The microenvironmental modelling approach to assess children's exposure to air pollution – A review

Abstract

Exposures to a wide spectrum of air pollutants were associated to several effects on children's health. Exposure assessment can be used to establish where and how air pollutants' exposures occur. However, a realistic estimation of children's exposures to air pollution is usually a great ethics challenge, especially for young children, because they cannot intentionally be exposed to contaminants and according to Helsinki declaration, they are not old enough to make a decision on their participation. Additionally, using adult surrogates introduces bias, since time–space–activity patterns are different from those of children. From all the different available approaches for exposure assessment, the microenvironmental (ME) modelling (indirect approach, where personal exposures are estimated or predicted from microenvironment measurements combined with time–activity data) seemed to be the best to assess children's exposure to air pollution as it takes into account the varying levels of pollution to which an individual is exposed during the course of the day, it is faster and less expensive. Thus, this review aimed to explore the use of the ME modelling approach methodology to assess children's exposure to air pollution. To meet this goal, a total of 152 articles, published since 2002, were identified and titles and abstracts were scanned for relevance. After exclusions, 26 articles were fully reviewed and main characteristics were detailed, namely: (i) study design and outcomes, including location, study population, calendar time, pollutants analysed and purpose; and (ii) data collection, including time–activity patterns (methods of collection, record time and key elements) and pollution measurements (microenvironments, methods of collection and duration and time resolution). The reviewed studies were from different parts of the world, confirming the worldwide application, and mostly cross-sectional. Longitudinal studies were also found enhancing the applicability of this approach. The application of this methodology on children is different from that on adults because of data collection, namely the methods used for collecting time–activity patterns must be different and the time–activity patterns are itself different, which leads to select different microenvironments to the data collection of pollutants' concentrations. The most used methods to gather information on time–activity patterns were questionnaires and diaries, and the main microenvironments considered were home and school (indoors and outdoors). Although the ME modelling approach in studies to assess children’s exposure to air pollution is highly encouraged, a validation process is needed, due to the uncertainties associated with the application of this approach.

Introduction

Duan (1982) and Ott (1982) introduced in the early 1980s the concept of human exposure (or simply exposure), which was defined as “an event that occurs when a person comes in contact with the pollutant” (Ott, 1982). Thus, exposure to air pollution occurs whenever a human being breathes air in a location where there are at least trace amounts of airborne pollutants (Klepeis, 2006). Although the first official efforts to control air pollution have traditionally focused on outdoor air, it is now apparent that elevated contaminant concentrations are common inside both private and public buildings (Spengler and Sexton, 1983). Attention should continue to be paid to outdoor air quality and its influence on human health, especially in urban and/or industrialized areas of developed countries. However, people spend up to 90% of their time indoors, making indoor air quality more important than outdoors (Harrison, 1997). Whilst this does not per se mean that indoor exposures will produce more harmful effects, the evidence is that indoor concentrations of many pollutants are often higher than those typically encountered outside (Jones, 1999, Sousa et al., 2012a).

Children are highly vulnerable to air environmental hazards, being considered a risk group (Nieuwenhuijsen et al., 2006, Peled, 2011, Sousa et al., 2009, Sousa et al., 2012b, Sousa et al., 2013) for several reasons including their relative higher amount of air inhalation (the air intake per weight unit in a resting infant is twice than in an adult) and their not fully developed immune system and lungs. As above referred, evidence has been made that children, as well as adults, spend most of their time in indoor environments and are therefore more exposed to indoor air pollution. As a consequence, exposures to a wide spectrum of air pollutants were associated to several effects on children's health, like the increasing of the occurrence of asthma, other allergies and respiratory diseases (Hulin et al., 2010, McGwin et al., 2010, Mendell, 2007, Rumchev et al., 2002, Salvi, 2007, Schwartz, 2004, Sousa et al., 2012a). Evidences of other health outcomes have been found: (i) Brook et al. (2004) and the World Health Organization (WHO, 2006) reported cardiovascular diseases associated with exposure to air pollutants; and (ii) a review from Beamish et al. (2011) suggested that there is a link between air pollution and intestinal disease.

In their daily routine, children move from one location to another and are exposed to a large number of air contaminants for different time durations, raising serious questions about whether such exposures are likely to cause adverse health effects, and what are pollutants' sources. Thus, a complex multifactorial approach for exposure assessment seems appropriate aiming to: (i) associate exposure with health effects; (ii) link health effects with pollution sources; and (iii) determine the exposure value of an individual or group of individuals relative to the population exposure distribution (Moschandreas and Saksena, 2002). In this field, epidemiologic studies provide the opportunity to assess the effects of exposure to air pollution on children's health, i.e., the exposure–response relationship. Multiple outcomes from this type of studies are of interest (Gilliland et al., 2005), including the prevalence of asthma and respiratory diseases, as well as the associated morbidity and mortality. In several countries, as the example of China (Ye et al., 2007), despite the increasing concern about environmental health, most risk-assessment activities are conducted focusing on adults, making environmental health policies inefficient in protecting children's health. Children exposure should be developed to characterize real-life situations, whereby (i) potentially exposed populations are identified; (ii) potential pathways of exposure are identified; and (iii) the magnitude, frequency, duration and time-pattern of contact with a pollutant are quantified (Hubal et al., 2000). Assessing children's exposure to air pollution cannot be merely reduced to the measurements of air pollutants concentrations in one or more environments. In fact, exposure studies can be used to establish where air pollutants exposures occur and the source of those air pollutants (Weisel, 2002).

Hubal et al. (2000) reviewed the factors that strongly influence children's exposure, and concluded that: (i) the physiologic characteristics and behavioural patterns of children result not only in exposure differences between children and adults, but also in differences in exposures among children of different developmental stages; (ii) significant challenges are associated with developing and verifying exposure factors for young children, so it is necessary to develop and improve the methods for monitoring children's exposures and activities; (iii) the data usually available for conducting children's exposure assessments are highly variable, depending on the route of exposure considered, so it requires the collection of physical activity data for children (especially young children) to assess exposure by all routes. Socioeconomic status also greatly influences children's exposure to air pollution (Chaix et al., 2006).

The study of exposure assessment has evolved significantly over the past 30 years (Lioy, 2010) through the appearance of a myriad of methods for assessing personal exposure levels to air pollution. Two different approaches, direct and indirect, described below, have been taken to assess personal exposure to air pollution (Ott, 1982).

There are two available direct methods: (i) personal monitoring, which monitors pollution concentrations using portable equipment worn by the subjects, which can work actively (pumped) or passively (diffusive); and (ii) biomonitoring, which is the use of biomarkers to assess exposure to air pollution, although its usability on exposure studies to air pollution is very specific. Simplicity of design and freedom from modelling assumptions are the advantages of the direct approach (Duan et al., 1991, Wallace and Ott, 1982). Despite direct measurements clearly reflect individual personal exposure levels best, measurements of personal exposures are expensive, time consuming and difficult to apply (Monn, 2001), especially to young children (Jones et al., 2007). It is important to note that a personal measurement does not a priori provide more valid data than a stationary measurement, i.e. a personal sample in a study investigating effects from a specific place or source is often influenced by other sources than those on focus of the investigation, and may thus confound the exposure–effect outcome. Nevertheless, in 1984 EPA performed two large studies of carbon monoxide (CO) exposure in Washington, DC and Denver Colorado, where 1987 persons were followed for 24 h in DC and 1139 persons were followed for two days in Denver. The specific personal monitor used provided exact times in each microenvironment without having to write them down in a questionnaire. This was the first and the most complete study to ever include actual ME measurements, and included many more MEs than in subsequent studies, although being a personal monitoring study (Akland et al., 1985). While biomarkers offer clear advantages, some important criteria must be met when using them for this purpose (Hubal et al., 2000): (i) biomarkers that can accurately quantify the concentration of an environmental contaminant and/or its metabolite(s) in easily accessible biological media (blood, urine, and breath) must be available; (ii) biomarkers must be specific to the contaminant of interest; (iii) the pharmacokinetics of absorption, metabolism, and excretion must be known; and (iv) the time between exposure and biomarkers sample collection must be known. Although there are a number of biomarkers that meet these criteria, few studies using biomarkers have collected all of the information required to accurately estimate exposure. In studies with large sample sizes, long duration and diverse outcomes and exposures, exposure assessment efforts should rely on modelling to provide estimates for the entire cohort, supported by subject-derived questionnaire data, although assessment of some exposures of interest requires individual measurements of exposures using snapshots of personal and microenvironmental exposures over short periods and/or in selected microenvironments (Gilliland et al., 2005). In addition, significant challenges are associated with collecting biomarkers' data from children (Weaver et al., 1998). Although findings from Sexton et al. (2000) indicated that, with proper care, it could be practicable to obtain personal volatile organic compounds (VOC) measurements from elementary school children wearing personal VOC badges samplers, direct methods are unusual on children studies due to their difficult applicability on their time–space–activity specifications. For example, personal monitors for suspended particles (PM) may be particularly impractical for infants or young children due to the requirement of attached pumps (Jones et al., 2007).

Exposure modelling is the indirect method that assesses (estimates or predicts) personal exposures derived from ambient measurements (i.e., measurements made in locations frequented by the study participants) combined with time–activity data, which results in exposure models (MacIntosh and Spengler, 2000, Monn, 2001, Ott, 1982). Some authors reviewed the existing exposure models and tried to classify them, by dividing them into different categories, like Klepeis (2006) and Zou et al. (2009), but the most common classification is into three major groups, as recently reviewed by Milner et al. (2011): (i) Statistical Regression models (not unanimously considered as models), in which linear and nonlinear regression techniques are used to relate personal exposure to its determinants based on measurement data (Kollander, 1991); (ii) Computational Fluid Dynamics (CFD), used to model the spatial and temporal variations in pollutants' concentrations at an extremely fine scale, working on the basic fluid dynamics principles; and (iii) Microenvironmental (ME) modelling, an approach in which weighted average exposure is calculated using time spent and time-averaged concentrations at various places where the population under observation is likely to circulate (Duan, 1981). There are also examples where different models can be complementary (Mölter et al., 2010a, Mölter et al., 2010b), increasing the amount of available data for assessing personal exposure to air pollution, or using both indirect and direct approach to compare the exposure values estimated by the indirect approach with the real personal sampling measured values, which can also be done to validate the model. It is feasible to believe that the indirect methods of exposure assessment can yield estimates closely matching those of the direct method (Malhotra et al., 2000). However, CFD is not considered appropriate for generic population exposure modelling, because it is primarily a research tool used for ventilation, air flow and contaminants' modelling, rather than individual or population exposure modelling. In the same way, and despite being frequently used in epidemiologic studies, regression models have major issues that could be constraints to their applicability, like their transferability to other locations and to other periods of time, when compared to a mechanistic approach like ME modelling (Ashmore and Dimitroulopoulou, 2009). In this field, ME modelling can be used to determine exposures to both individuals and large populations, because it is not often financially practical to make a sufficient number of exposure measurements to completely characterize the spatial and temporal range of exposures in large populations, and to predict what changes in emissions or activities are most effective to obtain reduced exposure (Weisel, 2002). Furthermore, it has several advantages, mainly the possibility to be rapidly and inexpensively used to calculate estimates of exposure over a wide range of exposure scenarios (Klepeis, 1999), and it is also the most appropriate way to examine the potential outcomes of future environmental and/or building interventions and policies, safeguarding the importance to consider indoor exposure modelling (Milner et al., 2011). However, and according to Klepeis (1999), a main disadvantage of this approach compared to the direct approach is the currently research need for its systematic validation, i.e., the results of a fully developed indirect exposure assessment must be compared to an independent set of directly measured exposure levels. The main advantages and limitations of the methods and approaches available to assess children's exposure to air pollution, as well as several examples of studies using them, are summarized in Table 1.

Exposure studies on children are usually a great ethics challenge especially for young children, because they cannot intentionally be exposed to contaminants and according to Helsinki declaration, they are not old enough to make a decision on their participation. Using adult surrogates for these studies introduce bias, because adults do not behave like young children, therefore they cannot mimic their contact activities (Hubal et al., 2000). This is why it is a challenge to develop a realistic estimation of children's exposures to air pollution.

Despite the several available methods within different approaches to assess human exposure to air pollution, the ME exposure modelling method seemed to have several advantages and a great application potential to the assessment of children's exposure to air pollution. With the time children spend in each location (microenvironment) and time-averaged pollutant concentrations, it is possible to estimate and quantify the exposure distribution of study subjects. Additionally, it is viable to examine the likely influence of each location and other exposure factors (Klepeis, 2006). Since children's time–space–activity patterns are different from those of adults, the performance of this modelling approach in estimating personal exposures may differ between these two different types of population (C.-F. Wu et al., 2005). Thus, this review aimed to explore the ME modelling approach methodology to assess children's exposure to air pollution. To meet this goal, this work reviewed studies from the last decade on the assessment of children's exposure to air pollution using this approach, focusing on the methodology, challenges and limitations, to provide a summary of the available scientific findings concerning study design and data collection (time–activity patterns information, microenvironments' selection and pollution measurements), and to some extent look at the outcomes and ME model type.