Local attributable burden disease to PM 2 . 5 ambient air pollution in Medellín , Colombia , 2010 – 2016

Background: Exposure to 2.5-micron diameter air pollutants (PM2.5) has been associated with an increased risk of illness and death worldwide; however, in Latin American health impacts assessment of this risk factor is scarce. Medellín is one of the most polluted cities in the region, with a population growth rate that is twice as high as that of other Colombian cities, which implies a growing population at risk. Methods: A descriptive study of the disease burden was carried out using the city as the unit of observation. Health events were selected based on epidemiologic evidence and the availability of the population attributable fraction associated with PM2.5. The mortality records were taken from the module of deceased of the Single Registry of Affiliates of the Health System; the morbidity records were taken from the Individual Health Services Registries. For the estimation of the burden of disease, the current Global Burden of Disease guidelines were followed. Results: Attributable disability-adjusted life years to exposure to ambient PM2.5 pollution (DALYsPM2.5) constituted 13.8% of total burden of the city. Males showed the greatest loss of DALYsPM2.5 due to acute events, while in women the greatest loss was due to chronic events. Ischemic heart disease, chronic diseases of the lower respiratory tract, and influenza and pneumonia were the events that contributed the most to DALYsPM2.5. 71.4% of the DALYsPM2.5 corresponded to mortality, mainly in the population over 65 years of age. Regarding attributable morbidity, acute events were more prevalent in both sexes, especially due to respiratory diseases Conclusion: Premature death among the elderly population has the greatest weight on burden of disease attributable to ambient PM2.5 pollution, mainly due to respiratory and cardiovascular diseases, without significant differences according to gender. Open Peer Review


Introduction
Air pollution is one of the main concerns surrounding public health worldwide due to its impacts on human health and ecosystem (Dominici et al., 2006). According to the Environmental Performance Index of Yale University, poor air quality is the greatest environmental threat and the most difficult challenge for public policies in middle-and low-income countries (Wendling et al., 2020). The attributable disease burden to ambient air pollution at a global, regional, and local level has been widely documented, this involves the measurement of disability-adjusted life years (DALYs), an aggregate value of years of life lost (YLLs) due to premature death and years lived with disability (YLDs) (HEI, 2019). Globally, the Institute for Health Metrics and Evaluation (IHME) has identified to air pollution as the fifth main health risk factor for the population, and it is estimated that the exposure to PM 2.5 contributes to 4.9 million deaths (8.7% of all deaths worldwide), and the loss of 147 million of healthy life years (5.9% of all DALYs). The main causes of mortality worldwide due to air pollution are ischemic heart disease (35.9%), stroke (21.1%), chronic obstructive pulmonary disease (COPD; 20.4%), acute respiratory infections (15.9%), and lung and respiratory tract cancer (6.9%) .
The World Health Organization (WHO) considers air pollution to be the main environmental health risk factor for the population in the Americas (Prüss-Ustün et al., 2016) due to its impact in susceptible populations, such as children younger than 5 years old, pregnant women, and elderly people (Piñeros-Jiménez et al., 2018). For Latin America and the Caribbean, it has been estimated that around 35,000 persons die annually to urban air pollution and 276,000 annual healthy life years are lost (Romieu et al., 2012).
In Colombia, few studies on ambient air pollution epidemiology have been carried out: a review of literature between 2008(Piñeros-Jiménez et al., 2018, 19 works were identified, which were mainly focused on the population-based analysis of the risk associated with the exposure to air pollutant on morbidity and mortality due to cardiovascular and respiratory events (Blanco-Becerra et al., 2014;Rodríguez-Villamizar et al., 2010;Salazar et al., 2011;Gaviria et al., 2011). Recently Rodríguez-Villamizar et al. (2018) conducted a multicity ecological time-series analysis with data from four major cities in the country. This analysis found for NO 2 , PM 10 , and PM 2.5, statistically significant percentage increases in emergency department visits for respiratory diseases in children between 5 and 9 years old, and for circulatory diseases in persons over 60 years of age.
Regarding studies of burden of disease attributable to ambient particulate matter pollution, few robust studies have been conducted (Romieu et al., 2012). In 2016, the Colombian National Health Institute estimated that 8% of 200,000 annual deaths in the country could be attributed to environmental risk factors, and calculated that 13.9% of ischemic heart disease deaths and 17.6% of chronic obstructive pulmonary disease deaths could be associated to ambient particulate matter pollution (INS, 2018). Also, the World Bank (WB) estimated 5,000 premature deaths and 69 million DALYs annually between 2002(Golub et al., 2014. These studies provided meaningful results of impact of air pollution in health in Colombia; however, they did not determine the magnitude of burden disease at the municipal levels where environmental phenomenon presents different magnitudes and particular local dynamics. Medellín is the second largest city in Colombia and one the most polluted in Latin America. Since 2016, Piñeros-Jiménez et al., have conducted several researches about health impacts associated with short-term exposure to PM 10 , PM 2.5 and ozone (Piñeros-Jiménez et al., 2018;Piñeros-Jiménez et al., 2019). They developed traceability techniques based on analytical methods to identify health events from the records of different sources of health information, which allowed them to have a more precise epidemiological baseline for the measurement of health impacts. For Medellín, an ecological time series study found that a 10 μg/m 3 increase in PM 2.5 was associated with increase of 25.2% in the risk of respiratory diseases for children younger than 5 years old, and of 29.7% for adults of 65 years of age and over (Piñeros-Jiménez et al., 2018). Furthermore, they could establish the increases in the risk percentage of emergency room visits for asthma (2.8%), acute respiratory infection (2.0%) and pneumonia (2.2%) due to population exposure to critical air pollution episodes occurred in February andMarch 2015 (Nieto-López et al., 2020).
Despite the advances in air pollution epidemiology research in Latin America, it has been recognized that establishing the attributable burden of disease to this environmental risk factor at the municipal level is a challenge for the academy and local management of environmental health, which requires updated information of the best quality to guide political decision-making and public planning with a territorial perspective. The aim of this study was to determine the local burden of disease to PM 2.5 (LBD PM2.5 ) for Medellín, for which an updated epidemiological baseline focused on respiratory and circulatory events.

Type of study and population
A descriptive study was conducted using the city as the unit of observation. The study population were all the residents in Medellín, the second major city in Colombia, between 1 January 2010 and 31 December 2016. Medellín is located in Aburra Valley in the Andes Mountains to the central-western of Colombia and an estimated population of around 2.9 million people. Additionally, it has the fastest demographic growth in the country, with a growth rate twice that of other cities.
Air pollutant data analysis PM 2.5 data were obtained from the air-quality monitoring network in Medellín. These data correspond to validated and adjusted information from eight monitoring stations distributed across the city between 1 January 2008 and 31 December 2016. Daily 24-h averages of PM 2.5 were calculated. Data quality analysis identified gaps in each available pollutant dataset which were filled with the R package nnet (RRID:SCR_001905) for data imputation using an artificial neural network (Villa-Garzon, 2018). This package was also used for to obtain a unique assembled dataset for PM 2.5 that represents the air pollution exposure for Medellín from information of different air monitoring stations. Daily, monthly and annually average concentrations of PM 2.5 were calculated from this dataset.
Data source and procedures for health information All available data on morbidity and mortality population residing in Medellín during the study period were used. Data of individual deaths was obtained from death records in the deceased module of the Single Registry of Affiliates of the Health System (Registro Único de Afiliados, módulo de Defunciones (RUAF-D)). Information related to emergency department visits, medical outpatient services visits, and hospitalizations were obtained of individual records on the Provision of Health Services (Registros Individuales de Prestación de Servicios de Salud (RIPS)). All data were provided by the Social Protection System (Sistema Integrado de la Información de la Protección Social (SISPRO)). The records of people whose basic cause of death or main medical diagnosis was an event related to air pollution were included, following International Classification of Disease, ICD-10.
For each health data source, a quality assessment of data was performed, considering the dimensions of completeness, consistency, accuracy, duplication and integrity for each individual record that complied with the inclusion criteria such as place of residence, year of death/medical attention, and causes of morbidity and mortality selected for this study. Underreporting of deaths was estimated with the Preston and Coale method, and the PAHO method of proportional distribution to address potential information biases (ONU, 1986). For morbidity, underreporting could not be controlled because there was no other source of information that could contrast the source used. The reference population were people living in Medellín based on time, age, and gender criteria, according to projection of the population census published by the National Department of Statistics of Colombia.
Traceability strategies were defined for each event in order to identify prevalent cases in each year of analysis. These were previously designed by the research team for local studies based on secondary data (Piñeros-Jiménez, 2018), which used the descriptive model of the natural history of disease, which presents the course of all biological events, the sequential action of causes (etiology), the evolution of the disease and its outcomes (recovery, chronicity, disability, or death), as well as the pre-pathogenesis and pathogenesis phases of the disease. Annual event per patient was included and pointprevalence was estimated.
Determining burden of disease WHO's methodology for estimation of burden of disease was used (WHO 2017). DALYs calculation incorporated the number of YLLs due to premature death and the number of YLDs (Rutstein et al., 1983). YLLs estimation used the standard method, which includes all deaths at any age within the total estimated disease burden. As a standard value, the frontier national life expectancy projected for the year 2050 was considered, with a life expectancy at birth (LEAB) of 91.9 years for both men and women (Murray, 1995). The equation used for calculations was: YLL ¼ P L x¼0 dx e * x , where L = the ultimate age of survivors; x = age of death; dx = number of deaths at x age in years; e x * = life expectancy at each age based on an ideal standard.
Due to the availability of aggregated data by cause of death according to sex, age group and year, the class mark was defined as the representative value of all age intervals in the calculation of the indicator. Premature death was calculated with the difference between the class mark of the respective age group and the LEAB standard value for each one record of the database.
For calculating YLDs, the methodology of 2013 GBD guidelines by WHO was used (WHO 2017;Salomon et al., 2015), not including discounting rate of 3% and age weights. For each study event, the distribution of cases in each year were calculated according to gender and age group. Them these were divided by the number of inhabitants in Medellín in order to find the point prevalence. The disability weights per event were calculated as the following equation: YLD ¼ D j Â p j where D j = the disability weight for each individual j cause, and p j = the prevalence of the j th disease. Therefore, the total of annual YLDs per event in the study period corresponds to the sum of individual YLDs per age group and gender.
DALYs were obtained from the sum of the total number of YLLs due to premature death and the total number of YLDs per year, gender, age group, and subgroup of diagnostic cause for each type of event (acute or chronic).
Estimating local burden disease to PM 2.5 pollution To determine the magnitude of LBDPM2.5, exposure is expressed as the fraction of disease or death to the risk factor in a population and referred to as the population-attributable fraction (PAF). Due to this requirement, in the case of Medellín, it began by defining the events to study. Air pollution-associated events were limited to those causes that have been examined in GBD studies, which already had PAF data for PM 2.5 pollution, according to the IHME measurement results for Colombia .
After obtaining the frequency of YLLs, YLDs, and DALYs per event, LBD PM2.5 was calculated. This was done by considering the standardized PAF by age estimated for Colombia  in relation to each one of the diagnostic groups examined in this study. LBD PM2.5 was estimated by using the following equation: LBD PM2.5 = (YLLs or YLDs or DAILYs) Â PAF PM2.5 Grisales-Romero et al., 2021).
Results are shown with absolute and relative frequencies along with rates/indices of each indicator considered with a constant value of 100,000 according to Medellín's population for each year. They are complemented, where necessary, with the 95% uncertainty interval (95% UI) generated using the Bootstrap method, a resampling technique (Efron & Tibshirani, 1993). Data capture, storage, and processing was performed using the database management software pgAdmin 4 v2.1 ® (RRID: SCR_021066). For the generation of results and graphs, the commercial software, Microsoft Excel ® (RRID:SCR_016137) was used. A free office suite alternative that could also be used for this process would be LibreOffice, which is available at https://www.libreoffice.org. February and April showed a higher trend towards a monthly average increase of PM 2.5 pollution during the study period. In March 2016, the highest average values of the whole time series data were identified. On the other hand, October and November from 2010 to 2014 presented monthly averages with slight increases, and even for the years 2015 and 2016 the trend was downward during these months ( Figure 1A-G).

Research ethics
Mortality and morbidity to PM 2.5 pollution There was a total 28,678 of deaths for acute and chronic diseases associated with air pollution in Medellín between 2010 and 2016, of which 3,873 deaths (13.5%) were attributed to PM 2.5 pollution. The attributable deaths to PM 2.5 pollution per year were similar during the seven years of studied period. 61.7% (n = 2,391) of them were for acute events, of which 75% (n = 1,793) were due to circulatory system diseases, mainly ischemic heart diseases (n = 1,550). Among diseases of the respiratory system, influenza and pneumonia showed the highest frequency in the study period (n = 598). Chronic lower respiratory tract diseases were the most frequent (n = 1081) of chronic events; 95.7% corresponded to chronic obstructive pulmonary diseases (Table 1).  During the study period, 567,505 prevalent cases for events associated with air pollution were identified, 75.6% (n = 428,858) were acute events, of which 92.3% (395,761) were due to respiratory system diseases. 88,083 (15.5%) prevalent cases were attributed to PM 2.5 pollution. Respiratory system diseases were the most frequent for acute attributable events, mainly for acute lower respiratory tract infections (n = 39,163). Among chronic events, the most frequent were chronic lower respiratory tract diseases (n = 21,479) ( Table 2).

Local attributable burden of disease
During the seven-year period there was a premature YLLs of 536,772 (95% UI 524,048-549,495) due to events related to air pollution in Medellín; 13.0% (n = 71,590 (95% UI 69,843-73,339)) corresponded to LBD PM2.5 (Table 3). The attributable YLLs to PM 2.5 pollution (YLLs PM2.5 ) showed little variation throughout the period, with an annual average of attributable mortality burden calculated in 10,227 (95% UI 9,950-10,505) of YLLs. The Figure 2A-B shows annual YLLs PM2.5 rates for males and females between 2010 and 2016. In both genders, stable losses were observed in the YLLPs PM2.5 rates with a tendency to decrease. The highest concentration of YLLs PM2.5 were presented in 2016, with annual rates for this year of 496.9 per 100,000 inhabitants for males and 378.6 cases per 100,000 inhabitants for females.   Figure 2C-D). Acute events constituted 56.3% of YLDs PM2.5 (n = 16,124 (95% UI 14,902-17,346)). Females had a higher percentage contribution, both for YLDs PM2.5 due to acute events (51.4%) and YLDs PM2.5 due to chronic events (50.8%). However, throughout the study period the highest rates occurred in men.
Of total LBDs associated with pollution (n = 726,353 DALYs, 95% UI 715,045-737,659), 73.9% corresponded to YLLs and the remaining to YLDs. The premature mortality due to acute events was higher than that estimated for chronic events, with 33 percentage points of difference, with a slightly greater magnitude in the male burden than the female burden (Table 1). Of total estimated DALYs, 13.8% (95% UI 13.7%-13.9%) was attributed to PM 2.5 pollution (DALYs PM2.5 ), with an average of 14,315 DALYs PM2.5 per year (95% UI 14,005-14,626), equivalent to rate of 592.2 DALYs PM2.5 per 100,000 cases. Compared with the initial year of study in 2016, a decrease of 8.3% in total DALYs PM2.5 was observed, with decreases of 8.1%in DALYs PM2.5 due to acute events and 8.6% due to chronic events. According to the gender, distribution of DALYs PM2.5 had variations by type of event; while DALYs PM2.5 due to acute events was higher in males (54.1%), the DALYs due to chronic events were slightly concentrated in females (50.8%) (Table 3) ( Figure 2E-F). Attributable burden to PM 2.5 pollution by age group and type of event LBD PM2.5 showed a positive gradient from 35 years of age, with significant differences among the five-year age groups. Those over 65 years of age contributed with 61.7% of the total burden, with the highest rates in the group of 80 years and older. In all the five-year age groups, LBD PM2.5 for acute events was higher than for chronic events, with considerable differences that were more marked until the age of 59 years. After 60 years of age, the differences were smaller although maintaining the trend. LBD PM2.5 for chronic events was comparable in the age groups between 15 years and 44 years; but after 45 years of age, it gradually increased until the age of 79 years, when it slowly decreases in the last age group. Only 10.3% of the total attributable burden was in the population younger than 39 years old (Table 4). Male showed the highest prevalence of LBD PM2.5 , with the highest difference in the 40-44 and 55-59 age groups, and an approximate ratio of 2 to 1.
Five groups provided the highest contribution to attributable premature mortality and disability according to the type of event by diagnosis group (chapter) of the seven groups analyzed (Table 5). 80.6% of premature deaths were caused by ischemic heart disease (40.1% of YLL PM2.5 ), chronic lower respiratory tract diseases (23.4% of YLL PM2.5 ) and acute infections of the lower respiratory tract (16.9% of YLL PM2.5 ). On the other hand, among the causes that most contributed to attributable disability were chronic lower respiratory tract diseases (43.3% of YLD PM2.5 ), other acute infections of the lower respiratory tract (24.3%) and influenza and pneumonia (20.1% of YLD PM2.5 ). In contrast, the group of neoplasms had a relatively lower contribution to LBD PM2.5 , explaining 8.8% of DALYs PM2.5 (12.2% of YLL PM2.5 and 0.3% of YLD PM2.5 ).   Period 2010-2016, Uncertainty ranges of 95% calculated using the bootstrap method for n = 10,000 samples. Figure 3 shows the relationship of the rates of YLL PM2.5 and YLD PM2.5 for each diagnosis group studied. The groups of causes were classified into four categories according to the relationship between the two rates: A (low mortality and low disability), B (low mortality and high disability), C (high mortality and low disability) and D (high mortality and high disability). Four of the seven groups of events studied (ischemic heart disease, cerebrovascular disease, influenza and pneumonia, and chronic lower respiratory tract diseases) were included in category D, due to their contribution to both mortality and disability. On the other hand, in situ neoplasms were included in category A, other acute lower respiratory tract infections in category B, and malignant neoplasms of respiratory and intrathoracic organs in category C. Table 6 shows the ranking of the diagnostic events according to their contribution to the LBD PM2.5 . Comparing the years of beginning and end of the study period, there were no changes in the order of the groups of events. In both years, the three highest DALY PM2.5 rates were found to be due to ischemic heart disease, chronic lower respiratory tract disease, and influenza and pneumonia. Cerebrovascular diseases, chronic lower respiratory tract diseases and in situ neoplasms showed a reduction of more than 10% in the rates of DALY PM2.5 by 2016 in relation to the year of initiation of the study. The opposite occurred with malignant neoplasms of respiratory and intrathoracic organs, which showed an increase of 22.3% in DALY PM2.5 rate.

Discussion
Nowadays, ambient air pollution by the criteria pollutant PM 2.5 is considered one of the biggest environmental problems at the global and local level due to the impact it causes on the health of populations (Apte et al., 2018). Although cities in Latin America and the Caribbean region have shown annual average PM 2.5 values that could be considered moderate if compared to cities located in Southeastern Asia and India, most of these have reported annual average higher than the levels recommended by WHO of 10 μg/m 3 (Riojas, 2016). For Medellín, one of the most polluted cities of the region, the annual average PM 2.5 levels during the seven years of study was 35.6 μg/m 3 , where more than 90% of the days had daily averages above 25 μg/m 3 (Piñeros-Jiménez et al., 2018), risk factor to which approximately 3 million people are exposed. Under these conditions, in the last five years, some studies have been carried out to establish the health impacts of criteria pollutants (PM 10 , PM 2.5 , ozone and nitrogenous) with population models of a single and multiple pollutants at a local level (Piñeros-Jiménez et al. 2018, Piñeros-Jiménez et al. 2019, Rodríguez-Villamizar et al. 2018. This is the first study that seeks to establish the magnitude of such an impact by using a holistic indicator.  Since the 1990s, multiple strategies have been used to gather knowledge about the burden disease caused by different risk factors. These strategies offer a holistic view of the joint effects of morbidity and mortality in the number of healthy life years lost. Environmental risk factors, mainly air pollutants, have had a growing interest among decision makers and the community at the local and global level, and have been prioritized in the political agendas promoted by research in environmental epidemiology in recent decades. Among the advances in research, the studies that analyze the magnitude of the impacts based on holistic indicators at the global and national level, supported by exposure-response functions and relative risk analysis, with the data available in the health and environmental information systems, stand out. All of which may help in transcending towards causality (Burnett et al., 2014).
Air pollution epidemiological research in Latin America and Colombia has been characterized by ecological studies of time series and some panel studies, which have confirmed the short-term effects associated to criteria pollutants (Romieu et al., 2012;Piñeros-Jiménez et al., 2018). Very few studies have been carried out at the local level to document the local burden of disease attributable to air pollution (Golub et al., 2014;INS, 2018). No one has evaluated the long-term effects.
In this study, we provided a detailed view of the local burden of disease attributable to air pollution, for Medellín, one of most development cities of the country and region. Which recognizes the local character of the epidemiological phenomenon associated to air pollution, and the need for updated information to influence public environmental policy in the city. It was focused on using a synthetic indicator: DALYs, an aggregate value YLLs due to premature death and YLDs, based on up-to-date methodologies validated by experts and international organizations World Bank, 2016;WHO 2016). In the absence of local cohort studies, the PAF estimates made in the GBD study were used, which assumed a non-linear relationship between the incidence of health events and short-and long-term exposure to particulate matter and developed the integrated exposure-response curve to estimate the long-term PM 2.5 exposureresponse association from low exposure level to concentration as high as 1000 μg/m 3 to avoid overestimating the magnitude of health effects (Burnett et al., 2014) We found that the attributable burden disease PM 2.5 pollution constituted 13.8% of the total local burden of DALYs for all the selected pollution-related events (100,208 DALYs PM2.5 out of 726,352 total DALYs,). In the study on environmental burden disease in Colombia carried out by National Health Institute of Colombia (INS), for 2016, 19% of the total national burden disease was associated to environmental risk factors (air, water, and other similar factors) (INS, 2018). Regarding air pollution, YLLs PM2.5 was calculated in 619.8 per 100,000 inhabitants. In our study, for the same year, 434.3 YLLs PM2.5 per 100,000 inhabitants was estimated. This difference can be considered reasonable, since the study cited above was carried out in a higher geographical area (national level), and it is possible that cities with a greater air pollution problem such as the capital of the country (Bogota D.C.) had a greater weight in the total estimate.
Acute diseases contributed with 62.0% of DALY PM2.5 , this is explained by the higher contribution of YLLs PM2.5 (74.0%) in the burden health index, mainly by ischemic heart disease, that constitute 30.5% of DALY PM2.5 for 2016 in contrast to the estimation of 15.8% from the study on the environmental burden disease in Colombia for this year (INS, 2018).
Our results are consistent with the results of the study by  regarding the events but not the magnitudes. The number of DALYs associated to ischemic heart disease was the highest in this study, similar results was found in India State-Level Disease Burden Initiative Air Pollution Collaborators study for 2017 (2019). There is more and more evidence from prospective studies that PM 2.5 exerts adverse effects particularly on the cardiovascular system, contributing substantially (mainly through mechanisms of atherosclerosis, thrombosis and inflammation) to coronary artery disease, but also to heart failure, hypertension, diabetes and cardiac arrhythmias (Hu, et al., 2018;Hajat, et al., 2019;Franklin, et al., 2015;Kauffman, et al., 2016a;Kauffman, et al., 2016b;Song, et al., 2020), which can help to explain the findings regarding this type of event in relation to the attributable burden disease.
In the case of the burden of disease attributable to acute respiratory diseases, our result differs from other studies conducted with national data. This can be observed in the estimated proportion of DALY PM2.5 for acute lower respiratory tract disease in the INS study was of 13.7% (INS, 2018), which contrasts with the 7.1% reported by us for the events of the diagnostic group of other acute lower respiratory tract infection events (J20-J22) in Medellín. Regarding the study by Golub et al. (2014), we found more approximate results, although the basis for calculating the attributable DALYs in the case of acute respiratory disease was the proxy for the reports of respiratory symptoms. These differences could be explained by the methodology used to identify the cases and calculate the indicators. In our study we used a data miningbased traceability strategy to select health events that was constructed based on the natural history of the disease. Which established that the duration of acute respiratory events was 15 days, and with this criterion grouped the records of each source of information into a data set that made up each case (Piñeros-Jiménez et al., 2018). This strategy allowed a more precise approach to the number of events and the calculation of the study indicators. This strategy allowed a more precise approach to the number of events and the calculation of the study indicators.
Chronic diseases of the lower respiratory tract constituted the chronic events with the greatest weight in the total of DALYs PM2.5 (38%) in our study. This result is consistent with estimations of DALYs due to chronic respiratory diseases in relation to the IHME reports. Between 2010 and 2016, the average rate of DALYs due to chronic diseases of the respiratory tract and ischemic heart diseases in Medellín, was of 172.9 and 181.7 per 100,000 inhabitants, respectively. Estimations published by the IHME show that, in 2016, around 112.2 DALYs due to chronic respiratory diseases were calculated per 100,000 inhabitants in Colombia. Also, in this period, 165.32 (106.5-217.9) DALYs due to ischemic heart disease were calculated per 100,000 inhabitants, which were associated to air pollution .
Among the events included in the diagnosis group of chronic lower respiratory tract diseases were asthma (J45) and COPD (J40-J42), two of the most relevant events in relation to the attributable burden disease to exposure to air pollution, specifically for the pollutants PM 2.5 , NO 2 and ozone (Achakulwisut et al., 2019;Anenberg et al., 2018. In the case of COPD, although its impact on GBD has been established with a global point prevalence of 3.918% (95% UI (3.5111-4.3201) and a mortality rate of 41.9 deaths per 100,000 people (5.7% of all deaths from all causes) (GBD Chronic Respiratory Disease Collaborators, 2020), its risk association with some air pollutants has had a positive sign but of low magnitude and in some cases without statistical significance (Schikowski et al., 2005;Schikowski et al., 2014), including its specific association with ambient PM 2.5 pollution (Atkinson RW et al., 2015;Dany Doiron et al., 2019). Globally, seven million deaths were attributed to the joint effects of environmental and domestic air pollution . And it is recognized that the effects can be divided into short and long-term effects, ranging from exacerbation of existing symptoms, impaired lung function and increase in hospitalization and mortality rates. Prolonged exposure to air with a high concentration of pollutants can also increase the incidence of COPD.
Air pollution can induce the development of asthma, increasing respiratory morbidity and mortality, particularly in minority groups (Nishimura et al., 2013). Annually, it is estimated that 40 million (95% UI 18-52) of new cases of pediatric asthma could be attributable to NO 2 contamination, with a higher burden of new asthma cases associated with NO 2 exposure per 100,000 children for the Latin American region (340 cases per year, 95% UI 150-440) (Achakulwisut et al., 2019). Likewise, it is estimated that 9-23 million and 5-10 million annual asthma emergency room visits globally could be attributable to ozone and PM 2.5 , respectively (Anenberg et al., 2018). Our results demonstrate an important impact of ambient PM 2.5 pollution on chronic respiratory diseases that should be analyzed in more detail, with the aim of generating useful knowledge for the design of interventions and decision-making in specific groups, especially taking into account, age groups where asthma and COPD are more frequent.
In contrast with other reports, the contribution of lung and airway cancer to local DALYs PM2.5 was relatively low (8.8%), and it was explained almost exclusively by YLLs PM2.5. The Cohen study found that lung and airway cancer accounted for 16.5% of the global burden attributable to PM 2.5 , while recently Yin et al. reported for China that 16.6% of deaths attributed to ambient particulate matters at the national level were due to this this cause Yin et al., 2020). This difference of more than double the DALYs PM2.5 could be due to the level of analysis carried out in the different studies, but it is mainly explained by the coverage of high-cost disease care in our country, where there is a health system with great inequalities in access, which makes these types of diseases often go undiagnosed and unattended, so they are not reported to health information systems. This situation is especially worrying in the economically most vulnerable sectors of the population, there is a hidden burden of the disease that makes difficult to carry out targeted interventions according to equity criteria.
The highest proportion of DALYs PM2 was related to YLL PM2.5 , at 71.4%. This apparent paradoxical effect of a significant premature mortality burden attributable to PM 2.5 pollution could be explained using the protracted polarized model, which is predominant in Latin America and other regions of the world sharing the same development features. Here we find a superposition of transmissible acute diseases and non-transmissible chronic diseases, along with the reappearance of emergent diseases, which affect vulnerable human groups discriminately and disparately, as observed in the population of Medellín. Despite the improvements in the control of childhood diseases, as well as demographic changes and the increased life expectancy, conditions of inequality, poverty and extreme poverty persist, which add to the new risk factors, such as PM 2.5 pollution caused by transportation and urbanization (GBD 2019Risk Factors Collaborators, 2020. This new risk factor generates a negative impact on air quality, affecting the exposed population, accelerating disease and increasing mortality, and resulting in an unbalanced pattern of death and disability. It is worth mentioning that the effects of air pollution on health are not directly proportional to the intensity of exposure in individuals, but it is clear that such effects exist and are dependent on other variables as well as the time and level of exposure. Air quality in high-income countries has improved in recent decades, however, adverse effects of external air pollution on health due to particulate matter continue to be a public health problem worldwide, even if levels are low Liu et al., 2021).
We can conclude that 71.4% of DALYs PM2.5 between 2010 and 2016 was due to premature death; especially YLLs PM2.5 due to acute events. A proportion of 28.6% was YLDs PM2.5 . The greatest concentration of YLLs PM2.5 was associated to ischemic heart diseases and chronic lower respiratory tract diseases with a high proportion of COPD, particularly in male older than 80 years of age. Considering YLDs PM2.5 , these were caused mainly by chronic lower respiratory tract diseases, influenza and pneumonia, and other acute lower respiratory tract infection which were more prevalent among the population older than 60 years of age, in both genders.
One of the greatest strengths of this study is that it was pioneers in Colombia and region in the use of the new proposed methodology to calculate the burden of disease indicators, as it uses the new GBD study (GBD 2019Risk Factors Collaborators, 2020, which gives a higher importance to prevalence over incidence as epidemiological indicators for disability calculations, this allows establishing the basis for future studies based on our findings. The quality assessment of data from the information sources we used exhaustive and included protocols that have been previously validated (Piñeros-Jiménez et al., 2018;Piñeros-Jiménez et al., 2019), in contrast with the majority of GBD studies where it has been shown to be one of the greatest obstacles in the application of the disease burden methodology. We obtained consistency percentages of more than 90% in morbidity aspects and more than 82% in mortality aspects, which are comparable to the results of INS and Ribotta studies (INS, 2018;Ribotta et al., 2019).
Among the limitations, we can mention that the multiple sources we used respond in an unarticulated way to events follow-up, which does not allow for an optimal traceability. The effects of migration could not be considered, and it was impossible to make an adjustment of results according to the coverage of morbidity information. Also, estimations of the Population Attributable Fraction proposed by the GBD methodology, which were used for a list of specific events, could not be representative for Medellín's population. The prevalence values that we used in this study were related only with those people who visited medical services; therefore, stating that this information can be extrapolated to obtain the actual prevalence of the diseases we examined may be risky.
We recommend developing cohort studies for the local context, which can help in the documentation of the actual attribution (attributable risk estimation) of morbidity and mortality to air pollution due to the environmental risk factors that have been explained by literature. Morbidity studies should be conducted with a focus on disability analysis at a local level, to improve the estimation of the disease burden attributable to environmental pollution, based on data that accurately show those events associated to environmental factors.
I recommend to consider the following issues and address them: The quality of air pollutant data.

○
The under-registration of RIPS. This has been presented previously in several publications (Primer Informe ONS. Aspectos relacionados con la frecuencia de uso de servicios, mortalidad y discapacidad en Colombia, 2011) (Burden of pneumococcal infection in adults in Colombia).

○
The influence of garbage codes to improve the comparability of data with respect to causes of death in the estimates obtained.
○ Use of the GBD 2013 methodology of the WHO and methodology GBD Institute for Health Metrics and Evaluation IHME. The GBD IHME was mentioned in strengths (page 17). ○