Respiratory diseases in Greater Buenos Aires and meteorological variables


  • Adelia P. Alessandro

    Corresponding author
    1. Dto de Cs. de la Atmósfera y los Océanos, Universidad de Bs As, Buenos Aires, Argentina
    • A. P. Alessandro, Dto de Cs. de la Atmósfera y los Océanos, Universidad de Bs As, Buenos Aires, Argentina.
    Search for more papers by this author


This paper provides a preliminary assessment of the relationship between atmospheric conditions and respiratory diseases based on the analysis of disease occurrence in patients during the period August 2004 to September 2007. To this effect, medical records from a medical services firm were examined, as well as temperature data (daily, maximum, minimum, dew point), pressure and relative humidity, measured at the meteorological station of Ezeiza (34.49°S, 58.32°W). The analysis revealed a strong seasonal relationship, with a peak in winter (low temperatures) and a minimum in summer. The annual cycle explains 76.24% of the total of the variance of the series. Most patients were below 10 years of age and older than 71. The best correlations were observed between daily values of the number of patients presenting with disease symptoms (N) and temperature, and between monthly values of N and maximum temperature (coefficients of − 0.73 and − 0.91, respectively). Multiple correlation between all meteorological variables obtained with the Stepwise method, made it possible to estimate the expected number of patients on daily and monthly scales. Mean geopotential height fields at 1000 hPa corresponding to days with the lowest number of patients have opposite synoptic characteristics to those of days with the largest number of patients. The same occurs with the fields corresponding to consecutive days with less than 40 patients per day and consecutive days with more than 300 patients per day. The fields corresponding to extremely high values show a positive pressure anomaly and consequent cold air advection over Buenos Aires and surrounding areas. Copyright © 2011 Royal Meteorological Society

1. Introduction

Every year, directly or indirectly, climatological and meteorological factors cause premature death to millions of people, as well as diseases or inabilities to hundreds of millions, all over the world. In spite of this, low priority is given to prevention of meteorotropic diseases by the policy and economic sectors. The relationship between weather and symptoms of different pathologies has been studied worldwide, but there is little awareness of risk attributable to environmental factors in the public health area. The study of environmental factors, including meteorological ones, is an essential contribution to better understand the health-disease process (Kashiwabara et al., 2002; Santic et al., 2002; Wardman et al., 2002; Alessandro et al., 2003).

Amongst respiratory diseases, asthma is a paradigmatic meteorotropic disease. Almost all affected people associate the exacerbation of their symptoms with changes in weather conditions. Asthma is a dynamic disease whose occurrence varies constantly. Variations may be seasonal, periodic and non-periodic.

Controlled experiments have been undertaken to describe the respiratory effects caused by changes in weather conditions or weather factors, both separately and in interaction. However, laboratory experiments were not able to reproduce what happens under the much more complex natural conditions, where the influence of these phenomena may be enhanced by air contaminants, infecting agents, or particles causing allergy or irritation (Rosas et al., 1998; Galán et al., 2000). There is now evidence of a worldwide increase in the occurrence of asthma and other allergic diseases, especially in industrial societies. Such an increase is related to anthropogenic and weather factors and especially to extreme weather events.

Due to the lack of sustained and good hospital databases, only a few studies of the effects of meteorological conditions on humans have taken place in Argentina. Among these, Rusticucci et al. (1996) found a close relationship between the persistence of high minimum temperatures and the number of hospitalized cardiovascular patients. Rusticucci et al. (2000) related 15 different diseases with four meteorological variables to study the relationships between hospital emergencies and weather conditions by analysing summer and winter cases of patients requiring attention at the emergency room of a hospital in the city of Buenos Aires, during 1996–1997. They found different rates of incidence of weather variables on seven groups of diseases. Alessandro et al. (2003) examined the irritability caused by given temperature ranges and the associated synoptic situation. This situation presents a general northerly flow that brings heat and moisture from the tropics into the city of Buenos Aires. Also in Buenos Aires, De Garín et al. (2006) found good multiple correlations with meteorological variables, especially a prominent role of the passage of fronts, and the incidence of heart diseases.

With regard to respiratory diseases, Hoffmann et al. (1983) studied the relationship between the presence of depressions and asthma occurrence in the city of Rosario (32.55°S; 60.47°W) between August 1978 and July 1979. They found an increase in daily asthma cases coincident with the presence of depressions located at 500 hPa at about 65–75°W and 40°S, and a decrease when the depressions passed over the city. Piccolo et al. (1988) determined the relationship between asthma emergency cases, meteorological variables and pollution in the city of Bahía Blanca (38.44°S, 62.1°W) during a short period. Tolcachier et al. (1994) studied the distribution and frequency of cases of bronchial asthma in paediatric emergencies in Capital Federal, and found that maxima took place in autumn. In the city of La Plata (35°S, 58°W), Nitiu et al. (2002), studied the correlation between total pollen concentration in the atmosphere and the degree of pollenosis in patients with respiratory allergies in the period September 1998 to June 1999. They found that the greatest pollen concentration is not always related with the maximum number of patients.

Various authors have used circulation indices for quantitative descriptions of atmospheric circulation. Among the first studies, Namías and Clapp (1951) introduced the zonal circulation index, defined as the difference between hemispherically averaged pressures, over subtropical and sub-polar latitude circles. They studied the behaviour of the mean and quasi-periodic variations and the relationship with the amplitude and motion of long waves. In the south of South America, Schwerdtfeger (1951) and Prohaska (1952) introduced local and regional circulation indices. Minetti et al. (1987) applied several circulation indices, calculating pressure differences between pairs of meteorological stations, to study long-period fluctuations (2–20 years) and relate them to precipitation. The use of indices, for instance the Southern Oscillation Index [SOI (Tahití-Darwin)], the Trans-Polar Index (Jones et al., 1999), the Antarctic Oscillation index (Jones et al., 2003) and the Monsoon Index (Kinter et al., 2002; Lee et al., 2005) has become generalized. In several works Alessandro et al. (1998a, 1998b, 2001, 2007a, 2007b, 2008b, 2008c, 2009) applied circulation indices such as Z, the zonal index, the southern index, a, and C, the curvature index, to characterize different synoptic situations associated with precipitation and temperature in Argentina. The last two indices will be applied in this work to characterize the prevailing atmospheric circulation related to the greater number of patients.

This paper aims at determining the distribution of patients with respiratory diseases in Greater Buenos Aires during a period August 2004 and September 2007, and analysing the possible connections between those diseases and meteorological variables. Some factors, such as the content of air pollutants, have not been considered because no data are available The lack of series of patients and of other factors strongly related to meteorotropic diseases, such as the content of air pollutants, in places and periods common to both, has not been considered in this work. Further studies are needed to achieve the final goal of obtaining good forecasts of imminent atmospheric phenomena associated all oriented to effective preventive measures that may reduce morbidity caused by respiratory diseases and many other meteorotropic diseases.

2. Data

The city of Buenos Aires (34°36′S, 58°26′W) is located on the margin of the Plata River and has amoist, temperate climate with hot summers and cool winters. Mean annual temperature is 17 °C. The coldest month is July, with a mean temperature of 10 °C, and the warmest month is January, when mean temperatures are 24–25 °C. Annual relative humidity is 70% and mean annual rainfall is 1146 mm. Rainy seasons are autumn and spring. Mean temperatures in adjacent areas are similar, though there are differences in daily maximum and minimum temperatures that result from differential effects of the river and the city. According to the National Institute of Statistics and Censuses of Argentina (INDEC), Greater Buenos Aires includes the city of Buenos Aires and 24 adjacent municipalities of the Province of Buenos Aires, and covers a total area of 3833 km2. The number of inhabitants in the city of Buenos Aires is about 3.2 million and above 8 million in the 24 administrative areas of the province, which represents slightly more than a quarter of the country's population.

Daily records of patients (N) for the period August 2004 to September 2007 were provided by the company ‘Ayuda Médica’, which has an at-home attention system covering the Federal Capital of Buenos Aires and surroundings. This home attention system allows patients to call the company, the physician visits the patient and makes a diagnosis of the disease based on the answers of the patient. No medical histories are consulted or prepared, so the physician does not know or record whether the disease is new or recurrent. Diagnoses were considered with no modifications as informed by the physicians. That is to say, the diagnoses depend on the physician on duty. However, pathologies such as asthma may be described by physicians as spasmodic bronchitis, broncho-spasm, allergic bronchitis, bronchial spasm and recurrent obstructive bronchitis, among others. These are diagnosis euphemisms that, according to an operational definition, must be considered asthma, even if some cases are in fact viral or bacterial bronchitis. According to these last considerations and due to the form in which the data of the series of patients were originated, it was impossible to achieve a good discrimination between the diverse diseases of respiratory origin: this is why the present study addresses respiratory diseases in general. Another problem found was the fact that data do not record the exact time when the affection started, thus the respiratory problem was considered to have started on the same day that the patient called the company. Furthermore, due to this lack of information, hourly meteorological data were not used, as well as other variables more difficult to address, such as wind. In spite of those problems, the number of data are significant for a first approach.

During the 3 years studied, the company Ayuda Médica assisted an average of 668.119 patients per year. Respiratory diseases affected 7.4% of the total number of patients assisted by the company.

Daily values of temperature (T), maximum temperature (Tmax), minimum temperature (Tmin), dew point temperature (Td), surface atmospheric pressure (P), and relative humidity (H%) from Ezeiza meteorological station (34.49°S, 58.32°W) were used in this study. Data were provided by the National Meteorological Service of Argentina (SMN). The 1000 hPa mean and geopotential height anomaly fields for the studied situations were obtained from graphical outputs of the National Centers for Environmental Prediction (NCEP). The reason for considering the 1000 hPa level instead of the near-surface level is that the former provide a better representation of reality in mountain areas, making it possible to avoid problems with topography in the data assimilation process of the NCEP model (Alessandro, 2008a).

3. Methodology

To know the annual distribution of series N, formed by patients for the chosen period, the number of patients recorded per month and gender (female (Nf) and male (Nm)) was obtained. Harmonic, or Fourier, analysis, studies the representation of functions or signals as a superimposition of ‘basic’ waves or harmonics. Fourier analysis was applied on the series N in order to know the temporal cycles that best explain the total variance of the series (Blackman et al., 1958). Patients were classified according to age and gender to establish disease incidence per group. Age intervals were 10 years, with the exception of the first and the last periods with patients below 1 year of age and older than 81 years. Histograms were made to study possible seasonal cycles in disease incidence according to gender and age.

The months corresponding to summer, autumn, winter and spring are December, January and February; March, April and May; June, July and August; September, October and November, respectively.

Since the year 2005 is included in the study period, data from the 2005 census projection performed in the area of Greater Buenos Aires were used to know the distribution of age and gender of inhabitants.

As the first age interval in the census is up to 10 years of age, occurrences in the first two intervals of N were added, for comparison purposes. The former were used to calculate relative frequencies of city inhabitants per gender and age categories. The same was done with the sample of patients with respiratory problems in 2005. For each age category, a quotient (r) was estimated between the relative frequencies of patients and inhabitants in 2005.

For example for ages in the 40–50 year interval:

  • pfn = number of female patients between 40 and 50 years of N series/total number of patients;

  • pmn = number of male patients between 40 and 50 years of N series/total number of patients;

  • pfs = number of female patients between 40 and 50 years according to the 2005 census/total number of persons in that year;

  • pms = number of male patients between 40 and 50 years according to the 2005 census/number total of persons in this year.

Thus, rf = pfn/pfs and rm = pmn/pms

Values of r greater than 1 imply significant disease prevalence.

In order to estimate the expected number of N in each month of the year, the number of patients in a given month was compared with the same month in other years (3 months in total for each month, according with the period studied). Student's (T) and Fisher (F) tests (Brooks et al., 1953) were applied to test mean values and variances of the various samples at the 95% confidence level.

Several linear correlations were calculated to describe the relationship between meteorological variables (T, Tr, Tmax, Tmin, P, and H%) and the number of patients (N). For each pair, linear correlations were estimated for the daily and monthly series and for different time lags to investigate whether meteorological conditions on the days preceding the call of the patient had influence on the appearance of the disease. The lag number corresponds to the number of preceding days. In addition, to establish the incidence on N of sudden changes in those variables in consecutive days, the differences were calculated between the value of the previous day (i − 1) and day i, when the patient record was taken. These differences were correlated with the number of patients on day i.

As greater variability of meteorological parameters indicates a more frequent passage of synoptic systems (fronts, troughs, instability lines), standard deviations were calculated for each of the monthly variables in order to determine which of the variations affect or are related with N. The deviations were correlated with N.

Multiple correlations were also calculated to determine the degree of influence of each variable on N using the Stepwise regression method. According to Draper and Smith (1981), this method is the most recommended for the selection of variables because of its computational economy compared with other methods and the possibility of knowing the association amongst the variables at the various stages. This method takes the independent variables and correlates them by pairs with the dependent variable, eliminating the independent variables that correlate worse with the latter. After comparisons of correlation values, the method selects the variables with the best coefficients.

Since the selected variables affect N to different degrees, the percentage of patients was calculated according to the daily value of each variable for the study period. This was made for future evaluations of the number of people affected in the different categories.

Finally, to know meteorological conditions on the days with highest and lowest number of patients, the mean geopotential height fields at 1000 hPa and their anomalies with respect to the 1960–2000 normal were analysed. Those situations may be represented partially by means of circulation indices, so an attempt was made to quantify the circulation conditions that favour disease prevalence.

The two meridional and curvature indices used in this paper were applied at 35°S. According to Alessandro (1998b), meridional index R at 1000 hPa presents a good association with temperature. Under mean conditions, R > 0 indicates higher atmospheric pressure over the Pacific Ocean or lower pressure over the Atlantic Ocean. At 500 hPa index R behaves similarly with respect to temperature. If R < 0 there is greater cyclonic activity over the Pacific Ocean or the Atlantic Ocean, where the western border of the cyclone may cause warm air advection. The curvature index describes cyclonic and anticyclonic circulation. Positive (negative) curvature indices represent troughs (ridges). The meridional and the curvature indices are defined as (Alessandro, 1998a):

equation image

where h is the geopotential height at 1000 hPa taken at 35°S, and 75 and 50°W.

4. Results

4.1. Distribution of identified patients (N)

4.1.1. Gender and annual distribution

Figure 1 shows the seasonal variation within the series, with greater prevalence in May, June, July and August 2007. Lower values were obtained for male patients, although distributions for both genders are almost the same. The Fourier analysis revealed that 76.24% of the variance is explained by harmonic 3, which represents the annual cycle. The other harmonics are very weak, decreasing to 5.58% of variance for the most important one in the rest of the series. However, the annual cycle was filtered out and the corresponding spectrum was recalculated. In this case, harmonics 4 (289 days) and 5 (231) contribute together with a variance of 31%. If harmonic 5 is filtered out, then harmonics 9 (128) and 10 (115) explain a total of 23 and 28% for harmonics 7 (13–65) and 8 (144) if harmonic 4 is filtered out. That is to say, there are no waves shorter than the annual that might explain a variance above 50%.

Figure 1.

Monthly distribution of N (patients), Nf (female patients) and Nm (male patients). equation image; N, equation image; Nf, equation image; Nm

4.1.2. Age and gender distribution

Figure 2 represents the age and gender distribution of N during the period August 2004 to July 2007. There is a dominance of patients below 10 years of age and males up to approximately 20 years old. For ages above 20 the number of female patients increases. This last result is consistent with those obtained by other investigators, at least referring to one of the respiratory diseases, i.e. asthma. Those studies indicate prevalence of this disease in male patients during childhood, the tendency reverting for adults (Tolcahier et al., 1994). Several causes, essentially hormonal, conspire for this, and trend lines cross each other in adolescence. The second maximum occurs between 21 and 30 years, due to the large number of people of working age in this category. The other two peaks in disease prevalence occur in children and elderly people, possibly because of susceptibility, which was already demonstrated in previous studies. Figure 3 shows gender and age distributions for the four seasons of the year, although monthly distributions had been estimated previously (not shown).

Figure 2.

Age distribution of N, Nf and Nm in 10 year age categories during August 2004 to September 2007. equation image; N; equation image; Nf; equation image; Nm

Figure 3.

Seasonal age and gender distribution of N during 2008/2004–2009/2007. (a) Summer, (b) Autumn, (c) Winter and (d) Spring. equation image; N, equation image; Nf, equation image; Nm

Again, it can be seen that patients below 21 years of age are mostly males and that this age category has the maximum value in the four seasons. Slight differences are observed among the four seasons for patients from 31 to 61 years of age. Relationships among the remaining categories are as the annual cycle. Figure 4 represents relationship, r, between the number of patients in 2005 and the census held in that year: r is greater than 1 in the categories from 0 to 10 years of age and above 71 years of age, which provides further confirmation for the two maxima found above. Although calculations were made using census values for year 2005 (no census is available for other years in the period 2004–2007), the difference with absolute values of patients represented in the previous histograms (Figure 2) is that men were relatively more affected than women in the last two periods.

Figure 4.

Relationship (r) between relative frequencies in the 2005 sample and in the 2005 census. equation image; N, equation image; Nf, equation image; Nm

4.1.3. Comparison between equal months of the series August 2004 to July 2007

Table I shows the monthly values of statistics T and F for three different comparison periods.

Table I. Values of statistics T (Student) and F (Fisher) calculated for groups of equal months of years 2004, 2005, 2006 and 2007
MonthT Between 2004 and 2005T Between 2005 and 2006T Between 2004 and 2006T Between 2004 and 2005T Between 2005 and 2006T Between 2004 and 2006
August− 0.01− 0.38− 0.381.050.881.08
September− 0.12− 0.12− 0.230.910.841.31
October− 0.220.09− 0.100.980.771.33
November− 0.03− 0.06− 0.100.931.120.95
 Between 2005 and 2006Between 2006 and 2007Between 2005 and 2007Between 2005 and 2006Between 2006 and 2007Between 2005 and 2007
May0.08− 0.50− 0.421.260.491.62
July− 0.500.07− 0.391.731.040.55

All T values are within the confidence level of 95 ( ± 1.96), and F values are below F1, ν2) (1.84), i.e. samples of each month may be considered similar at this level. Therefore, a similar number of patients may be expected for each month in the years to come.

4.2. Correlation with meteorological variables

4.2.1. Linear correlations (daily and monthly)

In spite of the logical high correlation between temperature and the other variables used, 0.96 for Tmax, 0.93 for Tmin and 0.86 for Td, and to lesser degree for P (−0.64) and for H% (−0.32), the author wanted to identify the variable that makes patients most vulnerable to disease.

Daily and monthly correlation coefficients are shown in Table II. Daily values calculated for lag = 0 and 1 (number of patients on day (i) with the variable of the previous day (i − 1)) are practically identical. That means that the value of the variable on the previous day affects N. Correlation coefficients decrease with increasing lag, although correlations with temperature are significant up to lag = 10, which would point to the influence of cold waves. Monthly values are greater than the daily ones, the best monthly correlation for lag = 0 responds to the smallest fluctuation in the monthly series of patients and the series of the variables considered. Although correlations for the monthly series are smaller than those for the daily series, for lag = 1 (number of patients of the month (j) with the variable of the previous month (j − 1)), for the variables related with temperature.

Table II. Coefficients of daily and monthly correlations between N (patients) and meteorological variables for lag = 0 (number of patients on day/month (i) with the variable of the same day/month (i)) and 1 (number of patients on day/month (i) with the variable of the previous day/month (i − 1))
 Lag = 0− 0.73− 0.72− 0.64− 0.620.250.37
 Lag = 1− 0.73− 0.72− 0.66− 0.630.240.38
 Lag = 0− 0.91− 0.92− 0.88− 0.840.780.43
 Lag = 1− 0.64− 0.66− 0.58− 0.490.570.57

This indicates greater independence between the mean number of patients and the mean temperatures of the previous month. On the other hand, mean humidity and pressure show greater dependence than in the daily values. All the correlations are significantly different from zero, at the 95% confidence level (theoretical correlation coefficient: 0.32 and 0.19 for monthly and daily data).

On the contrary, correlations were not significantly different from zero (0.01), when the number of patients was related with changes in any of the variables between the day of record and the previous day. Figure 5(a) and (b) show the daily cycle of the four temperatures. Figure 5(b) shows the daily cycle of N and P and Figure 5(c)) of H%. Maximum N values are consistent with minimum temperatures and with maximum pressure and H%, according to the signs of the correlations. The most pronounced maximum and minimum values are in agreement with high correlation values.

Figure 5.

Daily distribution of (a) equation image; T, equation image; Td, (b) equation image; Tmax and equation image; Tmin, (c) equation image; N, equation image; P and (d) equation image; H%

4.2.2. Correlation between monthly N and standard deviations of the variable

Table III shows that the number of patients is mostly related to temperature variability. The correlations are significantly different from zero at a significance level of α = 0.05 (where the theoretical coefficient is 0.32) but there is no correlation with pressure and relative humidity. This means that N is sensitive to the passage of a larger number of synoptic systems producing changes in temperature.

Table III. Correlation coefficients between monthly N and standard deviations of the meteorological variables
  1. Correlations significantly different from 0 at significance level α = 0.05 (ρt = 0.19) are in bold.

0.520.320.510.50− 0.130.10

4.2.3. Multiple correlation

In Section 4.2.1 simple correlations were calculated between the number of patients N and each of the selected variables. Although correlations between N and the four temperatures have similar values, the author wanted to identify the variable with the greatest influence on N, Nf and Nm, as well as the contributions of P and H%. Table IV shows the equations obtained with the Stepwise method and the correlation coefficients. The process identified T, Tmax, Td and P for the entire daily series, whilst the variables for the series consisting only of men or women, are not exactly the same, as shown in the same table. Correlation coefficients for the monthly series are greater. In this case the method selected Tmax.

Table IV. Equations of multiple correlations (with daily and monthly data) between all meteorological variables and N obtained with the Stepwise method
 Equations of multiple correlationsρξSD
  1. ρ = correlation coefficient; ξ = mean quadratic error; SD, standard deviation.

 N2081.226 − 4.818T − 1.762P − 3.302Tmax − 1.243Td0.7345.3767.52
 Nf1256.563 − 1.137P − 1.955Tmax − 3.583Td + 0.664H0.7125.5928.51
 Nm761.3183 − 2.69T − 0.626P − 1.111Tmax0.7021.1229.97
 N370.13 − 10.802Tmax0.9122.6441.12

Because of the high values of the correlations among the mentioned temperatures, the variation or relationship with N of one of them will be detected in any of the other three.

Figure 6 presents the daily (a) and monthly (b) distribution of real (grey) and calculated (black), via the stepwise method (sw), data for series total N. Maximum and minimum values in both curves are in phase, though peaks are more pronounced in the approximation. Both genders (not shown) have similar characteristics to the entire samples. The correlations and errors for both genders are practically the same. The mean quadratic errors (ξ) were below the standard deviations (DS) of the series, so the adjustment may be considered good (Rao, 1997).

Figure 6.

Real (grey) and calculated (black) data with the stepwise method (sw), for (a) daily series (N) and (b) monthly series

4.2.4. Daily number of patients according to T, Tmax, Tmin, Td, P and H%

Figure 7(a–f) presents the daily distribution of patients related with the daily value intervals of each variable, while Figure 7(g–l) represents the frequency of patients per day considering within each class interval of the entire period the number of days with the values of the variables (Tmax, Tmin, Td, P or H%). For instance, in Figure 7(a) the number of patients when temperature was between 1.5 and 3.8 °C was 3099 and between 3.9 and 6.2 °C was 8033, but given that the number of days with temperatures in the range 1.5–3.8 °C is smaller than in the following temperature interval, the number of patients per day is greater for the smallest temperature as can be seen in Figure 7(g).

Figure 7.

(a–f) Daily distribution of patients, per categories and (g–l) relative patients, during 2008/2004–2007/2007 of variables T, Tmax, Tmin, Td, H% and P. (a) T, (b) Tmax/nd, (c) Tmin/nd, (d) Td/nd, (e) H%, (f) Pr, (g) T/nd, (h) Tmax, (i) Tmin, (j) Td, (k) H%/nd, (l) Pr/nd

Again, these histograms make it possible to observe the increase in the number of patients per day with decreasing temperature. Patient distribution with relative humidity is smoother. The number of patients for any humidity range is at least 100 per day and 150 for humidity values above 80%. Although there are at least 100 patients per day in all pressure ranges, patients are twice as many when pressure is above 1010.5 hPa.

Considering the mean annual values of the six variables at Ezeiza in the entire period (16.38, 22.2, 11.8, 12.2 °C, 61% and 1011.5 hPa), 63–67% of the patients were admitted when the temperature was below the mean annual values of the four temperatures and 54–61% when relative humidity and pressure were above the mean annual values.

4.3. Mean synoptic situations and circulation indices

Although the annual temperature cycle was shown to have the greatest incidence on the number of patients, here the author wants to show the mean synoptic situations for the days when patient distribution reached maximum values. Also interesting are the mean synoptic situations with the highest and lowest number of patients. Mean geopotential height fields are only shown for the 1000 hPa level, as it provides the best explanation for temperature values.

Figure 8(a,c) shows the mean field of daily synoptic situations for 20 days with less than 40 patients per day. Those cases, as explained in Section 4.1.1, correspond to summer months. Figure 8(b,d) represents the mean situation of all days with 17 days with more than 300 patients per day and corresponds to the months of June and July. Figure 8(a,c) and Figure 8(b,d) represent different seasons of the year, and therefore have different characteristics. For instance the displacement of subtropical anticyclones to the north and the absence of the northwest low pressure system (DNOA) in the cold season, causing the absence of flows with northerly component associated with high temperatures. For comparison purposes, and to see the departures of mean situations from normal values, anomalies are shown in parts (c) and (d). In the area of Greater Buenos Aires, a zero-value anomaly in the field corresponding to fewer cases (<40), implies that it barely departs from normal conditions of flow with northerly component. In the opposite case, there is an important positive anomaly over almost the entire country, with 40 mgp in Ezeiza. At this station, the difference in mean surface pressure between the values of corresponding to days with less cases and the normal summer value is of − 0.5 hPa. On the other hand, the difference in pressure for more than 300 cases is 2 hPa, i.e. pressure is below normal in the lower extreme and vice versa.

Figure 8.

Mean 1000 hPa geopotential height field (a,b) and their anomalies (c,d) of days on which (a,c) the number of patients is smaller than 40 and (b,d) the number of patients is greater than 300

Given the greater interest raised by the greater number of patients in cooler months, Figure 9 shows the mean geopotential height field at 1000 hPa for all the months of June and July in the series with less than 170 patients per day in the series (15 days). The figure shows a trough over the continent, contrary to what happens in the mean field for the situations with more than 300 patients per day (Figure 8(b,d)) registered in winter.

Figure 9.

Mean 1000 hPa geopotential height field (a) and their anomaly (b) of the winter days when the number of patients was smaller than 170

To check whether persistence of high or low numbers of patients is determined by particular synoptic situations, mean 1000 and 500 hPa fields were counted and analysed for those days when large patient numbers were recorded on more than two consecutive days and those when the number was low.

Again, high or low persistent patient numbers occurred in those months with the highest and lowest number of patients respectively. With the exception of June 2006, the remaining periods showed positive anomalies in different locations over the continent. The mean fields obtained were practically equal to those of cases with more than 300 and with less than 170 patients per day respectively

To save space, Figure 10 presents two mean 1000 hPa geopotential height situations and the corresponding anomalies only. Parts (a) and (c) correspond to 4–7 January 2005 with 46–72 patients per day and parts (b) and (d) correspond to 4–7 July 2005 with more than 300 patients per day. In Figure 10(d)) a positive anomaly is observed in the study area and the mean field shows a great anticyclonic system with the corresponding cold and dry air advection, explaining the larger number of cases. On the contrary, in Figure 10(c)) a trough and a negative or neutral anomaly are observed over a large part of the country, associated with higher temperatures.

Figure 10.

Mean fields (a,b) and their anomalies (c,d) corresponding to two situations of consecutive days (a,c) with low number of patients (4–7 January 2005) and (b,d) with more than 300 patients per day (4–7 July 2005)

It is worth mentioning that the data assimilation method of the NCEP model causes a spurious maximum in the Argentine northwest which displaces the DNOA to the east, distorting the expected consequences on the meteorological variables, as demonstrated by Alessandro (2008a).

Characteristic situations were found for extreme and persistent values, but no clear relationships were found with pressure for other ranges of N, which is consistent with the low correlation coefficients.

As low temperature correlates well with N, flows with southerly component are expected to represent larger numbers of patients. This is associated with a decrease in the occurrence of winds with northerly component (Lichtenstein, 1976), with consequent cold air advection over Argentina.

Table V shows daily and monthly correlations between the number of patients and the indices R and C: R values are positive, which is associated with cold air advection, C is negative, which indicates a ridge dominating the country. This result is consistent with the mean situations found for greater values of N. All the correlations were significantly different from zero at the 95% level (greater than 0.19). Index C indicates a strong relationship with the daily number of patients and a better relationship with monthly values.

Table V. Daily and monthly correlations between N and the meridional and curvature indices for lag = 0 (number of patients on day/month (i) with the variable of the same day/month (i)) and 1 (number of patients on day/month (i) with the variable of the previous day/month (i − 1))
 Lag = 0Lag = 1Lag = 0Lag = 1
Daily0.210.20− 0.34− 0.37
Monthly0.630.33− 0.71− 0.76

5. Conclusions

The number of patients (N) with respiratory illness in Buenos Aires presents a strong seasonal relationship, with maximum values in the coldest months of the year and minimum values in summer. The annual cycle explains 76.24% of the variance. The greatest frequency of patients corresponds firstly to children below 10 years of age and secondly, to people above 71 years of age. Up to approximately 20 years of age, patients are mostly males; females prevail from that age on, based on the information of the series only. There are practically no differences in the age and gender distributions in the different seasons. Student's and Fisher's tests applied on the 12 samples, each of them formed by equal months of years 2004, 2005, 2006 and 2007 determined that the number of patients in each month is similar at the 95% confidence level. The number of expected patients in a given month may therefore be estimated using this sample.

Daily and monthly linear correlations between N and temperature (T), maximum temperature (Tmax), minimum temperature (Tmin), dew temperature (Td), pressure (P) and relative humidity (H%) were significantly different from zero at the 95% confidence level. T, Tmax and Tmin were the most significant correlations at that level. Correlations significantly different from zero at the 95% level were found between N and the passage of synoptic systems affecting temperatures in Ezeiza. Non-significant values were obtained when comparing the differences in variable values between the day of patient record and the previous day. Daily correlations between temperatures and N, are significantly different from zero, even for correlations of N with the temperature on the 10th day preceding the patient's call (lag = 10). The Stepwise method determined a multiple correlation of 0.74 among N and T, Tmax, Td and P. The calculated Nsw made it possible to determine the expected daily and monthly values of N with an error of ε = 45.37 and of 22.64 respectively and the results obtained for the period August 2004 to September 2007 make it possible to achieve a first approximation to the daily frequency of patients in the different ranges of the analysed parameters.

Mean geopotential fields for days with maximum and minimum numbers of patients are different. Consecutive or non-consecutive days with more than 300 patients per day are characterized by a high pressure system in the region under study and consequently a positive geopotential height anomaly, with the corresponding cold air advection over Buenos Aires or nearby, while on days with less than 40 patients per day a low pressure system is found in the northeast and negative or neutral anomalies. Correlations between circulation indices R and C and the number of patients are different from zero. Their value and sign indicate greater respiratory disease occurrence when the Atlantic Ocean Anticyclone weakens or when there is a ridge or an anticyclone near the City of Buenos Aires.

The results obtained are a first approach to the proposed problem. Databases indicating the time of beginning of the symptoms are necessary and if possible, a follow-up of a group of voluntary patients sensitive to this type of diseases.


University of Buenos Aires and CONICET supported this work through grants EX234 and PIP 02717. The author also thanks Ayuda Médica for the information provided.