The systematic evaluation of the results of time-series studies of air pollution is challenged by differences in model specification and publication bias.
We evaluated the associations of inhalable ...particulate matter (PM) with an aerodynamic diameter of 10 μm or less (PM
) and fine PM with an aerodynamic diameter of 2.5 μm or less (PM
) with daily all-cause, cardiovascular, and respiratory mortality across multiple countries or regions. Daily data on mortality and air pollution were collected from 652 cities in 24 countries or regions. We used overdispersed generalized additive models with random-effects meta-analysis to investigate the associations. Two-pollutant models were fitted to test the robustness of the associations. Concentration-response curves from each city were pooled to allow global estimates to be derived.
On average, an increase of 10 μg per cubic meter in the 2-day moving average of PM
concentration, which represents the average over the current and previous day, was associated with increases of 0.44% (95% confidence interval CI, 0.39 to 0.50) in daily all-cause mortality, 0.36% (95% CI, 0.30 to 0.43) in daily cardiovascular mortality, and 0.47% (95% CI, 0.35 to 0.58) in daily respiratory mortality. The corresponding increases in daily mortality for the same change in PM
concentration were 0.68% (95% CI, 0.59 to 0.77), 0.55% (95% CI, 0.45 to 0.66), and 0.74% (95% CI, 0.53 to 0.95). These associations remained significant after adjustment for gaseous pollutants. Associations were stronger in locations with lower annual mean PM concentrations and higher annual mean temperatures. The pooled concentration-response curves showed a consistent increase in daily mortality with increasing PM concentration, with steeper slopes at lower PM concentrations.
Our data show independent associations between short-term exposure to PM
and PM
and daily all-cause, cardiovascular, and respiratory mortality in more than 600 cities across the globe. These data reinforce the evidence of a link between mortality and PM concentration established in regional and local studies. (Funded by the National Natural Science Foundation of China and others.).
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.
Estimating air pollution exposure has long been a challenge for environmental health researchers. Technological advances and novel machine learning methods have allowed us to increase the geographic ...range and accuracy of exposure models, making them a valuable tool in conducting health studies and identifying hotspots of pollution. Here, we have created a prediction model for daily PM2.5 levels in the Greater London area from 1st January 2005 to 31st December 2013 using an ensemble machine learning approach incorporating satellite aerosol optical depth (AOD), land use, and meteorological data. The predictions were made on a 1 km × 1 km scale over 3960 grid cells. The ensemble included predictions from three different machine learners: a random forest (RF), a gradient boosting machine (GBM), and a k-nearest neighbor (KNN) approach. Our ensemble model performed very well, with a ten-fold cross-validated R2 of 0.828. Of the three machine learners, the random forest outperformed the GBM and KNN. Our model was particularly adept at predicting day-to-day changes in PM2.5 levels with an out-of-sample temporal R2 of 0.882. However, its ability to predict spatial variability was weaker, with a R2 of 0.396. We believe this to be due to the smaller spatial variation in pollutant levels in this area.
Most epidemiological studies estimate associations without considering exposure measurement error. While some studies have estimated the impact of error in single-exposure models we aimed to quantify ...the effect of measurement error in multi-exposure models, specifically in time-series analysis of PM
, NO
, and mortality using simulations, under various plausible scenarios for exposure errors. Measurement error in multi-exposure models can lead to effect transfer where the effect estimate is overestimated for the pollutant estimated with more error to the one estimated with less error. This complicates interpretation of the independent effects of different pollutants and thus the relative importance of reducing their concentrations in air pollution policy.
Measurement error was defined as the difference between ambient concentrations and personal exposure from outdoor sources. Simulation inputs for error magnitude and variability were informed by the literature. Error-free exposures with their consequent health outcome and error-prone exposures of various error types (classical/Berkson) were generated. Bias was quantified as the relative difference in effect estimates of the error-free and error-prone exposures.
Mortality effect estimates were generally underestimated with greater bias observed when low ratios of the true exposure variance over the error variance were assumed (27.4% underestimation for NO
). Higher ratios resulted in smaller, but still substantial bias (up to 19% for both pollutants). Effect transfer was observed indicating that less precise measurements for one pollutant (NO
) yield more bias, while the co-pollutant (PM
) associations were found closer to the true. Interestingly, the sum of single-pollutant model effect estimates was found closer to the summed true associations than those from multi-pollutant models, due to cancelling out of confounding and measurement error bias.
Our simulation study indicated an underestimation of true independent health effects of multiple exposures due to measurement error. Using error parameter information in future epidemiological studies should provide more accurate concentration-response functions.
•Studies with information on PM2.5 and NO2 measurement error structures were reviewed.•We derived outdoor source personal exposure to compare with ambient concentrations.•Outdoor sources contribute ...44% to total personal exposure to PM2.5 and 74% for NO2.•Mean difference (measurement error) was 5.72 μg/m3 for PM2.5 and 7.17 ppb for NO2.•Error variability was also greater for NO2. Error correlation was not reported.
The use of proxy exposure estimates for PM2.5 and NO2 in air pollution studies instead of personal exposures, introduces measurement error, which can produce biased epidemiological effect estimates. Most studies consider total personal exposure as the gold standard. However, when studying the effects of ambient air pollution, personal exposure from outdoor sources is the exposure of interest.
We assessed the magnitude and variability of exposure measurement error by conducting a systematic review of the differences between personal exposures from outdoor sources and the corresponding measurements for ambient concentrations in order to increase understanding of the measurement error structures of the pollutants.
We reviewed the literature (ISI Web of Science, Medline, 2000–2016) for English language studies (in any age group in any location (NO2) or Europe and North America (PM2.5)) that reported repeated measurements over time both for personal and ambient PM2.5 or NO2 concentrations. Only a few studies reported personal exposure from outdoor sources. We also collected data for infiltration factors and time-activity patterns of the individuals in order to estimate personal exposures from outdoor sources in every study.
Studies using modelled rather than monitored exposures were excluded. Type of personal exposure monitor was assessed. Random effects meta-analysis was conducted to quantify exposure error as the mean difference between “true” and proxy measures.
Thirty-two papers for PM2.5 and 24 for NO2 were identified. Outdoor sources were found to contribute 44% (range: 33–55%) of total personal exposure to PM2.5 and 74% (range: 57–88%) to NO2. Overall estimates of personal exposure (24-hour averages) from outdoor sources were 9.3 μg/m3 and 12.0 ppb for PM2.5 and NO2 respectively, while the corresponding difference between these exposures and the ambient concentrations (i.e. the measurement error) was 5.72 μg/m3 and 7.17 ppb. Our findings indicated also higher error variability for NO2 than PM2.5. Large heterogeneity was observed which was not explained sufficiently by geographical location or age group of the study sample.
Relying only on information available in published studies led to some limitations: the contribution of outdoor sources to total personal exposure for NO2 had to be inferred, individual variation in exposure misclassification was unavailable and instrument error could not be addressed. The larger magnitude and variability of errors for NO2 compared with PM2.5 has implications for biases in the health effect estimates of multi-pollutant epidemiological models. Results suggest that further research is needed regarding personal exposure studies and measurement error bias in epidemiological models.
The present study aimed at developing a standardized heat wave definition to estimate and compare the impact on mortality by gender, age and death causes in Europe during summers 1990-2004 and 2003, ...separately, accounting for heat wave duration and intensity.
Heat waves were defined considering both maximum apparent temperature and minimum temperature and classified by intensity, duration and timing during summer. The effect was estimated as percent increase in daily mortality during heat wave days compared to non heat wave days in people over 65 years. City specific and pooled estimates by gender, age and cause of death were calculated.
The effect of heat waves showed great geographical heterogeneity among cities. Considering all years, except 2003, the increase in mortality during heat wave days ranged from + 7.6% in Munich to + 33.6% in Milan. The increase was up to 3-times greater during episodes of long duration and high intensity. Pooled results showed a greater impact in Mediterranean (+ 21.8% for total mortality) than in North Continental (+ 12.4%) cities. The highest effect was observed for respiratory diseases and among women aged 75-84 years. In 2003 the highest impact was observed in cities where heat wave episode was characterized by unusual meteorological conditions.
Climate change scenarios indicate that extreme events are expected to increase in the future even in regions where heat waves are not frequent. Considering our results prevention programs should specifically target the elderly, women and those suffering from chronic respiratory disorders, thus reducing the impact on mortality.
We studied the potential synergy between air pollution and meteorology and their impact on mortality in nine European cities with data from 2004 to 2010. We used daily series of Apparent Temperature ...(AT), measurements of particulate matter (PM
), ozone (O₃), and nitrogen dioxide (NO₂) and total non-accidental, cardiovascular, and respiratory deaths. We applied Poisson regression for city-specific analysis and random effects meta-analysis to combine city-specific results, separately for the warm and cold seasons. In the warm season, the percentage increase in all deaths from natural causes per °C increase in AT tended to be greater during high ozone days, although this was only significant for all ages when all causes were considered. On low ozone days, the increase in the total daily number of deaths was 1.84% (95% CI 0.87, 2.82), whilst it was 2.20% (95% CI 1.28, 3.13) in the high ozone days per 1 °C increase in AT. Interaction with PM
was significant for cardiovascular (CVD) causes of death for all ages (2.24% on low PM
days (95% CI 1.01, 3.47) whilst it is 2.63% (95% CI 1.57, 3.71) on high PM
days) and for ages 75+. In days with heat waves, no consistent pattern of interaction was observed. For the cold period, no evidence for synergy was found. In conclusion, some evidence of interactive effects between hot temperature and the levels of ozone and PM
was found, but no consistent synergy could be identified during the cold season.
Objective
Quantitative estimates of air pollution health impacts have become an increasingly critical input to policy decisions. The WHO project “Health risks of air pollution in Europe—HRAPIE” was ...implemented to provide the evidence-based concentration–response functions for quantifying air pollution health impacts to support the 2013 revision of the air quality policy for the European Union (EU).
Methods
A group of experts convened by WHO Regional Office for Europe reviewed the accumulated primary research evidence together with some commissioned reviews and recommended concentration–response functions for air pollutant–health outcome pairs for which there was sufficient evidence for a causal association.
Results
The concentration–response functions link several indicators of mortality and morbidity with short- and long-term exposure to particulate matter, ozone and nitrogen dioxide. The project also provides guidance on the use of these functions and associated baseline health information in the cost–benefit analysis.
Conclusions
The project results provide the scientific basis for formulating policy actions to improve air quality and thereby reduce the burden of disease associated with air pollution in Europe.
An increasing number of studies suggest adverse effects of exposure to ambient air pollution on cognitive function, but the evidence is still limited. We investigated the associations between ...long-term exposure to air pollutants and cognitive function in the English Longitudinal Study of Ageing (ELSA) cohort of older adults.
Our sample included 8,883 individuals from ELSA, based on a nationally representative study of people aged ≥ 50 years, followed-up from 2002 until 2017. Exposure to air pollutants was modelled by the CMAQ-urban dispersion model and assigned to the participants' residential postcodes. Cognitive test scores of memory and executive function were collected biennially. The associations between these cognitive measures and exposure to ambient concentrations of NO
, PM
, PM
and ozone were investigated using mixed-effects models adjusted for time-varying age, physical activity and smoking status, as well as baseline gender and level of education.
Increasing long-term exposure per interquartile range (IQR) of NO
(IQR: 13.05 μg/m
), PM
(IQR: 3.35 μg/m
) and PM
(IQR: 2.7 μg/m
) were associated with decreases in test scores of composite memory by -0.10 (95% confidence interval CI: -0.14, -0.07), -0.02 -0.04, -0.01 and -0.08 -0.11, -0.05, respectively. The same increases in NO
, PM
and PM
were associated with decreases in executive function score of -0.31 -0.38, -0.23, -0.05 -0.08, -0.02 and -0.16 -0.22, -0.10, respectively. The association with ozone was inverse across both tests. Similar results were reported for the London-dwelling sub-sample of participants.
The present study was based on a long follow-up with several repeated measurements per cohort participant and long-term air pollution exposure assessment at a fine spatial scale. Increasing long-term exposure to NO
, PM
and PM
was associated with a decrease in cognitive function in older adults in England. This evidence can inform policies related to modifiable environmental exposures linked to cognitive decline.
Heat Effects on Mortality in 15 European Cities Baccini, Michela; Biggeri, Annibale; Accetta, Gabriele ...
Epidemiology (Cambridge, Mass.),
2008-September, Letnik:
19, Številka:
5
Journal Article
Recenzirano
Odprti dostop
Background: Epidemiologic studies show that high temperatures are related to mortality, but little is known about the exposure-response function and the lagged effect of heat. We report the ...associations between daily maximum apparent temperature and daily deaths during the warm season in 15 European cities. Methods: The city-specific analyses were based on generalized estimating equations and the city-specific results were combined in a Bayesian random effects meta-analysis. We specified distributed lag models in studying the delayed effect of exposure. Time-varying coefficient models were used to check the assumption of a constant heat effect over the warm season. Results: The city-specific exposure-response functions have a V shape, with a change-point that varied among cities. The meta-analytic estimate of the threshold was 29.4°C for Mediterranean cities and 23.3°C for north-continental cities. The estimated overall change in all natural mortality associated with a 1°C increase in maximum apparent temperature above the city-specific threshold was 3.12% (95% credibility interval = 0.60% to 5.72%) in the Mediterranean region and 1.84% (0.06% to 3.64%) in the north-continental region. Stronger associations were found between heat and mortality from respiratory diseases, and with mortality in the elderly. Conclusions: There is an important mortality effect of heat across Europe. The effect is evident from June through August; it is limited to the first week following temperature excess, with evidence of mortality displacement. There is some suggestion of a higher effect of early season exposures. Acclimatization and individual susceptibility need further investigation as possible explanations for the observed heterogeneity among cities.