Elsevier

Atmospheric Environment

Volume 65, February 2013, Pages 171-176
Atmospheric Environment

Which metric of ambient ozone to predict daily mortality?

https://doi.org/10.1016/j.atmosenv.2012.10.032Get rights and content

Abstract

It is well known that ozone concentration is associated with daily cause specific mortality. But which ozone metric is the best predictor of the daily variability in mortality?

We performed a time series analysis on daily deaths (all causes, respiratory and cardiovascular causes as well as death in elderly 65+) in Vienna for the years 1991–2009. We controlled for seasonal and long term trend, day of the week, temperature and humidity using the same basic model for all pollutant metrics. We found model fit was best for same day variability of ozone concentration (calculated as the difference between daily hourly maximum and minimum) and hourly maximum. Of these the variability displayed a more linear dose–response function. Maximum 8 h moving average and daily mean value performed not so well. Nitrogen dioxide (daily mean) in comparison performed better when previous day values were assessed. Same day ozone and previous day nitrogen dioxide effect estimates did not confound each other.

Variability in daily ozone levels or peak ozone levels seem to be a better proxy of a complex reactive secondary pollutant mixture than daily average ozone levels in the Middle European setting. If this finding is confirmed this would have implications for the setting of legally binding limit values.

Highlights

► We performed a time-series analysis over 19 years of ozone and daily mortality. ► Within-day variability of ozone was the best predictors of same day mortality. ► Previous day nitrogen dioxide did not confound the results.

Introduction

The adverse health effects of the gaseous air pollutant ozone are well established (WHO, 2006, Bates, 2005), but the magnitude of the effect of an incremental change in ozone concentration is still controversial (Smith et al., 2009). Goodman (2005) pointed out that the choice of different averaging times in the studies of Bell et al., 2005, Levy et al., 2005, and Ito et al. (2005) might partly explain the differences in effect estimates. But other methodological differences in the three papers preclude firm conclusions as to the impact of the choice of different ozone metrics. Also Fann et al. (2011) underline the concern that the optimal ozone metric still has to be determined.

We have previously reported an increase in daily mortality in Vienna, Austria, due to particulate air pollution in a time series analysis 1999–2004 (Neuberger et al., 2007). For the sake of comparability the methodology of this study followed that of the APHEA project (Samoli et al., 2003) which performed time series analyses in a range of European cities. In fact, the Viennese results were broadly in accordance with those from APHEA and similar studies: The increase of (all causes) daily deaths per 10 μg m−3 increase of PM10 (average of same and previous day) amounted to 0.2% (CI = −0.4; 0.7) and was higher for specific causes of death and when longer lags (up to 14 days) were considered. Daily maximum ozone 8 h moving average was analysed as well for the same period but no consistent effect was found even when the analysis was restricted to the summer seasons. This null finding for ozone effects is not uncommon in the Austrian situation where ozone both temporally and spatially is negatively correlated with the indicators of traffic related air contaminations PM10, CO, NO, and NO2. Under equilibrium conditions between NOX and ozone and no other precursor substances there is no net production of ozone and during the day with enough solar energy to fuel the reaction positively its concentration is correlated with NOX. However, in the presence of precursors like CO, methane or other carbohydrates there is an additional production of ozone that leads to an increase of ozone while NO2 is decreasing due to photo dissociation or remains stable. NO2 proved to be a good predictor of health effects in several studies in Austria (Neuberger et al., 2007, Neuberger et al., 2002, Moshammer et al., 2006).

Ozone is an irritant gas with potent oxidative capacity. It also serves as an indicator of a reactive mixture of secondary pollutants that includes also very short-lived radicals that are not easily directly monitored. The association between ozone and these radicals is not constant. In relatively clean air ozone formed under the influence of UV radiation is quite stable and only few radicals are generated while in more polluted air the generation and consumption of ozone is accelerated and thus more radicals are formed (Reid et al., 1996, Solomon et al., 2000).

Therefore, this study sets out to test the hypothesis that daily peak levels of ozone or even the difference between night time low and day time high ozone are better predictors of health effects than daily mean values or even maximal 8 h moving averages.

In fact the idea of that study resulted from a meeting of a working group of the Austrian Academy of Sciences that was charged with the drafting of a guideline document regarding air quality requirements for Austrian spas. In spas health-seeking people also expect good air quality. But in some well-known alpine mountain spas daily average ozone levels are sometimes rather high during summer month in spite or even because of otherwise very clean air. It was hypothesised that higher average ozone levels (measured as daily mean levels) with less daily variation are not as dangerous as the same average levels but with higher daily fluctuations.

Section snippets

Material and methods

Mortality data for the years 1991–2009 were provided by Statistics Austria. For each death in the database the following data were extracted: Day of death, age of deceased person (in completed years of life), sex, district of home address, primary cause of death: before 2002 according to ICD9 (WHO, 1977), starting with 2002 according to ICD10 (WHO, 1992). Analysis was restricted to all cases with a home address district in Vienna, the capital city of Austria with approx. 1.5 million inhabitants

Results

The monitoring station ZA performed best of all 4 stations for all metrics except for SO2 (Table 1). For the sake of consistency also for SO2 the data from ZA were used because the correlation coefficients were only slightly lower at ZA (range: 0.86–0.93) compared to the data from STE (r range: 0.89–0.93). For all metrics but one the correlation coefficients between the 4 stations were very high. Only NO2 displayed a somewhat stronger spatial heterogeneity.

Daily overall mortality in Vienna

Discussion

In this time series over nearly 19 years O3 and NO2 were adversely associated with daily mortality. The effect estimates were comparable with previous findings (Goldberg et al., 2001, Gryparis et al., 2004, Anderson et al., 2004, Bell et al., 2004, Bell et al., 2005, Bell et al., 2006, Levy et al., 2005, Ito et al., 2005, Parodi et al., 2005, Zhang et al., 2006, Zanobetti and Schwartz, 2008, Garrett and Casimiro, 2011). Table 3, Table 4 provide effect estimates per 10 μg m−3. For a 10 μg m−3

Conclusion

The results of this analysis indicate that the daily variability of ozone or the daily maximal ozone concentration are better predictors of short term health effects than daily mean values. Legally defined health based limit values for ozone (e.g. European Air Quality Directive: EP, 2008) are often based on 8-h moving average data. Should the findings of this investigation be confirmed for other regions these recommendations may have to be reconsidered.

Disclosure statement

The authors declare no conflict of interest. The data for this study were obtained from routine data sources. Therefore no financial support was needed and no ethical considerations regarding clinical studies were applicable.

References (31)

  • C. Chen et al.

    Assessing the influence of indoor exposure to “outdoor ozone” on the relationship between ozone and short-term mortality in U.S. communities

    Environmental Health Perspectives

    (2012)
  • F. Dominici et al.

    On the use of generalized additive models in time-series studies of air pollution and health

    American Journal of Epidemiology

    (2002)
  • F. Dominici et al.

    Issues in Semi-parametric Regression with Applications in Time Series Models for Air Pollution and Mortality

    (2003)
  • European Parliament, 2008. Directive 2008/50/EC of the European Parliament and of the Council of 21 May 2008 on Ambient...
  • N. Fann et al.

    Improving the linkages between air pollution epidemiology and quantitative risk assessment

    Environmental Health Perspectives

    (2011)
  • Cited by (14)

    • Influence of temperature to the short-term effects of various ozone metrics on daily mortality in Suzhou, China

      2013, Atmospheric Environment
      Citation Excerpt :

      In this time-series analysis, we used various ozone metrics to explore the short-term effects of O3 on daily mortality. The effect estimates of the O3-mortality relationship were consistent with previous studies (Moshammer et al., 2013; Peng et al., 2012; Tao et al., 2011; Wong et al., 2008), see the Supplementary Material, Fig. S9. Compared with the APHENA study, the percent increase (95%CI) in all-cause non-accidental mortality associated with an increase of 10 μg m−3 in 1 h maximum O3 concentrations were 0.21%(−0.08%, 0.529%) in our study, 0.85%(0.51%, 1.20%) in the Canada study, 0.18%(0.07%, 0.30%) in the Europe study, and 0.31%(0.09%, 0.52%) in the U.S. study (Peng et al., 2012) (Fig. S9).

    View all citing articles on Scopus
    View full text