Comparison of time series and case-crossover analyses of air pollution and hospital admission data

[Contribution of case-crossover design to the analysis of short-term health effects of air pollution: reanalysis of air pollution and health data]

BACKGROUND

During the last decades, numerous studies have shown significant links between short-term exposure to air pollution and health. Time series design have been widely used in order to study these associations. In recent years, the case-crossover design has been applied to the analysis of acute effects of environmental exposures, especially air pollution. The aims of this paper are to describe the case-crossover design and to compare this approach with time series design to assess the association between air pollution and health.

METHODS

In the case-crossover approach, a case-control study is conducted whereby each person who had a health event (case period) is matched with himself/herself on a nearby time period where he/she did not have the event (control period). Review of the literature shows that the referent selection strategies can be associated to a bias in the estimation of the health effect of air pollution. In comparison with time series design, the case-crossover design is easier to conduct, and individual factors can be taken into account. Nevertheless, it is not possible to take into account the overdispersion in the health indicator with this approach.

RESULTS AND CONCLUSION

In conclusion, we suggest to use time series analysis with population data and case-crossover design with individual data.

Short-Term Effects of Particulate Air Pollution on Male Smokers and Never-Smokers

BACKGROUND

Numerous studies have shown that ambient air pollution and smoking are both associated with increased mortality, but until now there has been little evidence as to whether the effects of these 2 factors combined are greater than the sum of their individual effects. We assessed whether smokers are subject to additional mortality risk from air pollution relative to never-smokers.

METHODS

This study included 10,833 Chinese men in Hong Kong who died at the age of 30 or above during the period 1 January to 31 December 1998. Relatives who registered for deceased persons were interviewed about the deceased's smoking history and other personal lifestyle factors about 10 years before death. Poisson regression for daily number of deaths was fitted to estimate excess risks per 10 microg/m increase in particulate matter with aerodynamic diameter <10 microm (PM10) in male smokers and never-smokers in stratified data, and additional excess risk for smokers relative to never-smokers in combined data.

RESULTS

In smokers there was a significant excess risk associated with PM10 for all natural causes and cardio-respiratory diseases for men age 30 years or older and men 65 or older. For all natural causes, greater excess risk associated with PM10 was observed for smokers relative to never-smokers: 1.9% (95% confidence interval = 0.3% to 3.6%) in men age 30 and older and 2.3% (0.4% to 4.3%) in those age 65 and older.

CONCLUSIONS

Ambient particulate air pollution is associated with greater excess mortality in male smokers compared with never-smokers.

Air pollution and postneonatal mortality in a tropical city: Kaohsiung, Taiwan

With growing evidence of the association between daily mortality and air pollution in adults, it is important to investigate whether infants are also susceptible to the adverse health effects of ambient air pollutants. The purpose of this study is to examine the relationship between air pollution and postneonatal mortality in Kaohsiung, Taiwan, a large industrial city with a tropical climate, during the period 1994-2000, using a case-crossover analysis. Case-crossover analysis provides an alternative to Poisson time-series regression for studying the short-term adverse health effects of air pollution. The air pollutants examined included particulate matter (PM(10)), sulfur dioxide (SO(2)), ozone (O(3)), nitrogen dioxide (NO(2)), and carbon monoxide (CO). The risk of postneonatal deaths was estimated to increase by 4.0% per 67 microg/m(3) (the interquartile range in daily ambient concentration of PM(10)) for PM(10), 1.8% per 17.84 ppb for NO(2), 5.1% per 0.31 ppm for CO, and 4.6% per 19.20 ppb for O(3). Although positive, none of these associations achieved statistical significance. The established link between air pollution levels and infant mortality may not be as strong in cities with tropical climates, although other factors such as differences in pollutant mix or the underlying health of the postneonates may explain the lack of a strong association in this study. Further studies of this type in cities with varying climates and cultures are needed.

Bayesian model averaging in time-series studies of air pollution and mortality

The issue of model selection in time-series studies assessing the acute health effects from short-term exposure to ambient air pollutants has received increased scrutiny in the past 5 yr. Recently, Bayesian model averaging (BMA) has been applied to allow for uncertainty about model form in assessing the association between mortality and ambient air pollution. While BMA has the potential to allow for such uncertainties in risk estimates, Bayesian approaches in general and BMA in particular are not panaceas for model selection., Since misapplication of Bayesian methods can lead to erroneous conclusions, model selection should be informed by substantive knowledge about the environmental health processes influencing the outcome. This paper examines recent attempts to use BMA in air pollution studies to illustrate the potential benefits and limitations of the method.

Health effects associated with exposure to ambient air pollution

The World Health Organization has identified ambient air pollution as a high public health priority, based on estimates of air pollution related death and disability-adjusted life years derived in its Global Burden of Disease initiative. The NERAM Colloquium Series on Health and Air Quality was initiated to strengthen the linkage between scientists, policymakers, and other stakeholders by reviewing the current state of science, identifying policy-relevant gaps and uncertainties in the scientific evidence, and proposing a path forward for research and policy to improve air quality and public health. The objective of this paper is to review the current state of science addressing the impacts of air pollution on human health. The paper is one of four background papers prepared for the 2003 NERAM/AirNet Conference on Strategies for Clean Air and Health, the third meeting in the international Colloquium Series. The review is based on the framework and findings of the U.S. National Research Committee (NRC) on Research Priorities for Airborne Particulate Matter and addresses key questions underlying air quality risk management policy decisions.

The impact of 9/11 on the association of ambient air pollution with daily respiratory hospital admissions in a Canada-US border city, Windsor, Ontario

The 11 September 2001 (9/11) terrorist attacks in the United States resulted in long lines of trucks at the border crossing in Windsor, Ontario. Public concern about the potential impact of these trucks spewing toxic pollutants into the air drew attention to the need to investigate the impact of 9/11 on the daily levels of air pollutants and respiratory hospitalization. In this study, significant increases in respiratory admissions were found one month and 6 months post-9/11. Mean daily respiratory admission was also significantly higher than the same period one year earlier and one year later. SO(2) and CO concentration levels were found to be generally higher after 9/11 than one year before and immediately before. Relative risk estimates of respiratory hospitalization after 9/11 showed that SO(2) (RR̂ = 1.15 for two-day, RR̂ = 1.18 for three-day, and RR̂ = 1.21 for five-day averages), NO(2) (RR̂ = 1.10 for current day), and COH (RR̂ = 1.09 for current day, RR̂ = 1.10 for two-day average) had the most significant effects after 9/11. These results suggest the need for more stringent regulatory efforts in air quality in the region in response to the changing transportation dynamics at this Canada-US border crossing.

A comparison of methods for the analysis of recurrent health outcome data with environmental covariates

Recurrent events such as repeated hospital admissions for the same health outcome occur frequently in environmental health studies. Dewanji and Moolgavkar proposed a flexible parametric model and a conditional likelihood analysis for recurrent events based on a Poisson process formulation. In this paper, we examine the statistical properties of the Dewanji-Moolgavkar (DM) estimator of the risk of an adverse health outcome associated with environmental exposures based on recurrent event data using computer simulation. We also compare the DM approach with both case-crossover analysis for multiple observations and time series analysis when there are no subject-specific covariates. When using a correctly specified model, the DM method produced better estimates with respect to relative mean square error when each subject had constant or curved baseline intensity functions than it did when baseline intensities were increasing or decreasing in a linear fashion. For under-specified models, the DM method outperformed case-crossover analysis for decreasing straight line intensity functions, was outperformed by case-crossover analysis for increasing straight line intensity functions, and was roughly equivalent to case-crossover analysis for constant and curved intensity functions. Case-crossover analysis produced superior risk estimates more frequently than the other two methods in the cases considered here, especially for linear representations of the baseline intensities.

Case???Crossover Analyses of Air Pollution Exposure Data

The case-crossover design has been widely used to study the association between short-term air pollution exposure and the risk of an acute adverse health event. The design uses cases only; for each individual case, exposure just before the event is compared with exposure at other control (or "referent") times. Time-invariant confounders are controlled by making within-subject comparisons. Even more important in the air pollution setting is that time-varying confounders can also be controlled by design by matching referents to the index time. The referent selection strategy is important for reasons in addition to control of confounding. The case-crossover design makes the implicit assumption that there is no trend in exposure across the referent times. In addition, the statistical method that is used-conditional logistic regression-is unbiased only with certain referent strategies. We review here the case-crossover literature in the air pollution context, focusing on key issues regarding referent selection. We conclude with a set of recommendations for choosing a referent strategy with air pollution exposure data. Specifically, we advocate the time-stratified approach to referent selection because it ensures unbiased conditional logistic regression estimates, avoids bias resulting from time trend in the exposure series, and can be tailored to match on specific time-varying confounders.

A call for reporting the relevant exposure term in air pollution case-crossover studies

The exposure term in the case-crossover design consists in the difference between the ambient concentration on the event day and the concentration(s) on some control day(s). So far, all air pollution case-crossover studies presented the distribution of the daily ambient pollutant concentrations but do not publish the distributional properties of the relevant exposure term--that is, the concentration difference. This article shows that this difference can be very small for a large fraction of event days, therefore, seriously limiting the statistical power to refute the null hypothesis. Publishing the distribution of the relevant differences will improve the interpretation and discussion of findings from case-crossover studies, particularly in cases with statistically non-significant associations.

Association of ambient air pollution with respiratory hospitalization in a government-designated "area of concern": the case of Windsor, Ontario

This study is part of a larger research program to examine the relationship between ambient air quality and health in Windsor, Ontario, Canada. We assessed the association between air pollution and daily respiratory hospitalization for different age and sex groups from 1995 to 2000. The pollutants included were nitrogen dioxide, sulfur dioxide, carbon monoxide, ozone, particulate matter 10 microm in diameter (PM10), coefficient of haze (COH), and total reduced sulfur (TRS). We calculated relative risk (RR) estimates using both time-series and case-crossover methods after controlling for appropriate confounders (temperature, humidity, and change in barometric pressure). The results of both analyses were consistent. We found associations between NO2, SO2, CO, COH, or PM10 and daily hospital admission of respiratory diseases especially among females. For females 0-14 years of age, there was 1-day delayed effect of NO2 (RR = 1.19, case-crossover method), a current-day SO2 (RR = 1.11, time series), and current-day and 1- and 2-day delayed effects for CO by case crossover (RR = 1.15, 1.19, 1.22, respectively). Time-series analysis showed that 1-day delayed effect of PM10 on respiratory admissions of adult males (15-64 years of age), with an RR of 1.18. COH had significant effects on female respiratory hospitalization, especially for 2-day delayed effects on adult females, with RRs of 1.15 and 1.29 using time-series and case-crossover analysis, respectively. There were no significant associations between O3 and TRS with respiratory admissions. These findings provide policy makers with current risks estimates of respiratory hospitalization as a result of poor ambient air quality in a government designated "area of concern."

Forecasting model for the incidence of hepatitis A based on artificial neural network

AIM: To study the application of artificial neural network (ANN) in forecasting the incidence of hepatitis A, which had an autoregression phenomenon.

METHODS: The data of the incidence of hepatitis A in Liaoning Province from 1981 to 2001 were obtained from Liaoning Disease Control and Prevention Center. We used the autoregressive integrated moving average (ARIMA) model of time series analysis to determine whether there was any autoregression phenomenon in the data. Then the data of the incidence were switched into [0,1] intervals as the network theoretical output. The data from 1981 to 1997 were used as the training and verifying sets and the data from 1998 to 2001 were made up into the test set. STATISTICA neural network (ST NN) was used to construct, train and simulate the artificial neural network.

RESULTS: Twenty-four networks were tested and seven were retained. The best network we found had excellent performance, its regression ratio was 0.73, and its correlation was 0.69. There were 2 input variables in the network, one was AR(1), and the other was time. The number of units in hidden layer was 3. In ARIMA time series analysis results, the best model was first order autoregression without difference and smoothness. The total sum square error of the ANN model was 9090.21, the sum square error of the training set and testing set was 8377.52 and 712.69, respectively, they were all less than that of ARIMA model. The corresponding value of ARIMA was 12291.79, 8944.95 and 3346.84, respectively. The correlation coefficient of nonlinear regression (R_NL) of ANN was 0.71, while the R_NL of ARIMA linear autoregression model was 0.66.

CONCLUSION: ANN is superior to conventional methods in forecasting the incidence of hepatitis A which has an autoregression phenomenon.