Röösli Martin, Künzli Nino, Braun-Fahrländer Charlotte, Egger Matthias
Department of Social and Preventive Medicine, University of Bern, Switzerland.
Int J Epidemiol. 2005 Oct;34(5):1029-35. doi: 10.1093/ije/dyi106. Epub 2005 May 23.
There is debate on how the effect of air pollution should be assessed. We propose an approach to estimate its impact on adult and infant mortality that integrates data from long-term epidemiological studies and studies of interventions to reduce pollution. We use the method to estimate the number of years of life lost (YLLs) attributable to air pollution during 1 year in Switzerland.
A dynamic exposure-response model was implemented, which uses an exponential function (exp(-kt)) to model the change in mortality after cessation of air pollution. The model was populated with relative risk estimates and estimates of time constant k from the literature. Air pollution exposure in Switzerland was modelled using data from emission inventories. YLLs attributable to air pollution were calculated by taking the difference between observed survival probabilities in Switzerland in 2000 and modified survival probabilities, assuming no air pollution during the year 2000.
Meta-analyses of three studies of adult mortality and five studies of infant mortality gave relative risks of 1.059 (95% confidence interval (CI) 1.031-1.088) and 1.056 (95% CI 1.026-1.088) per 10 mug/m(3) increase in PM10 concentration. Time constants k derived from two studies of the effects of the closing down of a steel mill in the Utah Valley and of the coal ban in Dublin were 0.88 and 0.11. Assuming a time constant k of 0.5 resulted in 42 400 (95% CI 22 600-63 600) YLLs, with 4.0% being ascribed to infant deaths. A total of 39% of the effect occurred in the same year and 80% within 5 years. The estimated number of YLLs was little affected by the choice of the time constant.
In contrast to traditional steady-state models the dynamic model allows changes in mortality following short-term increases or decreases in air pollution levels to be quantified. This type of information is of obvious interest to policy makers.
关于如何评估空气污染的影响存在争议。我们提出了一种方法来估计其对成人和婴儿死亡率的影响，该方法整合了长期流行病学研究和减少污染干预措施研究的数据。我们使用该方法来估计瑞士一年内空气污染导致的寿命损失年数（YLLs）。
实施了一个动态暴露-反应模型，该模型使用指数函数（exp(-kt)）来模拟空气污染停止后死亡率的变化。该模型用文献中的相对风险估计值和时间常数k的估计值进行填充。瑞士的空气污染暴露情况使用排放清单数据进行建模。通过计算2000年瑞士观察到的生存概率与假设2000年无空气污染情况下的修正生存概率之间的差异，得出空气污染导致的YLLs。
对三项成人死亡率研究和五项婴儿死亡率研究的荟萃分析得出，PM10浓度每增加10微克/立方米，相对风险分别为1.059（95%置信区间（CI）1.031 - 1.088）和1.056（95%CI 1.026 - 1.088）。从犹他山谷一家钢铁厂关闭和都柏林煤炭禁令影响的两项研究中得出的时间常数k分别为(0.88)和(0.11)。假设时间常数k为(0.5)，则导致42400（95%CI 22600 - 63600）个YLLs，其中4.0%归因于婴儿死亡。总影响的39%发生在同一年，80%在5年内发生。YLLs的估计数量受时间常数选择的影响较小。
与传统的稳态模型不同，动态模型能够量化空气污染水平短期上升或下降后死亡率的变化。这类信息对政策制定者显然具有重要意义。
