General Comments

One objective of this study is to make recommendations on the use of an "Acute Health Effect Index" (AHEI) to quantify the effect of ambient air pollution on the public health. No previous studies have attempted to extend their findings in this direction. Hence there is no standardized method for our reference. For regulatory purposes, however, some form of quantitative risk assessment (QRA) is often necessary. Moreover, because the results are expressed as a single number (e.g., excess number of cases of certain diseases for a level of pollutant exposure), they give the appearance of scientific certainty and simplicity, both of which make the methods appealing to decision-makers. In practice, however, the ability to quantify the health effects is often limited and valid methods of risk assessment are both complex and uncertain. Methods of QRA are highly dependent on a series of assumptions and subjective choices which can have critical effects on the resulting risk estimates. Considerable care is therefore necessary in both using and interpreting results of QRA. (Briggs et al., 1996)

Empirically, some form of index of health effects can be derived from the partial regression coefficients (b) of the various pollutants and their relative risks from the model. The partial regression coefficient provides information on the magnitude of change of the health outcome (in terms of number of hospital admissions or mortalities) per unit change in the level of the individual air pollutant. The relative risk or risk ratio (RR) refers to the ratio of the health outcome at a certain air pollutant level to that of a reference level. In a Poisson regression model, this corresponds to the anti-logn of . In this study, the reported RR of each air pollutant was calculated by taking the anti-logn of 100 x to denote the proportional increase in hospital admissions or deaths for every 100ug.m-3 increase in the level of that pollutant.

In deriving a health outcome based index of pollution, the following points have to be carefully considered. The numerical risk estimate is affected by the model chosen (e.g., Poisson regression using generalized estimating equations or GEE, quasi-likelihood estimation, maximum likelihood estimation etc.) and the presence of unknown confounders. The RR of a single air pollutant alone (based on the single pollutant model) cannot be used to quantify overall health risk. Also, summing the separate estimates of the health effects of individual pollutants would exaggerate the estimated effect. On the other hand, using a multiple pollutant model for two highly correlated pollutants will subsume the significance of the "weaker" pollutant. Moreover, there are interactions between pollutants (e.g., between NO2, O3, and RSP, and O3 and SO2 in the elderlies, shown in the multiple pollutant model), and in evaluating the health effect of air pollutants in a "real life" situation, one needs to investigate whether the "joint effects", if present, are additive, multiplicative or even antagonistic. There is also the possibility of a non-linear health effect (and a change in ) at different ranges of pollutant levels and temperatures. Finally, the presence of population heterogeneity (e.g., children, the elderly) imply that risk estimates for these groups have to be performed separately.

Health Benefits of Reduced Air Pollutant Levels

At the risk of over-simplification, the following expresses the health benefit of reduced ambient air pollutants in a non-technical language:

Based on the risk estimates using the single pollutant model, a reduction of 100 ug.m-3 of ambient O3 could be paralleled by a 28% reduction in hospital admissions for respiratory diseases (and we are 95% sure that this decrease ranges from 22% to 34%), and a 33% reduction of hospital admissions for asthma (and we are 95% sure that this decrease ranges from 21% to 43%). A reduction in ambient NO2 by the same amount would result in a 27% fall in hospital admissions for respiratory diseases (and we are 95% sure that this decrease ranges from 21% to 33%), and a 36% reduction of hospital admissions for asthma (and we are 95% sure that this decrease ranges from 24% to 46%). A similar reduction in ambient RSP results in a 22% fall in hospital admissions for respiratory diseases and also for asthma (and we are 95% sure that this decrease ranges from 17% to 27% in the former and from 11% to 33% in the latter disease). By contrast, a reduction in SO2 levels is associated with a smaller (12%) fall in hospital admissions for respiratory diseases, but still a substantial (21%) reduction in hospital admissions for asthma.

These health benefits are even more obvious in terms of the reduction of hospital deaths due to respiratory diseases. For ambient O3, a reduction of 100 ug.m-3 of results in a 55% fall in deaths from respiratory diseases. For NO2, we should see a 28% fall and for RSP, a 17% fall. For NO2, SO2 and O3, somewhat smaller decreases in the hospital admissions for cardiovascular diseases are observed. All these decreases are unlikely to have occurred by chance alone, indicating that the observed associations between air pollutants and health effects are probably true, and there are other epidemiological and animal studies which suggest that these associations have a cause-effect relationship. The magnitude of these health benefits for each air pollutant, however, were estimated without taking into consideration the simultaneous (synergistic in some, antagonistic in others) effects on exerted by the other air pollutants.

A special warning note should be made when the datasets for 1996 were analyzed. Levels of air pollutants, especially for O3 were generally higher compared with the previous two years. Also, the long time term trend of hospital admissions was rising. Although the latter may reflect a steady increase in the number of hospital beds and in the population size, there is an urgency to consider implementing measures to lower the levels of air pollutants in general and O3 in particular, as the RR of hospital admissions and deaths for O3 are the highest among all pollutants.