English (pdf)
Article in xml format
Article references
Automatic translation
Send this article by e-mail
Cited by SciELO 
Cited by Google 
Similars in
SciELO 
Similars in Google 
Permalink
Introduction
Food production is mainly based on monocultures that limit the entire agroecosystem, produce an intense im balance of homeostasis, and promote the appearance of different pests [1]. A control method for this production model is encouraging the intensive use of pesticides, which increases the exposure concentration of these pro ducts among farmers [2].
This exposure is usually linked to adverse health outcomes. Acute pesticide poisoning is the most fre quent event associated with pesticide exposure because signs and symptoms occur quickly after exposure and present a high possibility of complications [3]. On the other hand, chronic exposure to pesticides is linked to a greater risk of presenting chronic neurological disor ders, pathologies, such as cancer, and chronic kidney disease [4,5,6].
Even though there are no exact estimates, the World Health Organization (WHO)) reported one million cases of unintentional severe poisoning per year worldwide, leading to approximately 20 000 deaths [7]. Seventy percent of these poisonings occurred due to occupatio nal exposure. Additionally, one million hospitalizations were reported for suicide attempts using these substan ces. Of these suicide attempts, 300 000 resulted in death. However, a recent systematic review of the scientific literature published between 2006 and 2018 estimated that about 385 million cases of unintentional poiso ning for pesticides occur annually worldwide including around 11 000 fatalities [8,9]. Additionally, the annual incidence of acute pesticide poisoning in agricultural workers has risen to 18,2 cases per 100 000 workers [7].
Some authors have indicated that suicidal intention and living in rural areas are the most frequent causes of death associated with pesticide poisoning. In Jiang su, China, 38 513 poisoning cases were reported, of which 7891 were occupational pesticide poisonings in rural areas and 16 569 were suicides caused by poiso ning [10]. In Nepal, 439 cases of suicide were reported, with almost 90% of the cases carried out by self-harm and 71,3% of the cases occurred in rural areas [11]. Furthermore, in Gujarat, India, approximately 70% of poisoning cases occurred in people living in rural areas, and more deaths happened from suicidal poisoning than accidental poisoning [12].
In Colombia, between 2008 and 2015, 209 823 cases of chemical poisoning were detected by the surveillance health system, first by pesticides (32,2%), followed by medications (30,1%) [13,14]. However, few studies on rurality and pesticide poisoning in Colombia have been performed. Additionally, no analysis on the behavior of mortality and morbidity has been carried out over time, and the estimation of risk factors associated with fatalitydue to poisoning has not been established for pesticides.
Furthermore, most national and international studies have not compared the distribution of pesticide poiso nings between rural and urban areas to understand if the re is an additional risk of poisoning due to differences in the rural context.
Furthermore, agricultural policies and markets have led to an increase in the number of hectares dedicated to monocultures, which requires greater pesticide usage to ensure productive harvests. The increase in pesticide usage implies higher exposures to these chemicals, in creasing the risk for accidental and nonaccidental poi sonings.
In this context, this work aimed to analyze the be havior of pesticide poisoning in rural and urban popu lations through time series and to determine the risk factors associated with fatal cases in Colombia during 2007-2017.
Methodology
Type of study
This is a retrospective observational study that included two components: a descriptive cross-sectional study that considered the individual the unit of analysis and an eco logical design that analyzed aggregate as measures of morbidity and mortality of the population in rural and urban areas during 2007-2017.
Population
The population was constituted by the reported cases of exogenous intoxications compiled by the National Public Health Surveillance System (SIVIGILA) in the Chemical Substances Poisoning Surveillance System.
Data source
The information was compiled by the National Institute of Health between 2007 and 2017. These records came from all public and private hospitals and clinics in the country. The SIVIGILA collects the mandatory notifi cations of intoxication by chemical substances weekly. Each report constitutes an epidemiological week (a year is constituted by 52 epidemiological weeks).
Definition of the case and data curation
All the reported cases of exogenous intoxication invol ving a pesticide in Colombia between 2007 and 2017 were included. At the time of the development of this research, 2017 was the last year available for analysis. The database was reviewed, excluding duplicate cases. No imputation method was used to manage the missing data. The name of the pesticide was not included due to a loss of more than 20% of the data.
Variables
Sociodemographic, exposure, clinical outcome and pla ce of occurrence of the poisoning (rural or urban) were analyzed. The Public Health Surveillance Protocol Poi soning by chemical substance from INS (2011, 2014 and 2017) was used as source of definition of external in toxication [15,16,17]. The definition remained unchan ged during the described period of time. In addition, the definition of rural areas from the National Department of Statistics (DANE) was used: “rural areas are outside of the location of the county seat, which is characterized by the dispersed locations of houses and farms” [18].
In addition, the variable of cases with suicidal in tentions was defined as all intoxication associated with self-inflicted injury behavior. Additionally, the variable of ethnicity was defined as intoxication in a person who belonged to an indigenous, gypsy, Raizal, Palenquero or Afro-Colombian group [18].
Statistical analysis
In the cross-sectional study, continuous variables were presented as proportions or means, with standard devia tion. Then, a bivariate analysis was performed, compa ring the variables with the place of occurrence (urban or rural), using a chi-square test. Statistical significan ce was defined as a p-value < 0,05. Statistical analysis was performed with SPSS software, version 25, licensed from EAN University.
The risk-related measures for lethality associated with pesticide poisoning were estimated using Poisson regression with a logarithmic function.
The dependent variable was fatal events due to pes ticide exposure. The following independent variables were considered: sex, age, suicidal intention, ethnici ty, place of occurrence, educational level, and clinical treatment. The Poisson regression model allowed us to estimate the risk of lethality, calculated from the beta exponent measurement. The RR was estimated for ad justed and unadjusted models [19].
Finally, each risk measure was evaluated stratifying by rural and urban areas. In addition, the measure of as sociation of the interaction between the independent va riables and the rural population were calculated, consi dering a significant pint> 0,05. All models were adjusted for sex and age. Morbidity and mortality measures were adjusted for age by the indirect method with the WHO standard population [20].
For the ecological study, morbidity and mortality rates were estimated in the general population and in rural and urban areas. These measures were presented as proportions, where the numerator was the total num ber of cases of poisonings and fatal poisonings reported by the SIVIGILA. In addition, the denominator was the population estimate of the DANE for rural and urban populations and each year of the time series [21]. Mor bidity and mortality measures were adjusted for age by the indirect method [22].
Additionally, to establish the excess risk of morta lity associated with pesticide poisoning in the popula tion living in rural areas, the population attributable risk (PAR) and the corresponding 95% confidence intervals for each year were calculated [23]. The excess risk ex presses the difference between mortality in the general population and mortality in the rural population.
Finally, the aggregated means were analyzed through an exploratory time series analysis, using the epidemiological week as the time unit. First, mortality, morbidity and PAR trends were described in the time period with a simple linear regression. Then, the coe fficient (R2) was estimated for these indicators. The R2 indicator provided information on the strength of the as sociation and the direction of the relationship between mortality, morbidity, and PAR with time [24].
Subsequently, the time series was constructed using 52 epidemiological weeks for each year. Additionally, a seasonal decomposition was performed using the mul tiplicative model of the time periods. The data was pre sented as a graphical representation of the time series [22]. Additionally, a predictive time series model for one year was built. Finally, the trend component of the mo del was expressed with seasonal R2
Model fit tests
Different tests were used to evaluate the sensitivi ty of the models. In the case of Poisson regression, the chi-square goodness-of-fit test and the omnibus contrast were used [19]. In the case of the time series analysis, the independence of distribution was tested using the Box-Ljung test [21]. The Box-Ljung test allowed for the identification of the independence of the distribution.
Ethical Considerations
This project used the records of the SIVIGILA public health vigilance database. To carry out the study, all the variables that allowed the identification of the cases were eliminated, which guaranteed confidentiality and data management, complying with the biomedical re search regulations of the country.
Results
Between January 2007 and December 2017, 89 490 ca ses were reported to the SIVIGILA. A list of sociodemo graphic characteristics of the reported cases is presented in Table 1. A higher proportion of poisonings among men living in rural areas compared to those who lived in urban areas (66% vs. 49,9%: p < 0,001) was observed.
Regarding the types of cases, occupational exposure was found to be higher in rural populations (rural areas 26,3% vs. urban areas 9,6%, p < 0,001). Furthermore, suicidal intention was lower in rural areas than in urban areas (67,7% vs. 47,9%). In addition, it was observed that fatal cases occurred more often in populations living in rural areas (rural 2,8% vs. urban 1,1%), with poor ac cess to health services.
Table 1 Characteristics of Poisoning and Pesticides, Colombia, 2007-2017
| Characteristics of poisoning by pesticides 2007-2017 | Urban | Rural | Total | p | ||||
|---|---|---|---|---|---|---|---|---|
| N | % | N | % | N | % | |||
| Sex | Feminine | 25038 | 50.1% | 13410 | 34.0% | 38448 | 43.0% | .000* |
| Masculine | 24957 | 49.9% | 26085 | 66.0% | 51042 | 57.0% | ||
| Age groups (years of age) | < 14 | 8269 | 16.5% | 5194 | 13.2% | 13463 | 15.0% | .000* |
| 15-39 | 33430 | 66.9% | 26242 | 66.4% | 59672 | 66.7% | ||
| 40- 64 | 7184 | 14.4% | 6971 | 17.7% | 14155 | 15.8% | ||
| > 65 | 1104 | 2.2% | 1085 | 2.7% | 2189 | 2.4% | ||
| Level of education | Secondary or higher level | 28023 | 56.1% | 12758 | 32.3% | 40781 | 45.6% | .000* |
| Basic level | 21972 | 43.9% | 26737 | 67.7% | 48709 | 54.4% | ||
| Declaring an ethnic identity | No | 46426 | 92.9% | 34869 | 88.3% | 81295 | 90.8% | .000* |
| Yes | 3569 | 7.1% | 4626 | 11.7% | 8195 | 9.2% | ||
| Marital status | Married or domestic partnership | 19091 | 38.2% | 18399 | 46.6% | 37490 | 41.9% | .000* |
| Single | 30904 | 61.8% | 21096 | 53.4% | 52000 | 58.1% | ||
| Type of case with suicidal intention | No | 16139 | 32.3% | 20560 | 52.1% | 36699 | 41.0% | .000* |
| Yes | 33856 | 67.7% | 18935 | 47.9% | 52791 | 59.0% | ||
| Type of case as occupational exposure | No | 45181 | 90.4% | 29122 | 73.7% | 74303 | 83.0% | .000* |
| Yes | 4814 | 9.6% | 10373 | 26.3% | 15187 | 17.0% | ||
| Deadly case | No | 49440 | 98.9% | 38383 | 97.2% | 87823 | 98.1% | .000* |
| Yes | 555 | 1.1% | 1112 | 2.8% | 1667 | 1.9% | ||
| Case of hospitalizing because of poisoning | Yes | 30244 | 60.5% | 19594 | 49.6% | 49838 | 55.7% | .000* |
| No | 19751 | 39.5% | 19901 | 50.4% | 39652 | 44.3% | ||
| Type of case confirmation | Laboratory confirmation | 2000 | 4.0% | 1489 | 3.8% | 3489 | 3.9% | .002* |
| Clinical confirmation | 47478 | 95.0% | 37677 | 95.4% | 85155 | 95.2% | ||
| Epidemiological link | 517 | 1.0% | 329 | 0.8% | 846 | 0.9% | ||
| Exposure manner | Respiratory | 5431 | 10.9% | 9038 | 22.9% | 14469 | 16.2% | .000* |
| Oral | 38112 | 76.2% | 22624 | 57.3% | 60736 | 67.9% | ||
| Dermic/mucous | 5587 | 11.2% | 6956 | 17.6% | 12543 | 14.0% | ||
| Other exposure | 865 | 1.7% | 877 | 2.2% | 1742 | 1.9% |
For the Poisson regression model, it was found that the risk of lethality for suicide intention was higher among the rural population risk: 5,9 (95% CI: 5,0-6,9) than in urban areas risk: 3,0 (95% CI: 2,3-3,8). (pint < 0,001). The risk of lethality of ethnic and low educational levels had a significant association, increasing lethality in both rural and urban areas [Table 2].
In urban areas, the nonhospitalized population had a protective factor for lethality (RR: 0,856 (95% CI: 0,72-1,018): However, these associations were not sta tistically significant. conversely, the risk of dying was higher among the people who resided in rural areas who were not hospitalized (RR: 1,51 (95% CI: 1,34-1,7)). This difference of receiving hospital care and living in rural areas was statistically significant (pint < 0,001) [Ta ble 2]. Nevertheless, it was observed that the population older than 65 years living in urban areas had an RR of 4,3 (95% CI: 3,2-5,7).
Cases of suicidal intention for people living in rural areas who reported ethnic belonging showed a greater RR: 8,6 (95% CI 6,9-10,6) compared to the same popu lation in the urban areas, RR: 5,2 (95% CI % 3,7-7,4) (pint < 0,001) [Table 2].
Table 2 Factors associated with suicide and crude and adjusted Relative Risk (RR) in rural and urban populations
| Rural area | Urban area | P INT | |||||||||||
| Factor | RR unadjusted | IC 95% lower | IC 95% higher | RR adjusted* | IC 95% lower | IC 95% higher | RR unadjusted | IC 95% lower | IC 95% higher | RR adjusted* | IC 95% lower | IC 95% higher | |
| Suicide | 5.371 | 4.590 | 6.285 | 5.906 | 5.028 | 6.937 | 2.790 | 2.204 | 3.531 | 3.000 | 2.366 | 3.805 | <0.001 |
| Ethnic identity | 1.870 | 1.614 | 2.166 | 1.870 | 1.613 | 2.167 | 1.786 | 1.383 | 2.306 | 1.847 | 1.430 | 2.385 | <0.001 |
| No Hospitalization | 1.481 | 1.314 | 1.670 | 1.514 | 1.341 | 1.708 | 0.876 | 0.737 | 1.042 | 0.856 | 0.720 | 1.018 | <0.001 |
| Low Education Level | 1.303 | 1.141 | 1.488 | 1.332 | 1.164 | 1.524 | 1.626 | 1.375 | 1.922 | 1.568 | 1.324 | 1.857 | <0.001 |
| Older than 65 years | 2.234 | 1.742 | 2.865 | 2.340 | 1.822 | 3.006 | 4.578 | 3.441 | 6.091 | 4.307 | 3.227 | 5.749 | <0.001 |
| Suicide and ethnic identity (interaction) | 7.803 | 6.290 | 9.681 | 8.601 | 6.915 | 10.698 | 4.686 | 3.325 | 6.604 | 5.258 | 3.722 | 7.429 | <0.001 |
*Adjusted for age and sex
In addition, the accumulated rate of morbidity in the general population for all time periods was 24,18 cases per 100 000 inhabitants adjusted per age. In rural areas, the accumulated rate of morbidity was 36,03 cases per 100 000 inhabitants [Figure 1A). This indicator was three times higher than that in urban areas (12,33 ca ses per 100 000). The accumulated mortality rate for all cases was 0,57 cases per 100 000 inhabitants. The mor tality rates in rural and urban areas were 1,00 and 0,13 cases per 100 000, respectively.
In addition, an upward mortality and morbidity trend was found in the time periods in rural areas (R2 =0,65 and 0,16, respectively) [Figure 1B). Finally, there was a notable difference in the behavior of the mortality rate according to the place of occurrence of the event. The PAR was 69,6% (95% CI 54,2-89,3%). Additionally, an increasing trend of this indicator was observed throughout the time period (R2=0,2498) [Figure 1C).
Figure 1 A) Morbidity by pesticide poisoning in rural and urban areas. B) Mortality by pesticide poisoning in rural and urban areas C) Risk attributable to mortality rural areas, 2007-2017
Regarding the time series analysis, there was no evi dence of a seasonal factor in mortality and morbidity be cause this indicator did not vary more than 20% between time periods. Additionally, an exponential smoothing model was selected only for morbidity in rural areas, because this indicator adjusted for distribution indepen dence (p: Ljung-Box Q: 0,180); this model showed an upward trend (R2 =0,279) [Figure 2]
Discussion
The objective of this work was to analyze pesticide poi soning in rural and urban populations through a time se ries analysis and estimate the relative risks associated with lethality in Colombia. More than 87 000 cases of exogenous intoxication were found during 2007-2017. To our knowledge, this is one of the largest reported re cord series of acute pesticide poisoning [7].
A seasonal factor was not observed in the time se ries analysis. Nevertheless, a growing trend in morbidity and mortality was observed, mainly in rural areas.
An increase in the population attributable risk of mortality was observed among rural populations. Addi tionally, an increased risk of lethality related to pesticide poisoning was found associated with living in rural areas and ethnicity. These results were similar to those of se veral studies, including that of Wang and associates [25], who performed a descriptive cross-sectional study in the province of Jiangsu, and reported 38 513 cases of pesti cide poisoning over 12 years. Out of the cases reported in that study, 71,69% were due to suicidal intention, but the report did not study the difference between rural and urban populations. The study by Gyenwali and associa tes [11], a descriptive epidemiological study, found 439 cases of acute pesticide poisoning from 12 districts of Nepal, of which 89,5% were due to suicidal intention; 71,3% occurred in rural residents and 37,8% reported ethnic belonging.
Additionally, a positive interaction was found bet ween suicidal intention and living in a rural region with an increase in lethality for pesticide poisoning. These results are comparable with some reports from the lite rature that describe the practice of taking pesticides as a usual method of suicide in rural areas [26]. In Gujarat, India, of 1092 identified poisoning cases, 69,59% were in rural areas, and 42,94% were by suicide, in Malaysia, a total of 11 087 pesticide poisoning cases were recorded and 93.6% of cases of intentional pesticide poisoning were caused by suicide attempts [12,27]. In addition, the WHO indicates that suicide has increased in agricultural workers due to poor harvests [28].
A high proportion of cases of suicidal intention due to acute pesticide poisoning was reported, particularly in rural areas. This may be explained by the easy access to pesticide products in the rural context. Troya and asso ciates [29] showed that access to pesticides can be iden tified as a suicide risk factor in rural areas. Likewise, in Latin America, it has also been stated that the characte ristics of suicides in rural areas include the use of pes ticides as a common method [30]. Finally, several stu dies reported high suicide and lethality rates for suicide with pesticides in ethnic populations [31,32]. Therefore, the WHO and The Food and Agriculture Organization highlight the effects of restrictive measures against the use of pesticides. For example, programs adopted in countries such as Sri Lanka, where prohibition and se vere restriction of pesticides led to a 70% reduction in suicides, and an estimated 93 000 lives were saved bet ween 1995 and 2015 [28].
Studies have documented the increase in suicide cases in the population with chronic exposure to pestici des in recent years [33,34,35]. This suggests that future research with prospective designs should consider the chronic and acute exposure to pesticides, the thoughts and behaviors in different ethnic groups, and the social factors associated with suicide in rural populations.
One of our main findings was an increasing trend of suicidal behavior, especially in rural areas, but this trend did not have an association with seasonality. This may be explained by the fact that Colombia has rainy and dry seasons. This allows farmers to have intensive farming practices during the entire year that require a greater usage of pesticides. Growing pesticide commercializa tion has been observed in Colombia during the last ten years [36]. In contrast, some studies carried out on far mers from nontropical areas have reported an increase in suicide cases during the summer period [25,37].
In addition, our analysis showed an increased risk of lethality in reported cases that occurred in rural areas and for those who did not receive medical attention. The lack of medical attention in rural areas has been highlighted as a determinant of poor clinical outcome in these areas [28].
An accumulated mortality rate of 0,57 cases per 100 000 inhabitants was observed in the 2007 - 2017 period. In contrast, other studies in a similar time period have reported decreasing trends in mortality for populations in India and Germany [25,37]. However, a higher morta lity rate was reported for the populations in El Salvador and Nicaragua in a similar period [38]. Finally, a study carried out in Colombia using death certificates to de termine deaths from pesticide poisoning, reported 4835 deaths between 1998 and 2011, and an age-adjusted rate of 2,38 deaths per 100 000 inhabitants with annual death averages higher than our findings and a decreasing trend in mortality since 2002 [14].
Our study had several limitations. First, there are li mitations related to the source of the information. This study was based on records from epidemiological sur veillance systems for patients who received medical at tention. This may introduce a record bias. This disparity shows a potential source of underregistration for fatal and nonfatal events, especially for cases of unintentional mild work-related poisoning. Furthermore, these data are not associated with ICD-10 codes, introducing an additional bias in the analysis. In addition, the definition of intoxi cation of the Surveillance Protocol Poisoning by chemi cal substance from INS 2011 and 2017 was used for our study. The definition remained almost unchanged during the described period of time. Nevertheless, the definition and protocol for the external poisoning from 2007 to 2010 wasn’t available and this may represent a bias in the data collected in these period of time.
Second, it was not possible to establish which pro ducts were involved in suicidal intentions. There was a lack of information related to the active ingredient as sociated with poisoning in more than 20% of the repor ted cases. Although Colombia has adopted the Globally Harmonized System since 2015, information related to the active ingredient in the product is not often known by rural farmers. Workers and relatives tend to report the popular names of chemical products, which can vary from region to region. This implies a limitation for clini cal attention and the surveillance system.
Finally, a common bias of time-series studies inclu des confusion resulting from measured and unmeasured time, affecting the relationship between the main expo sure and the result. We tried to control this bias by inclu ding additional explanatory variables in the regression model. However, this research did not have access to medical records to include additional variables that may have affected the outcomes.
In conclusion, the mortality and morbidity for pesti cide poisonings in rural areas of Colombia have shown a trend of increase throughout the period 2007-2017. A higher risk of lethality associated with these events occurred in rural populations belonging ethnic. And, in population was reported as suicidal intent. These results can be used by health and academic institutions to pre vent non-occupational and occupational poisoning by pesticides. The Surveillance Protocol Poisoning by che mical substance should include a differential rural and ethnic perspective that should be included in the public policy instruments. In addition, it should be tacked into account that the clinical guidelines may consider socio demographic aspects in order to prioritize management of referral to specialized clinical centers. Finally, the sur veillance may include a preventive actions through the local health and safety committees and health bureaus.
