04–1.16), dementia (1.05; 1.00–1.10), psychiatric disorders (1.38; 1.27–1.50), asthma (1.13; 0.94–1.36), and ALRI (1.14; 1.09–1.20). Associations with long-term exposure to ozone (mean: 80.2 µg/m3) were generally negative but became significantly positive for several endpoints in two-pollutant models. Generally, associations were attenuated but remained significant after indirect adjustment for smoking and BMI.
Long-term exposure to PM2.5, NO2, and/or BC in Denmark were associated with mortality beyond cardiorespiratory diseases, including diabetes, dementia, psychiatric disorders, asthma, and ALRI.
Abstract
Due to the wealth of exposome data from longitudinal cohort studies that is currently available, the need for methods to adequately analyze these data is growing. We propose an approach in ...which machine learning is used to identify longitudinal exposome-related predictors of health, and illustrate its potential through an application. Our application involves studying the relation between exposome and self-perceived health based on the 30-year running Doetinchem Cohort Study. Random Forest (RF) was used to identify the strongest predictors due to its favorable prediction performance in prior research. The relation between predictors and outcome was visualized with partial dependence and accumulated local effects plots. To facilitate interpretation, exposures were summarized by expressing them as the average exposure and average trend over time. The RF model’s ability to discriminate poor from good self-perceived health was acceptable (Area-Under-the-Curve = 0.707). Nine exposures from different exposome-related domains were largely responsible for the model’s performance, while 87 exposures seemed to contribute little to the performance. Our approach demonstrates that ML can be interpreted more than widely believed, and can be applied to identify important longitudinal predictors of health over the life course in studies with repeated measures of exposure. The approach is context-independent and broadly applicable.
Self-perceived general health (SPGH) is a general health indicator commonly used in epidemiological research and is associated with a wide range of exposures from different domains. However, most ...studies on SPGH only investigated a limited set of exposures and did not take the entire external exposome into account. We aimed to develop predictive models for SPGH based on exposome datasets using machine learning techniques and identify the most important predictors of poor SPGH status.
Random forest (RF) was used on two datasets based on personal characteristics from the 2012 and 2016 editions of the Dutch national health survey, enriched with environmental and neighborhood characteristics. Model performance was determined using the area under the curve (AUC) score. The most important predictors were identified using a variable importance procedure and individual effects of exposures using partial dependence and accumulated local effect plots. The final 2012 dataset contained information on 199,840 individuals and 81 variables, whereas the final 2016 dataset had 244,557 individuals with 91 variables.
Our RF models had overall good predictive performance (2012: AUC = 0.864 (CI: 0.852-0.876); 2016: AUC = 0.890 (CI: 0.883-0.896)) and the most important predictors were "Control of own life", "Physical activity", "Loneliness" and "Making ends meet". Subjects who felt insufficiently in control of their own life, scored high on the De Jong-Gierveld loneliness scale or had difficulty in making ends meet were more likely to have poor SPGH status, whereas increased physical activity per week reduced the probability of poor SPGH. We observed associations between some neighborhood and environmental characteristics, but these variables did not contribute to the overall predictive strength of the models.
This study identified that within an external exposome dataset, the most important predictors for SPGH status are related to mental wellbeing, physical exercise, loneliness, and financial status.
Long-term exposure to air pollution and noise is detrimental to health; but studies that evaluated both remain limited. This study explores associations with natural and cause-specific mortality for ...a range of air pollutants and transportation noise.
Over 4 million adults in Switzerland were followed from 2000 to 2014. Exposure to PM
, PM
components (Cu, Fe, S and Zn), NO
, black carbon (BC) and ozone (O
) from European models, and transportation noise from source-specific Swiss models, were assigned at baseline home addresses. Cox proportional hazards models, adjusted for individual and area-level covariates, were used to evaluate associations with each exposure and death from natural, cardiovascular (CVD) or non-malignant respiratory disease. Analyses included single and two exposure models, and subset analysis to study lower exposure ranges.
During follow-up, 661,534 individuals died of natural causes (36.6% CVD, 6.6% respiratory). All exposures including the PM
components were associated with natural mortality, with hazard ratios (95% confidence intervals) of 1.026 (1.015, 1.038) per 5 µg/m
PM
, 1.050 (1.041, 1.059) per 10 µg/m
NO
, 1.057 (1.048, 1.067) per 0.5 × 10
/m BC and 1.045 (1.040, 1.049) per 10 dB Lden total transportation noise. NO
, BC, Cu, Fe and noise were consistently associated with CVD and respiratory mortality, whereas PM
was only associated with CVD mortality. Natural mortality associations persisted < 20 µg/m
for PM
and NO
, < 1.5 10
/m BC and < 53 dB Lden total transportation noise. The O
association was inverse for all outcomes. Including noise attenuated all outcome associations, though many remained significant. Across outcomes, noise was robust to adjustment to air pollutants (e.g. natural mortality 1.037 (1.033, 1.042) per 10 dB Lden total transportation noise, after including BC).
Long-term exposure to air pollution and transportation noise in Switzerland contribute to premature mortality. Considering co-exposures revealed the importance of local traffic-related pollutants such as NO
, BC and transportation noise.
Long-term exposure to outdoor air pollution increases the risk of cardiovascular disease, but evidence is unclear on the health effects of exposure to pollutant concentrations lower than current EU ...and US standards and WHO guideline limits. Within the multicentre study Effects of Low-Level Air Pollution: A Study in Europe (ELAPSE), we investigated the associations of long-term exposures to fine particulate matter (PM2·5), nitrogen dioxide (NO2), black carbon, and warm-season ozone (O3) with the incidence of stroke and acute coronary heart disease.
We did a pooled analysis of individual data from six population-based cohort studies within ELAPSE, from Sweden, Denmark, the Netherlands, and Germany (recruited 1992–2004), and harmonised individual and area-level variables between cohorts. Participants (all adults) were followed up until migration from the study area, death, or incident stroke or coronary heart disease, or end of follow-up (2011–15). Mean 2010 air pollution concentrations from centrally developed European-wide land use regression models were assigned to participants’ baseline residential addresses. We used Cox proportional hazards models with increasing levels of covariate adjustment to investigate the association of air pollution exposure with incidence of stroke and coronary heart disease. We assessed the shape of the concentration-response function and did subset analyses of participants living at pollutant concentrations lower than predefined values.
From the pooled ELAPSE cohorts, data on 137 148 participants were analysed in our fully adjusted model. During a median follow-up of 17·2 years (IQR 13·8–19·5), we observed 6950 incident events of stroke and 10 071 incident events of coronary heart disease. Incidence of stroke was associated with PM2·5 (hazard ratio 1·10 95% CI 1·01–1·21 per 5 μg/m3 increase), NO2 (1·08 1·04–1·12 per 10 μg/m3 increase), and black carbon (1·06 1·02–1·10 per 0·5 10−5/m increase), whereas coronary heart disease incidence was only associated with NO2 (1·04 1·01–1·07). Warm-season O3 was not associated with an increase in either outcome. Concentration-response curves indicated no evidence of a threshold below which air pollutant concentrations are not harmful for cardiovascular health. Effect estimates for PM2·5 and NO2 remained elevated even when restricting analyses to participants exposed to pollutant concentrations lower than the EU limit values of 25 μg/m3 for PM2·5 and 40 μg/m3 for NO2.
Long-term air pollution exposure was associated with incidence of stroke and coronary heart disease, even at pollutant concentrations lower than current limit values.
Health Effects Institute.
Ambient air pollution exposure has been linked to mortality from chronic cardiorespiratory diseases, while evidence on respiratory infections remains more limited.
We examined the association between ...long-term exposure to air pollution and pneumonia-related mortality in adults in a pool of eight European cohorts.
Within the multicenter project ELAPSE (Effects of Low-Level Air Pollution: A Study in Europe), we pooled data from eight cohorts among six European countries. Annual mean residential concentrations in 2010 for fine particulate matter, nitrogen dioxide (NO
), black carbon (BC), and ozone were estimated using Europe-wide hybrid land-use regression models. We applied stratified Cox proportional hazard models to investigate the associations between air pollution and pneumonia, influenza, and acute lower respiratory infections (ALRI) mortality.
Of 325,367 participants, 712 died from pneumonia and influenza combined, 682 from pneumonia, and 695 from ALRI during a mean follow-up of 19.5 years. NO
and BC were associated with 10-12% increases in pneumonia and influenza combined mortality, but 95% confidence intervals included unity (hazard ratios, 1.12 0.99-1.26 per 10 μg/m
for NO
; 1.10 0.97-1.24 per 0.5 10
m
for BC). Associations with pneumonia and ALRI mortality were almost identical. We detected effect modification suggesting stronger associations with NO
or BC in overweight, employed, or currently smoking participants compared with normal weight, unemployed, or nonsmoking participants.
Long-term exposure to combustion-related air pollutants NO
and BC may be associated with mortality from lower respiratory infections, but larger studies are needed to estimate these associations more precisely.