Check the next ↓ link to find more info:
AIR POLLUTION IN BOGOTÁ, COLOMBIA:
A Concentration-Response Approach
Air pollution has become one of the most important concerns of the local authorities of Latin-American cities and Bogotá, Colombia is no exception. This paper will develop a model to define a concentration response function between three of the most important air pollutants in Bogotá and the daily respiratory hospital admission counts in the city during the year of 1998. This article won't concentrate on the estimation of the costs but rather will motivate further work on this area by giving the first input needed for that type of analysis.
I. Introduction Air pollution has become one of the most important concerns of the local authorities of Latin American cities. Bogotá, like as other urban centers in South America such as Sao Paulo, Mexico City and Santiago de Chile, shows significant levels of air pollution, levels that may represent a high risk for the population's health and certainly a reduction in the quality of life of its inhabitants. Bogotá, capital of Colombia, is one of the largest cities of Latin America; with a population of around 6.5 million and an annual growth rate of 2.081 percent it is the largest urban center in Colombia; it also has the highest rates of environmental deterioration of the country. Air pollution has increased dramatically lately, due mainly to the uncontrolled increase in the number of vehicles in the city.2 Although air pollution has been monitored in Bogotá since 1967, it wasn't until 1990 that the monitoring stations were spread widely throughout the city. At that time the Secretary of Health of the District with the collaboration of the Japanese International Cooperation Agency (JICA) pursued a study in order to determine the air quality of the city. This study concluded that the most important source of pollution in Bogotá was automobiles; 70% of the pollution could be attributed to cars. Another very important source of pollution was found to be bricks and battery plants, among others.3 The study conducted with the support of JICA identified for the first time the composition of air pollution in Bogotá and its principal components. These were identified to be the following: Sulfur Dioxide (SO2), Nitrogen Oxides (NOx), Total Suspended Particles (TSP), Carbon Monoxide (CO), Hydrocarbons (HC), and Ozone (O3). It was estimated that 75% of the pollutants‟ annual emissions correspond to Particulate Matter.4 The study determined that the levels of CO, HC, SO2 and Particulate Matter were not above the limits defined as safe by the WHO. This led to JICA „s conclusion that: in 1990-1991 air pollution in Bogotá did not reach levels of concern to the local authorities. Nevertheless, the rapid growth in the number of cars in Bogotá during the last decade originated additional interest in this matter. The JICA pointed out in 1996 that the number of cars registered in Bogotá had increased from 324.902 in 1991 to 570.000 in 1996; this meant that around 40% of the cars of the whole country were circulating in Bogotá. Currently, half of the localities of the city where the monitoring stations are exceed the emission limits stated as safe by the WHO, with Particulate Matter (PM10) and ozone levels being the major problems. Most of the largest cities in Latin America also share this problem. In Mexico, Santiago de Chile, and Sao Paulo vehicles account for almost all of the carbon monoxide emissions, between 50 to 90 percent of hydrocarbons, at least three-quarters of NOx and around 40 percent of suspended particulate matter (PM10)5. The great concern around pollution levels stems from the connection that has been found between exposure to these kinds of gases and human health problems; inhalation of these gases in certain concentration levels may cause serious respiratory illnesses as well as injuries to the neural system, especially in children. Most air pollutants have effects on human health although their effects are different. Consider first Carbon Monoxide. This pollutant reduces the level of oxygen in the blood forcing the heart to work harder. At high exposure levels it may affect the capacity of thinking, reduce the reflexes and cause nausea, dizziness, unconsciousness and even death. On the other hand, a pollutant such as nitrogen dioxide will affect mainly persons susceptible to respiratory infections, especially children. Nevertheless, a strong and direct effect on human health from exposure to this pollutant has not been proven to exist yet. On the contrary, there is strong evidence of the effect of sulfur dioxide on human health with long as well as shorter time exposure to it Recent studies have associated changes in the 24-hour average exposure to SO2 to lung function, increase in the incidence of respiratory symptoms and diseases, and even risk of death.
Particulate matter is another main pollutant that presents serious health effects on humans. Epidemiological studies have shown that the presence of particulate matter in the environment may affect the human respiratory apparatus causing a notorious reduction in lung function. Lead is also present in the air in most urban centers and its presence has been proven to be a serious problem especially for children. Lead may cause loss of memory, reading and spelling difficulties, vision problems, and deficiencies in perception among others. Finally, there is ozone, the principal component of smog. This gas has been associated with an increase in respiratory illnesses, eye problems and a reduction of lung activity.
The strong connection between air pollutants and health problems described in the previous paragraphs has, under these circumstances, become a concern for Bogotá‟s authorities. Statistics of the Secretary of Health showed that for 1996 around 14% of the visits to the hospitals were related to respiratory problems. The evidence is even stronger for the infant population where 30% of the visits to the hospitals were associated with Acute Respiratory Illnesses (ARI).
Local authorities now face the challenge of supporting the growth and development of the city and at the same time minimizing the adverse effects of the associated air pollution and its consequences on health. In order to find the best way to do so, cost-benefits analysis can take a very important role. Economists would suggest that policy makers, when making decisions on air pollution regulation, should weigh the costs and benefits associated with the different options they have; therefore, it is essential to estimate the effect of air pollution on human health to estimate the benefits related to human health of a reduction in air pollution. This paper does not concentrate on the benefit-cost analysis but gives a first step towards this final objective by estimating a concentration-response function for several pollutants using information available for Bogotá, Colombia. The remainder of this paper is organized as follows: Section II gives a general description of the data used and the sources from where they are extracted. Section III presents the model that will be estimated and section IV gives a short description of the status of air pollution in Bogotá. The results of the econometric models estimated are presented IN section V, and finally the conclusion of this article is stated in section VI. II. The Data The data used in this study come from two main sources and can be classified into two main categories: environmental data and morbidity information. The environmental data was provided by the Administrative Department for the Environment (DAMA). They include information from thirteen environmental stations that are part of the net of environmental quality of Bogotá. For all of them we have geographical information such as station address, latitude, altitude, precipitation, and temperature readings. The information on pollutants is not uniform across the different stations; measures for PM10, SO2, and NO2 are collected in nine stations while measures for CO and O3 are gathered in only six of them. The information on these measures comes in an hourly basis, for daily records for the year of 1998. The morbidity information available for this study consists of counts of daily Hospital Admissions. The information was gathered by the Secretariat of Health for the District and comes from the reports that each Hospital in the city fills on a daily basis. The Respiratory Hospital admissions were taken from the original dataset and aggregated in order to obtain the total number of daily respiratory hospital admissions for the city in 1998. The original dataset contained information for each individual that was received at each hospital: date of admission (day, month, year), code of the hospital at which the individual was admitted; sex and age; neighborhood where the person lives; type of “visit” to the hospital (external, domestic or emergency); whether or not the person has been previously admitted to the hospital and if so, if this is the first time this year; is the person new in the year; referred patient; and type of insurance that the patient uses. Given the nature of this study however, only the daily number of respiratory hospital admissions is useful.
As mentioned above, the daily Respiratory Hospital Admissions for all hospitals in the city were extracted from these data and aggregated to daily counts. These data were combined with the environmental information in order to create a dataset with daily information on RHA as well as on pollution levels and meteorological data in order to estimate the concentration-response function for selected air pollutants in Bogotá, Colombia.
III. The Model Different types of models have been used to establish the relation between human health and air pollution. A broad classification of these models could be based on the unit of observation that they use.6 The first group uses the individual as its observation unit. Among these studies there are cross-sectional ones, which look for a relation between health outcomes and different levels of exposure to pollutants at a specific moment in time. Usually the levels of exposure are differentiated by the geographical distribution of individuals among the area in study. Cohort studies would be included in this group. These are very similar to cross-sectional studies but include also variation of exposure in time; cohort studies allow to include more exposed and less exposed individuals as cross-sectional studies, but also account for changes in exposure over time. They result very useful in analyzing which accumulating effects of exposure through time are to be studied. Nevertheless, they require the collection of individual level data through time, which makes them very expensive and lengthy to complete.
On the other hand, there are studies whose unit of observation is a group of people rather than the individual. These are known as ecological studies; they study the relation between pollutants and health, as the exposure to air pollution occurs in the community. These models were first developed for the analysis of mortality incidence of air pollution, and then expanded into the area of its morbidity effects. Epidemiological analysis is very common among morbidity studies because the information that it uses is in most cases easily accessible. Measures of morbidity traditionally used in these studies are the number of hospital admissions or visits to the emergency room. The fact that epidemiological models are based on previously collected morbidity data and pollution measures makes these models the most inexpensive to complete. British investigators are responsible for the development of ecological models7. Their studies showed that pollution, measured as particles and sulfur oxides, was associated with excess mortality as well as with morbidity indicators such as respiratory symptoms and infections, reduced lung function and exacerbation of chronic respiratory diseases. In the USA, ecological studies grew in number in the seventies, with the establishment of the US EPA. Studies such as Ferris et al. (1979) concentrated on large datasets that included several cities. As time passed ambient pollution levels have declined and these large scale studies have been changed for studies that look for relatively smaller effects of air pollution. Another change in the studies developed in this area has been the inclusion of indoor pollution in the analysis. In the beginning, only outdoor pollution measures were used, but some studies published in the eighties and nineties have showed that outdoor pollution also affects indoor measures, and moreover, that indoor pollution also has additional sources (such as cooking) that are of great interest in morbidity studies. It has been shown that indoor sources are an important source of individuals‟ exposure to particles, nitrogen dioxide and ozone.8 For morbidity, the fit of the models measured as the R2, increases dramatically when indoor pollution measures are included in the analysis.9 Another concern in epidemiological studies is the measurement of exposure levels. Indirect as well as direct instruments have been used in this effort. Direct instruments are based on individual monitoring systems for each person involved in the study that collect information both on pollutant levels and on exposure times. These are not only expensive but are sometimes also difficult to carry out. Indirect techniques to account for exposure usually collect information on concentrations of pollutants over time in different locations, and if possible, they estimate exposure time of the population; with this information, individuals at similar locations are assigned the concentration measure that corresponds to that area, say the place where they live.10 The use of either exposure or ambient concentrations leads to the distinction between dose-response and concentration-response functions. Since this study will use pollution measures that come from monitoring stations and assign those levels to individuals, it is clear that the model falls within the latter. Ecological models have also used several measures of morbidity. Among these there are work loss days; school loss days; days of restricted activity, rates of utilization of outpatient medical services and facilities, visits to the emergency room and hospitalizations11. There are two groups of ecological models: cross-sectional and time-series studies. The first group usually compares pollution and morbidity measures from different locations at one point in time; the second group is usually limited to a single location that is followed through a period of time, i.e. a year. Time series designs have the advantage of avoiding problems that are driven from the generalization of results and findings from groups to individuals, especially if they use a short period of collection of the data, say a day. The principal advantage of following a single population over time is that it is not necessary to control for individual-level confounding factors such as education, income or percentage of smokers, as long as they stay roughly constant over time.12 Ecological models also have limitations, and it is in the best interest of this article to identify them. Long-term cycles of pollutant and morbidity measures may cause wrong associations and give biased estimates for pollutants‟ health risk. These wrong associations may come from shared seasonal trends, driven for example from the transition from winter to summer. Addressing seasonal cycles in respiratory disease time- series is therefore important. Different modeling options have been used to model the seasonal behavior of morbidity and pollutants relation. Among these there are Fourier techniques, that fit sine/cosine waves to the data; auto regression methods; and the use of dummy variables that account for changes in time (day of the week, month or a specified season). Some recent studies show that no matter which method is used, the coefficient of the pollution variable does not change much, as long as seasonality is taken into account.13 The model of this article is an application of the ecological approach, since it examines the relation between air pollution in Bogotá-Colombia, and a health outcome –daily respiratory admissions to hospitals (RHA). The concentration level is measured as the average of daily maximums across the whole city. Geographical or individual distinctions are not taken into account due to data limitations. A concentration-response model relating respiratory admissions in hospitals in Bogotá and air pollutant levels will be constructed. The daily number of RHA in Bogotá is assumed to be a function of 4 pollutants and some meteorological variables such as rain and temperature; seasonal factors related to weather, pollen and diseases such as the flu and colds, are taken into account by including a dummy variable for each quarter of the year.
A semi-log specification is used to define the relationship between the health outcome (RHA) and pollution. All pollutants are expected to have a positive relationship with the number of respiratory hospital admissions in the city, and therefore the expected sign of each coefficient is positive. The expected signs of the meteorological variables are not clear a priori. One would expect a negative sign of the coefficient of rain, since rain acts as a cleaning device for the environment. Higher levels of rain will then result in lower respiratory hospital admissions, as rain reduces pollution in the air. By contrast, the expected effect of the daily average temperature in Bogotá is unclear. On one hand, most pollutants are the result of chemical processes that take place with solar radiation, which suggests a positive association with the dependent variable. On the other hand, cold weather is usually associated with illnesses such as cold and flu and hence respiratory illnesses. In developed cities, special warnings are issued on warm summer days in order to discourage people from exercising outdoors and getting exposed to pollutants such as ozone. This self defensive attitude may lead to a decreasing effect of temperature on the dependent variable. With the aim of further investigating this issue, a quadratic term for temperature is included in the model. Dummy variables have been a common way to avoid the problems associated with the presence of seasonality in morbidity to identify seasonal behavior of morbidity. One modeling option useful to separate seasonality is the inclusion of dummy variables that account for the different relevant periods (seasons). Bogotá is located in the tropics and therefore it is very difficult to clearly divide the year in seasons, as it has been done in several studies for the U.S.A and Canada. Four dummy variables are created in this article; one accounting for each quarter of the year, as an attempt to identify some pattern of seasonality in Bogotá.
The pollutants covered in this study are PM10, NO2, and O3. SO2 is not included in this study because for the year analyzed most of the monitoring stations did not have measures for this pollutant for the second part of the year. Although the possibility of including CO was considered, the relationship between this pollutant and health outcomes is left to future research; CO is related with heart diseases rather than with respiratory illnesses, which are the main concern of this article. In order to determine the relevance of the pollutants selected for this study, the first step will be to estimate what will be referred to as single pollutant models. In this first step, for exploratory purposes models will be run that relate the dependent variable to only one of the air pollutants here examined. In this case the weather variables and seasonal dummy variables will still be included in the model. After a series of exercises of this type, the full model will be estimated.
...
No hay comentarios:
Publicar un comentario