EXCESS MORTALITY in London during the December 1952 smog episode was clearly associated with the extremely high particulate and sulfur dioxide ([SO.sub.2]) pollution concentrations.[1] Analyses of daily mortality in subsequent winters[2] have suggested that increased daily mortality was associated with particulate and [SO.sub.2] air pollution, even during nonepisodic conditions. Several recent studies in the United States[3-5] have reported an association between increased daily mortality and particulate and [SO.sub.2] air pollution at concentrations below the World Health Organization (WHO) recommendations for air quality levels.[6] These studies have also suggested that increased daily mortality is causally associated with particulate air pollution, and is associated with [SO.sub.2] only indirectly through the correlation of [SO.sub.2] and particulate pollution concentrations.[2-5]

In the United States, the major sources of particulate and [SO.sub.2] air pollution are electric power production, industrial processes, and automobile exhaust. In many parts of the world, domestic coal burning continues to be the dominant source of local air pollution, as it was in London in the 1950s and 1960s. Data from two residential areas in Beijing with very high ambient air pollution concentrations of particulates and sulfur dioxides from domestic coal burning provide an opportunity to evaluate the association between daily mortality and air pollution in a setting very different from that considered in the United States studies.

Material's and methods

Study area. Dongchen (East District) and Xichen (West District) are two residential areas, about 30 [km.sup.2] in area, located in the center of Beijing (Fig. 1). The total population in 1989 was 1 419 123 (713 607 men, 705 516 women), with 645 192 in Dongchen and 773 931 in Xichen (approximately 25% of the urban population of Beijing). Population density was approximately 24 000 persons/[km.sup.2]. Demographic and social characteristics were compatible between the two areas. There are no major industries in these residential districts. Bicycles are the primary form of transportation, and automobile, bus, and commercial traffic is very light on most streets. Coal stoves, used for heating in 90% of households and for cooking in 50% of households, are the major air pollution source in the area. Few homes (or hospitals) are air conditioned, and windows are kept open most of the time from May to September.

Mortality data. In the event of a death in Beijing, the decedent's family is required to obtain a death certificate from the hospital or local community clinic, which must be submitted to the police station to cancel the decedent's household registration, and to local public health station to have the decedent's home "sterilized." The decedent's family then obtains two certificates (one from the police station and another from the local public health station), which are required to have the body cremated.

Daily mortality data for calendar year 1989 were obtained from death certificates recorded at the public health stations in the Dongchen and Xichen districts. Deaths were first coded in Chinese Classification of Causes of Disease (CCD) and then transformed into International Classification of Disease, Revision 9 (ICD9)[7] by the collaborators from the Center for Health Statistics Information, Ministry of Public Health. Deaths resulting from accidents (CCD = E1-E15; ICD9 = 800+) were excluded, as were all deaths that occurred outside of the city. Total mortality was subdivided by cause of death: cardiovascular disease (CCD = 42, 44-47, 49-51; ICD9 = 390-414, 417-448); pulmonary heart disease (CCD = 48; ICD9 = 415-416); chronic obstructive pulmonary disease (CCD = 54; ICD9 = 490-493); cancer (CCD = 22; ICD9 = 140-208); and other causes. Because the death certificates are sorted according to report date, rather than recorded death date, the 1989 death certificates included 189 people who died in December 1988 but were reported in 1989. Data from December 1989 were excluded from this analysis to prevent such partial ascertainment.

Air monitoring and weather data. The outdoor total suspended particle (TSP) and [SO.sub.2] concentrations are measured at the World Health Organization Global Environmental Monitoring System sites in each of the two districts (Fig. 1). TSP is measured gravimetrically and [SO.sub.2] is measured by colorimetric pararosaniline methods.[6] Daily air samples are collected and analyzed for 2 to 3 wk each month, starting with the second week of the month. The annual mean concentrations of [SO.sub.2] and TSP were 108 [mu]g/[m.sup.3] (SD = 141) and 350 [mu]g/[m.sup.3] (SD = 172), respectively, in Dongchen district, and 93 [mu]g/[m.sup.3] (SD = 122) and 390 [mu]g/[m.sup.3] (SD = 180) in Xichen district. The daily measurements of the pollutants were highly correlated (Pearson correlation = 0.93 for TSP and 0.92 for [SO.sub.2]) between the two monitors. Mean TSP and [SO.sub.2] across the two districts were used as the exposure estimates. [SO.sub.2] and TSP measurements were performed for 194 d from January to November. Temperature and humidity data were obtained from Beijing Weather Bureau.

Statistical method. Daily counts of deaths were regressed, using Poisson regression with a Markov approach because of the correlation in time series data.[8] The model can be expressed as

log E([y.sub.t] = [alpha] [x.sub.t] + [beta] [z.sub.t] + [sigma][[gamma].sub.j] [y.sub.t-j] where t = 1, 2, ..., 365 (day of the year), j = 1, 2, 5 (lag day for mortality), and y, is the number of deaths on day t;[x.sub.t] is the vector of controlling indicator variables on day t for quintiles of temperature, quintiles of humidity, and an indicator for Sunday (the only day of the week significantly different); and [z.sub.t] is the vector of air pollution variables on day L The regression coefficients were estimated, using a quasi-likelihood approach,[8] and the variances were estimated robustly.[9] Because the model included previous days' mortality, the estimated coefficients for air pollution represent the effects conditioned on the previous days' mortality.

Results

Table 1 shows the distribution of air pollution, weather, and daily mortality in 1989 in Dongchen and Xichen districts. TSP concentrations (mean = 375 [mu]g/[m.sup.3], maximum = 1 003 [mu]g/[m.sup.3]) and [SO.sub.2] concentrations (mean = 102 [mu]g/[m.sup.3], maximum = 630 [mu]g/[m.sup.3]) were both far above WHO recommended criteria. An average of 21.6 persons died cause day. Cardiovascular disease was recorded as the cause in 47% of the deaths, cancer in 22%, pulmonary heart disease in 9%, and chronic obstructive pulmonary diseases in 3%.

[TABULAR DATA 1 OMITTED]

Both TSP and [SO.sub.2] were associated negatively with temperature, indicating that coal combustion for heating was an important source. High monthly mean mortality rates were observed in winter months when TSP and [SO.sub.2] were high, and temperature and humidity were low (Fig. 2).

Initial Poisson models included indicator variables of quintiles of temperature, humidity, and [SO.sub.2] or TSP, plus an indicator for Sunday. The lowest pollution quintile was the reference group. The estimated increase in the logarithm of total mortality (relative risk) was then plotted against the mean [SO.sub.2] and TSP of each quintile ([mu]g/[m.sup.3]) in the original scale (not presented) and in the logarithmically transformed scale (Fig. 3). Adjusted daily mortality increased linearly with the logarithm of [SO.sub.2]. The association with [SO.sub.2] appeared much stronger than with TSP.

When In(TSP) and In([SO.sub.2]) were included separately in the model, In([SO.sub.2]) on the same day was a highly significant predictor of daily mortality ([beta] = .148, SE = .032). In a separate analysis, In(TSP) was not significant ([beta] = .060, SE = .048). In this dataset, In([SO.sub.2]) and In(TSP) were highly correlated (R = approximately 0.6). The assessment of independent effects of [SO.sub.2] and TSP was achieved by including both pollutants simultaneously in the model. Inclusion of In(TSP) as a covariate did not change the magnitude or significance of the In([SO.sub.2]) estimate ([beta] = .158, SE = .034), whereas the estimated effect of In(TSP) was reduced substantially and was far from significance ([beta] = -.001, SE = .048).

Save

Print

Send

Link

Subscribe