Chronic exposure to odorous chemicals in residential areas and effects on human psychosocial health: Dose–response relationships

Science of the Total Environment 490 (2014) 545–554

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Chronic exposure to odorous chemicals in residential areas and effects on human psychosocial health: Dose–response relationships Victoria Blanes-Vidal a,⁎, Jesper Bælum b, Esmaeil S. Nadimi c, Per Løfstrøm d, Lars P. Christensen a a

Department of Chemical Engineering, Biotechnology and Environmental Technology, Faculty of Engineering, University of Southern Denmark, Niels Bohrs Alle, 1, DK-5230, Odense M, Denmark Institute of Public Health, Research Unit of General Practice, University of Southern Denmark, J.B. Winsløws Vej 9A, DK-5000 Odense C, Denmark Maersk Mc-Kinney Moller Institute, Faculty of Engineering, University of Southern Denmark, Campusvej 55, DK-5230 Odense M, Denmark d Department of Environmental Science, Aarhus University, Frederiksborgvej 399, DK-4000 Roskilde, Denmark b c

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Odor exposure is a stressor that can affect human health and well-being. • We examined relationships between residential NH3 exposure and psychosocial effects. • Psychosocial effects were annoyance, risk perception and behavioral interference. • We quantified these associations and obtained dose–response models. • We provided insights into the underlying mechanisms that result in these responses.

a r t i c l e

i n f o

Article history: Received 13 January 2014 Received in revised form 28 March 2014 Accepted 13 May 2014 Available online xxxx Editor: Lidia Morawska Keywords: Air pollution exposure Health Dose–response Nuisance Waste Slurry

a b s t r a c t Perceived air pollution, including environmental odor pollution, is known to be an environmental stressor that affects individuals' psychosocial health and well-being. However, very few studies have been able to quantify exposure–response associations based on individual-specific residential exposures to a proxy gas and to examine the mechanisms underlying these associations. In this study, individual-specific exposures in non-urban residential environments during 2005–2010 on a gas released from animal biodegradable wastes (ammonia, NH3) were calculated by the Danish Eulerian long-range transport model and the local-scale transport deposition model. We used binomial and multinomial logistic regression and mediation analyses to examine the associations between average exposures and questionnaire-based data on psychosocial responses, after controlling for person-specific covariates. About 45% of the respondents were annoyed by residential odor pollution. Exposures were associated with annoyance (adjusted odds ratio [ORadj] = 3.54, 95% confidence interval [CI] = 2.33–5.39), health risk perception (ORadj = 4.94; 95% CI = 1.95–12.5) and behavioral interference (ORadj = 3.28; 95% CI = 1.77–6.11), for each unit increase in loge(NH3 exposure). Annoyance was a strong mediator in exposure–behavior interference and exposure–health risk perception relationships (81% and 44% mediation, respectively). Health risk perception did not play a mediating role in exposure–annoyance or exposure–behavioral interference relationships. This is the first study to provide a quantitative estimation of the dose–response associations between ambient NH3

⁎ Corresponding author. Tel.: +45 27781937; fax: +45 65507374. E-mail address: [email protected] (V. Blanes-Vidal).

http://dx.doi.org/10.1016/j.scitotenv.2014.05.041 0048-9697/© 2014 Elsevier B.V. All rights reserved.

546

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

exposures and psychosocial effects caused by odor pollution in non-urban residential outdoor environments. It further shows that these effects are both direct and mediated by other psychosocial responses. The results support the use of NH3 as a proxy gas of air pollution from animal biodegradable wastes in epidemiologic studies. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Perceived air pollution, including environmental odor pollution, has been associated with a series of adverse physical and psychosocial health impacts in humans. Common physical health symptoms among people exposed to odorous chemicals at their residences from e.g. industries or agricultural activities, include eye, nose, and throat irritation, headache, nausea, diarrhea, hoarseness, sore throat, cough, chest tightness, nasal congestion, palpitations, shortness of breath and drowsiness (Schiffman and Williams, 2005). In many cases, conventional toxicological paradigms are not able to explain the association between exposure and symptoms, because reporting of physical effects occurs even when exposures are far below toxicity thresholds. Previous authors have suggested that psychosocial (stress-mediated) responses, such as perceived odor annoyance, may play an important role in physical symptom reporting (Sucker et al., 2009). Environmental odor pollution is indeed an important ambient stressor, since exposure is physically perceptible, negatively valued, unpredictable, uncontrollable, and requires moderate adjustments (Campbell, 1983). Psychosocial effects can be defined as the complex of distress, dysfunction and disability manifested in a wide range of social, psychological and behavioral outcomes, as a consequence of actual or ‘perceived’ environmental contamination (Elliott et al., 1993). Psychosocial effects not only may exacerbate chemical intolerance and symptoms, but also they are considered ‘negative health effects’ per se, in accordance with the World Health Organization's (WHO) definition of health (WHO, 1987). Previous studies have reported annoyance, health risk perception and behavioral interference as important psychosocial responses to environmental odors. Annoyance can be defined as “a feeling of displeasure associated with any agent or condition, known or believed by an individual or group to adversely affect them” (Lindvall and Radford, 1973). Annoyance is considered a form of psychological stress, and it may coincide with other negative emotions, e.g., anger, disappointment, dissatisfaction, helplessness, anxiety, and agitation (Schiffman and Williams, 2005). Risk perception is a subjective judgment by which a person assesses a particular threat or hazard, based on life experiences, information, beliefs, and ideas, that are individual-specific or shared by his/her social group. Odor-related health risk perception can be defined as a concern about the potential consequences that exposure to odorous chemicals in the environment may have on the health. Finally, behavioral interference refers to any disruption of the lifestyle, interference with intended activities and unwanted changes in social behavior. Behavioral interference caused by outdoor odor pollution may lead to significant social impacts, especially in communities where lives are rooted in enjoying the outdoors (Thu, 2002). The relations between odor exposure and psychosocial responses are complex. According to the Transaction Model of Stress and Coping (Lazarus and Folkman, 1984; Cavalini et al., 1991), people's responses to stressors are influenced by individual differences in sensory perception (e.g. olfactory sensitivity), stressor appraisal and coping strategies. Stressor appraisal is people's evaluation of any potential harm or benefit derived from the stressor. When an odor is appraised as harmful or objectionable (primary appraisal), coping strategies are developed, which can be “problem-oriented”, i.e. aimed at minimizing or managing the source of stress; for instance, by closing windows to prevent malodorous air from entering inside the dwelling or not going outside; or “emotion-oriented”, i.e. aimed at reducing or managing the emotional distress caused by the stressor; for instance, by comforting cognition about health effects. When the person has little control over the source and perceives that the only way of managing his/her exposure to the

stressor is by altering intended plans to avoid exposure (behavioral interference), this situation may be a secondary source of annoyance. Therefore, problem-oriented coping itself may result in annoyance, even when it successfully resolves the exposure problem. Based on these coping strategies, people re-appraise their situation (secondary appraisal). To sum up, a person will experience annoyance when he/she appraises the odor as objectionable or harmful (primary appraisal), and, considering their options to cope with the situation (secondary appraisal), find that his/her coping resources are insufficient or detrimental. Therefore, chemical exposures explain part, but not all, of the variation in people's reactions to odors. Individual variations arise partly from differences in how individuals react to the same odor environments, which also depend on the context where exposures occur (Dalton, 1996). Consequently, observational epidemiological studies (as opposed to laboratory trials) provide in principle more realistic quantifications of exposure–response relationships. Given that odor concentrations over regions show high spatial variability (NRC, 2003), one big challenge in resolving the air quality/well-being relationships in observational studies is the assessment of household-specific residential exposures to odorous chemicals. Quantifications of exposure–psychosocial response associations from field studies based on individual exposure assessments are very scarce, since most previous observational studies have been semi-ecological studies, in which responses and important covariates are measured on the individual level, while exposure is assigned on a group level (Oglesby et al., 2000; Claeson et al., 2013). Besides, while some studies have discussed what factors influence the psychosocial responses to odor exposure, it is yet to be determined how these factors exert their combined effects. The objective of our study was to investigate the associations between exposure to a proxy indicator of odor pollution in non-urban residential environments and psychosocial effects (i.e. odor annoyance, health risk perception and behavioral interference) and the mechanisms through which these associations occur; based on individualspecific exposures estimations. In our study, three hypotheses were formulated and tested: 1. The exposure hypothesis was that higher chemical exposures will result in increased likelihood of odor annoyance, health risk perception and behavioral interference. Besides, higher exposures will result in higher level of annoyance. 2. The appraisal hypothesis was that health risk perception is a mediator in exposure–annoyance and exposure–behavioral interference relationships. That is, that appraisal of exposure as harmful (with possible health effects) should account for at least part of the annoyance and behavioral interference experienced by the exposed residents. 3. The coping hypothesis was that annoyance is a mediator in the exposure–behavioral interference relationship. That is, that exposure may result in annoyance perception, which in turn may initiate an active coping strategy involving modifications of individuals' behaviors. The reverse mediation was also considered, that is: given that modifying behavior due to the exposure may result in secondary annoyance, exposure–annoyance responses could be partly mediated by behavioral interference. 2. Materials and methods 2.1. Population data collection A cross-sectional, population-based study was conducted in six 12 km × 12 km non-urban Danish regions. The six regions guaranteed

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

a gradient of exposures to animal waste/farming odors. A total number of 1120 households within the six regions were randomly selected and a structured questionnaire based on a standard questionnaire on indoor climate (Brauer et al., 2000) was mailed from October 2011 to February 2012. The survey was performed, when application of animal/farming wastes to the fields is banned by Danish policies, so that temporary variations caused by different short-term odor exposures were minimized. Adults (N 18 years old) living at the household (1 adult/household) were requested to fill and return the questionnaire. Research was conducted in accordance with the principles of the Declaration of Helsinki. The study was notified to and registered by Datatilsynet (the Danish Data Protection Agency). The questionnaire was anonymous and consisted of two main sections. The first part included general socio-demographic and lifestyle data (i.e. age, gender, smoking habit, job (occupation), time spent at home per week, existence of household residents below 18 years old, and years living in the region), while the second part referred to environmental stressors, i.e. annoyance, health risk perception and behavioral interference experienced during the years 2010 and 2011. Questions regarding odor pollution included: degree of perceived annoyance (estimated using the 5-point verbal annoyance scale, i.e. “0 = not annoyed”, “1 = slightly annoyed”, “2 = moderately annoyed”, “3 = very annoyed” and “4 = extremely annoyed”), duration (i.e., residents estimation of the number of days per year when annoyance was experienced) and origin (i.e., traffic, industry, farm, livestock waste spreading, unknown, or others) (Fields et al., 2001). Concerns about the adverse health impacts of environmental odor perceived at their residences were evaluated using a verbal scale (0 = not concerned; 1 = slightly concerned; 2 = very concerned). Finally, residents were asked whether the existence of environmental odor at their properties prevented them from properly ventilating their homes or from performing outdoor activities that they wished to (0 = no behavioral interference; 1 = behavioral interference). 2.2. Exposure assessment Individual exposure to environmental odorants was estimated by objective air pollution estimations at each individual household. Although odor is the result of a mixture of a large number of gases (Blanes-Vidal et al., 2009), ammonia (NH3) concentration was chosen as a proxy of airborne exposure to odors, as previous studies carried out in the selected areas have shown that farming activities are the main source of offensive odors in the study areas and that NH3 exposure in the residential outdoor environment could be used as a predictor of farming odor (Blanes-Vidal et al., 2012a, 2012b). Ammonia concentrations throughout the six regions were estimated by emission/dispersion modeling, by combining information from two validated models: The Danish Eulerian long-range transport model (DEHM) and the local-scale transport deposition model (OML-DEP) (Geels et al., 2012). The DEHM model was used to estimate the background NH3 concentration (from medium and long-range transport of NH3) at a 16 km × 16 km resolution. The OML-DEP model was used to calculate NH3 dispersion from local point and surface sources in the study areas. The OML-DEP model is a Gaussian dispersion model based on the boundary-layer theory, which also accounts for chemical transformation of NH3 to NH+ 4 . In the OML-DEP, detailed annual emission inventories with an hourly resolution and a spatial resolution of 400 m × 400 m were used. Emission inventories for years 2005–2010 were used for all regions except one, where inventories were only available for 2009. The main significant sources of NH3 in the study areas were considered. These are classified into six main categories: agricultural sources that emit NH3 throughout the year (barns and manure stores), emission from plants, application of manure and mineral fertilizers, grassing (pastures and paddocks), field burning and straw treatment, and emission from traffic and industry. The emission sources

547

considered in the model are described in detail in Gyldenkærne et al. (2005) and Nielsen et al. (2012). Concentrations were calculated in a regular grid of 400 m × 400 m. Finally, ammonia concentrations at each household were estimated by inverse distance weighting (IDW) multivariate interpolation, from UTM coordinates and OML-DEP modeling results. The DEHM and OML-DEP models have been validated in previous studies (Geels et al., 2012; Hertel et al., 2013). Additionally, the Danish National Air Quality Monitoring Programme includes five air pollution monitoring stations located in five of our study areas, which are equipped with semiautomatic filter pack samplers to measure NH3 concentrations in the air on a daily basis. In Blanes-Vidal et al. (in press), annual averaged NH3 concentrations measured at these monitoring stations during the study period were compared with NH3 concentrations estimated by the emission–dispersion model at the locations of the monitoring stations, showing a coefficient of determination (R2) of 0.90. 2.3. Statistical analysis The exposure hypothesis was tested by exposure–response binomial and multinomial logistic models using log-logit sigmoid equations. Odor detection thresholds of individuals within a population are typically lognormally distributed. We assumed that the individual thresholds for odor annoyance, health risk perception and behavioral interference in the population were also lognormally distributed as it has been demonstrated for odor annoyance (Nicell, 2003). Binomial logistic models were used to describe the association between exposure and probability of odor annoyance at any level (“not annoyed”, score = 0; “annoyed”; score N 0), health risk perception (“not concerned”, score = 0; “concerned”; score N 0) and behavioral interference (“no interference”, score = 0; “interference”, score = 1). Multinomial logistic regression models were used to test the significance of the exposure–response associations between NH3 exposure level and probability of experiencing the different annoyance degrees. Multinomial logistic regression can be used in both ordered and not ordered nominal variables and it does not require the proportional odds assumption of ordinal logistic regression (i.e. that the cumulative odds ratio for any two values of the covariates is constant across response categories) to be met. For the purposes of multinomial logistic regression, the response variable annoyance level was categorized into four categories: “no annoyance” (score = 0), “slight annoyance” (score = 1), “moderate annoyance” (score = 2) and “high or extreme annoyance” (score N 2). Unadjusted or crude odd ratios provide a quantification of the exposure–response association when no other factors are taken into account; whereas adjusted odds ratios consider a set of covariates that could affect the exposure–response relationship. Two adjusted models were obtained. In the first adjusted model, the covariates included in the model were four basic demographic features which are usually considered as potential confounders in many epidemiologic studies. In the second adjusted model, the effect of exposure on psychosocial variables was adjusted for three additional potential confounders (i.e. time spent at home per week, existence of household residents below 18 years old, and years living in the region). In a previous publication (Blanes-Vidal et al., 2012a), we discussed the existence of person-related factors that can significantly influence the way each subject perceives and evaluates odors. These characteristics of the exposed individuals can be related with sensory factors (i.e., those that determine the individual's olfactory sensitivity) and nonsensory factors (those that determine the cognitive evaluation of the environment), affecting odor responses. The appraisal and coping hypotheses were tested by mediation analyses according to Baron and Kenny (1986) (Fig. 1). The overall total effect of the independent variable (air pollution exposure, X), on the response (psychosocial variable, Y), path c, is estimated in the first

548

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

Step 1 c

Exposure (X)

Response (Y)

were male and average age was 54 (SD = 14). The non-response analysis showed no significant differences between respondents and nonrespondents regarding gender, age and level of exposure (Table 3).

Covariates (C) 3.2. Prevalence of annoyance, health risk perception and behavioral interference

Step 3

Exposure (X) Step 2

a

Response (Y)

c´

b

Step 3

Mediator (M) Covariates (C) Fig. 1. Path diagram of mediation analysis, where c is the coefficient estimate relating the independent variable and the dependent variable, a is the coefficient estimate relating the independent variable to the mediator, c′ is the coefficient estimate relating the independent variable to the dependent variable adjusted for the mediator, and b is the coefficient estimate relating the mediator to the dependent variable adjusted for the independent variable.

step. In the second step, the effect of independent variable on the mediator (psychosocial variable M), path a, is estimated. In the third step, we estimate path c′, the direct effect of exposure (X) on psychosocial variable (Y), controlling for psychosocial variable M (path b). In the fourth step, the significance of the mediation effect is tested with the Sobel test (Sobel, 1982). Methods of estimating mediation in linear regression do not directly apply when the mediator, the dependent variable, or both, are categorical because the estimated regression coefficients are not comparable across equations (Lockhart et al., 2011). To solve this problem, the coefficients were standardized before applying the Sobel test (MacKinnon and Dwyer, 1993). In mediation analysis, the psychosocial variable not considered in each model as mediator or response was introduced as covariate in all steps (in addition to the personrelated covariates). All statistical analyses were performed in R software. In order to avoid complete or quasicomplete separation, Firth bias-reduction method was used in all logistic analyses (Firth, 1993). A non-response analysis comparing exposure, age and gender of respondents vs. nonrespondents was performed by Chi-squared and two sample t-tests. 3. Results 3.1. Socio-demographic characteristics and air pollution exposure estimates A total of 454 (response rate 40.5%) subjects returned the completed questionnaire. Residential NH3 exposures ranged from 0.14 to 11 μg/m3 (total mean = 2.18 μg/m3, standard deviation = 1.33 μg/m3 and median = 1.91 μg/m3) (Table 1). Fig. 2 shows NH3 concentrations in one of the regions as an example. The socio-demographic characteristics of the respondents are summarized in Table 2. Of 454 respondents, 54%

About 45% of the residents (N = 205) were annoyed by odor pollution at their residences, 151 individuals being “slightly annoyed”, 31 “moderately annoyed”, 15 “very annoyed” and 8 “extremely annoyed” (Table 1). The perceived odor was characterized by all residents as farming/animal waste odor. About 4% of the residents (N = 19) were concerned about the potential negative effects that odor exposure may have on their health, 14 residents being “slightly concerned” and 5 residents “very concerned”. A total of 51 residents (11%) stated that their behavior was affected by the existence of odor in their residential areas. In general terms, the prevalence of psychosocial responses tended to increase as the residents were exposed to higher levels of NH3 concentrations at their households (Table 1). Less than 1% of the residents exposed to NH3 concentrations b2 μg/m3 perceived odor as a potential health risk; while this percentage was 6% and 10%, among residents exposed to 2–3 μg/m3 and N 3 μg/m3, respectively. The percentage of residents that reported that they were not able to aerate their homes or enjoy outdoor activities, due to the existence of malodor outside their homes was 6%, 16% and 20% among residents exposed to b 2 μg/m3, 2–3 μg/m3 and N3 μg/m3, respectively. Furthermore, the degree of reported annoyance generally increased with NH3 exposure. 3.3. Exposure hypothesis 3.3.1. Association between exposure and psychosocial responses The binomial univariate logistic models showed that NH3 concentration at the residence was significantly associated with reported odor annoyance (OR = 3.71; 95% CI = 2.46–5.60), health risk perception (OR = 4.66; 95% CI = 1.95–11.6) and behavioral interference (OR = 3.03; 95% CI = 1.71–5.37) (for each unit increase in loge(NH3 exposure)) (Table 4). The continuous mathematical functions in Fig. 3a show the exposure–response univariate logistic models for the binomial response variable (N = 454). The proportions of participants which expresses the response at eight exposure levels (from Table 1) are also shown for comparison. The results controlled for person-specific variables were consistent with crude OR's, with ORadj = 3.48 (95% CI = 2.29–5.30) for annoyance, ORadj = 5.08 (95% CI = 1.99–13.0) for health risk perception and ORadj = 3.25 (95% CI = 1.75–6.05) for behavioral interference, for each unit increase in loge(NH3 exposure) (Table 4). 3.3.2. Association between exposure and degree of odor annoyance The univariate multinomial exposure–annoyance model (N = 454) characterized by the probability density function (PDF) for the four category response is shown in Fig. 3b. The PDF shows the probability of being slightly annoyed, moderately annoyed and very or extremely

Table 1 Summary of individual NH3 exposures (2005–2010) and prevalence of psychosocial responses (2011) (n = 454). NH3 exposures, μg/m3

n (%)

Level

Mean ± STD

b0.5 0.5–1 1–1.5 1.5–2 2–2.5 2.5–3 3–3.5 N3.5 Total

0.15 0.89 1.27 1.74 2.24 2.69 3.21 5.10

± ± ± ± ± ± ± ±

0.00 0.09 0.15 0.16 0.16 0.14 0.13 1.67

Annoyance, n (%)

Health risk perception, Behavioral interference, n (%) n (%)

Not annoyed Slightly annoyed Moderately annoyed Very annoyed Extremely annoyed 14 (3) 14 (100) 37 (8) 28 (76) 91 (20) 60 (66) 101 (22) 64 (63) 74 (16) 36 (49) 45 (10) 18 (40) 43 (9) 12 (28) 49 (11) 17 (35) 454 (100) 249 (55)

0 (0) 8 (22) 25 (27) 34 (34) 27 (36) 16 (36) 20 (47) 21 (43) 151 (33)

0 (0) 0 (0) 4 (4.4) 2 (2.0) 9 (12) 5 (11) 5 (12) 6 (12) 31 (6.8)

0 (0) 1 (2.7) 2 (2.2) 1 (1.0) 0 (0) 5 (11) 3 (7.0) 3 (6.1) 15 (3.3)

0 (0) 0 (0) 0 (0) 0 (0) 2 (2.7) 1 (2.2) 3 (7.0) 2 (4.1) 8 (1.8)

0 (0) 0 (0) 2 (2.2) 1 (1.0) 5 (6.8) 2 (4.4) 4 (9.3) 5 (10) 19 (4.2)

0 (0) 1 (2.7) 6 (6.6) 7 (6.9) 11 (15) 8 (18) 8 (19) 10 (20) 51 (11)

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

549

Fig. 2. Exposure estimations from the Danish Eulerian long-range transport model (DEHM) and the local-scale transport deposition model (OML-DEP), for one of the regions (Lindet, Denmark), in 2009.

annoyed, for an individual exposed to a certain air pollution at the residence. For a given value of the predictor NH3 exposure, Fig. 3c shows the cumulative distribution function, i.e. the probability of a resident of being “very or extremely annoyed” by a given exposure (lower curve), the probability of being “moderately, highly or extremely annoyed” (middle curve) and the probability of a resident of being annoyed at any level (upper curve). Respondents exposed to NH3 concentrations at their households higher than 2.25 μg/m3 were more likely to be annoyed by odor than not being annoyed. Respondents exposed to NH3 concentrations at their households higher than 10.1 μg/m3 were more likely to be moderately, highly or extremely annoyed, than slightly annoyed or not annoyed (Fig. 3c). Multinomial models controlled for person-related variables showed that the odds for a citizen of being “slightly annoyed” by environmental odors (compared to not being annoyed) significantly increased with NH3 concentration at the residence (ORadj = 2.86; 95% CI = 1.84–4.44), for each unit increase in loge(NH3 exposure). Higher NH3 exposure also increased the odds of being “moderately annoyed” (ORadj = 8.77, 95% CI = 3.53–21.8) and “very or extremely annoyed” (ORadj = 13.4, 95% CI = 4.34–41.2) rather than not being annoyed (Table 5).

3.4. Appraisal and coping hypotheses The Spearman correlation coefficients between the three psychosocial responses showed that the correlation coefficients (p b 0.05) were 0.23 between annoyance and risk perception, 0.39 between annoyance and behavioral interference, and 0.38 between risk perception and

behavioral interference. Table 6 shows the percentage of the total effect of exposure on each psychosocial variable that is attributed to the mediating role of another psychosocial variable, after controlling for the covariates. Mediation tests indicated that health risk perception did not have a significant mediation effect in the exposure–annoyance relationship or in the exposure–behavioral interference relationship. The effect of exposure on behavioral interference was almost completely (81%) mediated by annoyance. The reverse mediating analysis showed that exposure had both a direct effect on annoyance, and an indirect effect mediated by behavioral interference (38%). The exposure–health risk perception relationship was mediated by annoyance (44%), but not by behavioral interference.

4. Discussion The present study provides evidence of the existence of strong dose– response associations between chronic exposure to a proxy chemical for waste odor in non-urban residential environments (NH3) estimated at individual level and three important psychosocial variables: annoyance, health risk perception and behavioral interference. Besides, we investigated the mechanisms of the relationship between odor exposure and each psychosocial variable, in order to explain how exposure is related to the outcomes. The prevalence of reported odor annoyance surpassed the WHO threshold levels (defined as “5% of the population affected at 2% of the time”) (WHO, 1987), as in our study we can infer that at least 18% of the approached population was annoyed by odors about 10% of the

550

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

Table 2 Socio-demographic characteristics of the respondents, n = 454a. n (%) Gender Male Female Ageb b40 years 40–60 years N60 years Smoking No Yes Years living in the householdb ≤25 years N25 years Children at home No Yes Time spent at homeb ≤100 h/week N100 h/week Jobc Not source-related Source-related

245 (54) 209 (46) 80 (18) 210 (46) 164 (36) 389 (86) 65 (14) 218 (48) 236 (52) 343 (76) 111 (24) 184 (41) 270 (59) 412 (91) 42 (9)

a Source population demographics: The total number of adult habitants (N18 years old) in the municipalities where the study areas are located is 247,197 (gender: males = 49.7%; females = 50.3%; age: average age = 50 years; adults age distribution: b40 years = 30.7%; 40–60 years = 38.2%; N60 years = 38.3%; children at home: yes = 29.6%; no = 70%). b Mean ± STD: age: 54 ± 14 years, years living in the area, 30 ± 20, time spent at home, 114 ± 37 h/week. c Source-related jobs are jobs related to animal production, agriculture or farming.

time. This percentage is in agreement with previous studies that have reported that between 13% and 20% of the population in some European countries are annoyed by environmental odors (Hudon et al., 2000; Blanes-Vidal et al., 2012a). About 20% of residents exposed to high NH3 concentrations reported that odor altered their daily activities. The results are in accordance with previous studies that have shown that odors from swine farms and agricultural lands interfere with daily activities and opportunities to socialize with family and friends (Tajik et al., 2008; Lowman et al., 2013). The first main objective of the present study was to test the exposure hypothesis. Long-term household NH3 exposures had a positive and significant association with increased frequency of reported odor annoyance, health risk perception and behavioral interference in non-urban residential communities. These results are in agreement with other investigations (Cavalini et al., 1991; Dalton, 1996; Wing and Wolf, 2000; Radon et al., 2007; Wing et al., 2008; Horton et al., 2009), but some considerations have to be made. First, momentary psychosocial responses from short-term exposures have shown that variations in short-term (1 h) exposures to H2S measured at community level were associated with reporting farming odor annoyance shortly after the exposure, and that average odor over a 12 h period was related to changes in

activities (Wing et al., 2008; Horton et al., 2009). Our study refers though to psychosocial responses to long-term residential exposures, i.e. average exposures over several years, rather than measures concentrations corresponding to the times when residents were annoyed by odors. According to Dalton (1996), responses to environmental odors are highly influenced by past experience and by information that the stimulus activates in long-term memory, and Cavalini et al. (1991) showed that average (long-term) concentrations were better in predicting annoyance, compared to momentary (short-term) concentrations, concluding that annoyance operates on a long term basis. Second, comparison of our results with studies on long-term farming odor impacts is not straightforward because most previous observational studies rely on distance-based exposures, assessed at community level, i.e., using the distance from each neighborhood to e.g. farms and waste outdoor storage units (Wing and Wolf, 2000; Radon et al., 2007), while in our study we performed assessment of individualspecific residential exposures. Personal-related factors had an influence on psychosocial responses. Our study showed that, in general, females, non-smokers, younger residents, new residents, residents spending more time at home, residents having children at home, and residents whose job was not related to the source of odors were positively associated with psychosocial responses. The results are in general agreement with previous studies and a detailed discussion can be found in Blanes-Vidal et al. (2012a). In relation to this, uncontrolled confounding can cause spurious, magnified, or minimized dose–response association results. Comparing the relative change in estimated odds ratios between crude and adjusted logistic regression models is a confounder detection method in statistical and epidemiological applications. Confounder detection methods often compare the relative difference in regression coefficients obtained from “crude” models that exclude the possible confounder(s) and “adjusted” models that include the variable(s), to a cutoff of 10% with differences of at least 10% providing evidence of confounding (Bliss et al., 2012). In our study, the relative changes in OR were N10% in some cases (e.g. 44% change in OR for very or extremely annoyance, in the model adjusted for seven covariates). The second main objective of the study was to get insight about the mechanisms through which exposure affects psychosocial responses, by testing the appraisal and coping hypotheses. Our results did not support the hypothesis that health risk perception mediated the exposure– annoyance or exposure–behavioral interference relationships. That is, annoyance and behavioral interference responses to farming/animal waste odors exposures were not the result from a “worry about its potential health effects”. In contrast, some field studies on perceived air pollution from different sources, such as steel and biofuel industries, have shown that health risk perception may play mediating roles in the exposure–annoyance relationship (Stenlund et al., 2009; Claeson et al., 2013); and in a laboratory study, Dalton (1996) showed that subjects who were led to believe that an odor was a healthy extract showed adaptation; while those told that the odor was potentially hazardous showed apparent sensitization. The fact that risk perception did not

Table 3 Non-response bias analysis of residential NH3 exposures, gender and age based on respondents and non-respondents of three regions (Nrespondents = 280, Nnon-respondents = 338)a. Respondents

Genderb b

Age

Residential NH3 exposurec a b c

Males Females b40 years 40–60 years N60 years

Non respondents

p-Value

Number

%

Number

%

163 117 40 128 112 Mean 2.04

58 42 14 46 40 STD 1.08

190 148 66 141 131 Mean 2.12

56 44 20 42 39 STD 1.16

Data from non-respondents were provided by local authorities (Region Syddanmark) and they cover three study regions. Chi-squared test of proportions. Two sample t-tests.

0.60 0.23

0.39

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

551

Table 4 Associations [OR (95% CI)] between residential exposures (logeNH3 exposures, in μg/m3) and psychosocial variables, from binomial logistic regressionsa,b. OR (95% CI)b

p

loge(NH3 exposure) loge(NH3 exposure) loge(NH3 exposure)

3.71 (2.46–5.60) 4.66 (1.95–11.2) 3.03 (1.71–5.37)

b.001 b.001 b.001

loge(NH3 exposure) Age Job loge(NH3 exposure) loge(NH3 exposure) Job Smoking

3.46 (2.28–5.23) 0.98 (0.97–1.00) 4.97 (2.11–11.7) 4.85 (1.93–12.2) 3.22 (1.74–5.97) 4.66 (0.86–25.3) 3.13 (0.85–11.5)

b.001 .011 b.001 b.001 b.001 .075 .086

3.48 (2.29–5.30) 5.03 (2.12–11.9) 5.08 (1.99–13.0) 3.25 (1.75–6.05) 4.47 (0.83–24.0) 3.06 (0.84–11.2)

b.001 b.001 b.001 b.001 .081 .090

Unadjusted Annoyance Health risk perception Behavioral interference Adjusted for age, gender, job and smoking Annoyance

Health risk perception Behavioral interference

Adjusted for age, gender, job and smoking, children at home, years living in the area and time at home Annoyance loge(NH3 exposure) Job Health risk perception loge(NH3 exposure) Behavioral interference loge(NH3 exposure) Job Smoking a b

Only personal variables with p b 0.10 are shown, i.e. age (in years), job (reference: source-related job) and/or smoking (reference: yes). Number of cases: Annoyance: Ncases = 205; health risk perception: Ncases = 19; behavioral interference: Ncases = 51.

play a mediating role in our model compared to other studies can be explained by two main facts: First, our study was observational, and we did not control or induce any belief about risk on the study population. Second, perceived risk in observational studies is very dependent on the knowledge and attitudes towards a particular exposure and source, and in our study the prevalence of health risk perception was very low (i.e. 4%) compared to health risk perception related to other odor sources generally considered “more noxious” (e.g. Claeson et al., 2013). The reverse mediation analysis showed that the perception of odor as a potential health risk is partially explained by the perception of odor annoyance. The result is in agreement with the traditional notion of unpleasant odor sensation, which is considered to serve as an early warning system for the detection of environmental health hazards (Schiffman and Williams, 2005). Regarding the copying hypothesis, mediating analysis showed that annoyance is a strong mediator in the exposure–behavior interference relationship. Studies on the mediating role of annoyance in exposure– behavioral interference relationships are very scarce, but ethnographic interviews have shown that irregular and unpredictable odor events from animal wastes may create stress and anxiety about daily routines and social events (Tajik et al., 2008). The implications of this result for public health policy and practice include the need of directing efforts not only towards reducing odor emissions but also towards minimizing the negative appraisal of the odor and providing basis to the residents for developing effective coping strategies. In our study, the results suggest that “perception of the stressor as a potential health risk” is not the main dimension of appraisal, but it is the hedonic stimulus quality itself, which is appraised as being unpleasant and in some way inappropriate in the context, what prevents residents to enjoy the outdoors. In this context, efforts to reduce the psychosocial effects of odor exposure could be directed towards, e.g. 1) increasing benefits, i.e. promoting that local residents benefit from the source and perceive it as valuable for themselves or the community; 2) matching expectations, i.e. providing enough information about the exposure times and intensities and assisting in establishing realistic expectations about the environmental quality in the region; 3) improving the perceived control, i.e. providing opportunities to the neighbors to participate and influence decisions regarding management of emissions and exposures; and 4) promoting social justice, i.e. improving the perceived and actual fairness of distributions of exposures among individuals. Finally, reverse mediating analysis showed that behavioral interference was a mediator in the exposure–annoyance relationship. This bidirectional relationship is in accordance to Lazarus and Folkman

Transaction Model of Stress and Coping (Lazarus and Folkman, 1984), who stated that causal relationships between coping strategies and outcomes are likely to be multidirectional rather than linear. Several variables have been reported in the literature to be involved in psychosocial responses to odor exposures, but it is important to note that it is not uncommon for authors to confuse or to not specify whether they refer to meditation or moderation variables. Moderation studies have shown for example, higher annoyance in individuals with a problem-oriented coping style (Steinheider and Winneke, 1993) and that coping, rather than health risk perception, is related to annoyance (Cavalini et al., 1991). Our study has several limitations. First, the fact that mediation analysis intrinsically assumes a sequence of events connected by causal chains, which is valid when hypothesized causes occur earlier in time than hypothesized outcomes. We assumed that the most plausible explanation for the exposure–response association was that odor exposures cause psychosocial responses (and not the reverse effect). This assumption was further based on the fact that the exposure estimated from emission/dispersion modeling occurs earlier in time than the measured response. However, in the case of mediators and outcomes, the mediator cannot be measured temporally before the outcome variable (as appraisal and coping process is dynamic and occurs simultaneously, (Folkman, 1982)). Despite this potential limitation, the results were fully supported by logical/theoretical arguments (Lazarus and Folkman, 1984; Cavalini et al., 1991; Kraemer et al., 2008) and mediation analysis can provide invaluable information for the design of experimental studies of causal processes (Shrout and Bolger, 2002). Another limitation of the study is the relatively low response rate which could increase the risk for non-response bias. Non-response analysis however did not provide an indication of significant bias in terms of exposure, age and gender. Another potential limitation of our study is its temporal resolution. Our study examines the association between residents' responses and annual averaged exposure levels, based on Cavalini et al. (1991) who showed that averaged long-term concentrations were better in predicting annoyance, compared to short-term peak concentrations. However, future studies should investigate the temporal fluctuations in exposures and the association between residents' responses and the magnitude and frequency of short-term peak exposures. This could be done based on the “odor impact criterion”, which is defined by a threshold of exposure and an exceeding probability of this threshold. The threshold of exposure is determined based on the physiological reaction that is expected from a population when the individuals are exposed to a certain gas or odor concentration. The

552

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

0.01

0.10

1.00

10.00

100.00 1000.00

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.01

0.10

NH3 exposure, µg/m3

10.00

100.00 1000.00

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.01

0.10

1.00

10.00

100.00 1000.00

NH3 exposure, µg/m3

NH3 exposure, µg/m3

c

b

1.0

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.01

Probability of annoyance

ecnayonafoytilibaborPn

1.00

Behaviouralinterference (prevalence and probability)

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0

Health risk perception (prevalence and probability)

Annoyance (prevalence and probability)

a

0.10

1.00

10.00

NH3 exposure, Slightly annoyed Highly or extremely annoyed

100.00

1000.00

0.9 0.8

Slightly annoyed

0.7 0.6

Moderately annoyed

Not annoyed

0.5 0.4

Highly or extremely annoyed

0.3 0.2 0.1 0.0 0.01

µg/m3 Moderatly annoyed

0.10

1.00

10.00

100.00

1000.00

NH3 exposure, µg/m3 Annoyed at any level Moderately, highly or extremely annoyed Highly or extremely annoyed

Fig. 3. Exposure–response univariate logistic models of the binomial response variables (a), probability density function of annoyance responses from univariate multinomial logistic regression (b) and cumulative density function (c) (N = 454). Note that although the full log-logit sigmoidal curve is shown (probability from 0 to 1); residents in our observational study were not exposed to the full range of exposures. The maximum residential exposure was 11 μg/m3. Population-level data (dots and stars) on the proportion of residents showing the response at each averaged level of NH3 exposure (based on N = 8 from Table 1) are shown for comparison.

Table 5 Associations [OR (95% CI)] between residential exposures (logeNH3 exposures, in μg/m3) and annoyance degrees from multinomial logistic regressionsa,b,c. Unadjusted Slightly annoyed Moderatly annoyed Very/extremely annoyed Adjusted for age, gender, job and smoking Slightly annoyed

Moderately annoyed Very/extremely annoyed

OR (95% CI)

p

loge(NH3 exposure) loge(NH3 exposure) loge(NH3 exposure)

3.10 (2.01–4.76) 6.32 (2.87–13.9) 8.88 (3.62–21.8)

b.001 b.001 b.001

loge(NH3 exposure) Age Job loge(NH3 exposure) Job loge(NH3 exposure) Gender

2.87 (1.86–4.42) 0.98 (0.96–0.99) 4.43 (1.81–10.9) 6.35 (2.75–14.6) 8.47 (0.99–72.6) 12.3 (4.16–36.3) 2.78 (1.01–7.66)

b.001 .004 .001 b.001 .051 b.001 .048

2.81 (1.84–4.36) 0.98 (0.96–1.00) 4.73 (1.90–11.8) 1.88 (1.04–3.41) 8.47 (3.42–21.0) 0.48 (0.20–1.14) 6.69 (0.78–57.7) 0.97 (0.95–1.00) 1.02 (1.01–1.03) 12.8 (4.23–38.5)

b.001 .077 b.001 .038 b.001 .095 .084 .048 .006 b.001

Adjusted for age, gender, job, smoking, children at home, years living in the area and time at home Slightly annoyed loge(NH3 exposure) Age Job Children Moderately annoyed loge(NH3 exposure) Gender Job Years living area Hours at home Very/extremely annoyed loge(NH3 exposure) a b c

Natural logarithms (base e) were used in all the analyses of this study. Note that using another base would also have been valid (based on the logarithm change of base rule). For each annoyance level, compared to not being annoyed. Only personal variables with p b .10 are shown, i.e. age (in years), job (reference: source-related job), gender (reference: male), and/or children (reference: no).

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

553

Table 6 Mediation effects in exposure (X)-response (Y) models and Sobel mediation testsa,b,c. Mediating effect (M)

Health risk perception Behavioral interference Annoyance Behavioral interference Annoyance Health risk perception

Response (Y)

Annoyance Annoyance Health risk perception Health risk perception Behavioral interference Behavioral interference

Path c

Path a

Path c′

Path b

c

p

a

p

c′

p

b

p

0.390 0.402 0.407 0.377 0.304 0.201

b.001 b.001 0.018 0.022 0.007 0.10

0.407 0.304 0.390 0.201 0.402 0.377

0.018 0.007 b.001 0.11 b.001 0.022

0.357 0.296 0.286 0.315 0.077 0.112

b.001 b.001 0.045 0.045 0.33 0.33

0.345 0.599 0.562 0.436 0.786 0.468

0.040 b.001 0.022 b.001 b.001 b.001

Zmedd

p

%e

1.54 2.11 2.08 1.47 3.05 1.91

0.12 0.034 0.039 0.15 0.002 0.063

– 38 44 – 81 –

a

Covariates: socio-demographic variables and the psychosocial variable not considered as mediator or response. Path c = total effect of X on Y; path a = effect of X on M; path c′ = effect of X on Y, controlling for M; path b = effect of M on T, controlling for X. c c = coefficient estimate relating the independent variable and the dependent variable, c′ = coefficient estimate relating the independent variable to the dependent variable adjusted for the mediator, b = coefficient estimate relating the mediator to the dependent variable adjusted for the independent variable, a = coefficient estimate relating the independent variable to the mediator. d A Zmed N |1.96| indicates a significant mediated effect with α = 0.05. e Percentage of total effect that is mediated. b

exceeding probability is defined as the percentage of time, when the pre-selected exposure threshold is exceeded (Schauberger et al., 2006). Finally, it should be noted that negative health effects may be caused in part by other components present in the mixture, such as endotoxins and other bio-aerosols that do not have odorant properties. Future epidemiological studies are needed to investigate the role of these components in the occurrence of psychosocial health effects. 5. Conclusions In conclusion, this study demonstrates that exposure to outdoor NH3 in residential environments is strongly associated with the occurrence of psychosocial effects, which can affect the social health and quality of life of residents, even at concentrations where traditional toxicity is not expected. To the best of our knowledge, this is the first study to quantify exposure–psychosocial response relationships based on longterm exposure to a proxy odor chemical assessed at individual level in non-urban/agricultural regions. The study provides insights about the potential underlying mechanisms for the associations between exposure and psychosocial responses, in which annoyance and behavioral interference play most significant roles. Apart from the theoretical interest of understanding causal pathways, the present study has practical implications for the development of effective programs able to reduce the impacts of odor exposures on the residents, by identifying and targeting critical variables. More systematic studies are needed to further investigate psychosocial effects of odor exposures and to assess the general validity of the results. Knowledge about these variables would allow integrating them into comprehensive stress models able to explain the relation between exposure to odorous chemicals and physical symptom reporting. Conflict of interest The authors have no conflicts of interest to disclose. References Baron RM, Kenny DA. The moderator–mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51:1173–82. Blanes-Vidal V, Hansen MN, Adamsen APS, Feilberg A, Petersen SO, Jensen BB. Characterization of odor released during handling of swine slurry: part I. Relationship between odorants and perceived odor concentrations. Atmos Environ 2009;43:2997–3005. Blanes-Vidal V, Suh H, Nadimi ES, Løfstrøm P, Ellermann T, Andersen HV, et al. Residential exposure to outdoor air pollution from livestock operations and perceived annoyance among citizens. Environ Int 2012a;40:44–50. Blanes-Vidal V, Nadimi ES, Ellermann T, Andersen HV, Løfstrøm P. Perceived annoyance from environmental odors and association with atmospheric ammonia levels in non-urban residential communities: a cross-sectional study. Environ Health 2012b; 11:27.

Blanes-Vidal V, Bælum J, Schwartz J, Løfstrøm P, Christensen LP. Respiratory and sensory irritation symptoms among residents exposed to low-to-moderate air pollution from biodegradable wastes. J Expo Sci Environ Epidemiol 2014. http://dx.doi.org/10.1038/ jes.2014.20. (in press). Bliss R, Weinberg J, Webster T, Vieira V. Determining the probability distribution and evaluating sensitivity and false positive rate of a confounder detection method applied to logistic regression. J Biometrics Biostat 2012;3:142. Brauer C, Mikkelsen S, Skov PReliability and validity of a new questionnaire for investigation of symptoms related to “The Sick Building Syndrome” and perceived Indoor Air Quality [in Danish, Report, own print]. Glostrup, Denmark: Department of Occupational Medicine, Copenhagen University Hospital; 2000. p. 1–73. Campbell JM. Ambient stressors. Environ Behav 1983;15:355–80. Cavalini PM, Koeter-Kemmerling LG, Pulles MPJ. Coping with odour annoyance and odour concentrations: three field studies. J Environ Psychol 1991;11:123–42. Claeson AS, Liden E, Nordin M, Nordin S. The role of perceived pollution and health risk perception in annoyance and health symptoms: a population-based study of odorous air pollution. Int Arch Occup Environ Health 2013;86:367–74. Dalton P. Odor perception and beliefs about risk. Chem Senses 1996;21:447–58. Elliott SJ, Taylor SM, Walter S, Stieb D, Frank J, Eyles J. Modelling psychosocial effects of exposure to solid waste facilities. Soc Sci Med 1993;37:791–804. Fields JM, De Jong RG, Gjestland T, Flindell IH, Job RFS, Kurra S, et al. Standardized generalpurpose noise reaction questions for community noise surveys: research and a recommendation. J Sound Vib 2001;242:641–79. Firth D. Bias reduction of maximum likelihood estimates. Biometrika 1993;80:27–38. Folkman S. An approach to the measurement of coping. J Occup Behav 1982;3:95–107. Geels C, Ellermann T, Frohn L, Løfstrøm P, Christensen JH, Hertel O, et al. A coupled model system (DAMOS) improves the accuracy of simulated atmospheric ammonia levels over Denmark. Biogeosciences 2012;9:2625–47. Gyldenkærne S, Ambelas Skjøth C, Hertel O, Ellermann T. A dynamical ammonia emission parameterization for use in air pollution models. J Geophys Res 2005; 110:D07108. Hertel O, Geels C, Frohn LM, Ellermann T, Skjøth C, Løfstrøm P, et al. Assessing atmospheric nitrogen deposition to natural and semi-natural ecosystems — experience from Danish studies using the DAMOS. Atmos Environ 2013;66:151–60. [ISSN 1352–2310]. Horton RA, Wing S, Marshall SW, Brownley KA. Malodor as a trigger of stress and negative mood in neighbors of industrial hog operations. Am J Public Health 2009;99:610–5. Hudon G, Guy C, Hermia J. Measurement of odor intensity by an electronic nose. J Air Waste Manage Assoc 2000;50:1750–8. Kraemer HC, Kiernan M, Essex M, Kupfer DJ. How and why criteria defining moderators and mediators differ between the Baron & Kenny and MacArthur approaches. Health Psychol 2008;27:S101–8. Lazarus RS, Folkman S. Psychological stress and the coping process. New York, NY: Springer; 1984. Lindvall T, Radford E. Measurement of annoyance due to exposure to environmental factors. Environ Res 1973;6:1–36. Lockhart G, MacKinnon DP, Ohlrich V. Mediation analysis in psychosomatic medicine research. Psychosom Med 2011;73:29–43. Lowman A, McDonald MA, Wing S, Muhammad N. Land application of treated sewage sludge: community health and environmental justice. Environ Health Perspect 2013;121:537–42. MacKinnon DP, Dwyer JH. Estimating mediated effects in prevention studies. Eval Rev 1993;17:144–58. Nicell JA. Expressions to relate population responses to odor concentration. Atmos Environ 2003;37:4955–64. Nielsen O-K, Winther M, Mikkelsen MH, Hoffmann L, Nielsen M, Gyldenkærne S, et al. Annual Danish Informative Inventory Report to UNECE. Emission inventories from the base year of the protocols to year 2010. Scientific Report from DCE — Danish Centre for Environment and Energy No. 18. Aarhus University, DCE — Danish Centre for Environment and Energy; 2012. [669 pp.]. NRC (National Research Council). Air emissions from animal feeding operations: current knowledge, future needs. Washington, D.C.: National Academy Press; 2003 [286 pp.].

554

V. Blanes-Vidal et al. / Science of the Total Environment 490 (2014) 545–554

Oglesby L, Kunzli N, Monn C, Schindler C, Ackermann-Liebrich U, Leuenberger P. Validity of annoyance scores for estimation of long term air pollution exposure in epidemiologic studies. Am J Epidemiol 2000;152:75–83. Radon K, Schulze A, Ehrenstein V, van Strien RT, Praml G, Nowak D. Environmental exposure to confined animal feeding operations and respiratory health of neighboring residents. Epidemiology 2007;18:300–8. Schauberger G, Piringer M, Petz E. Odour episodes in the vicinity of livestock buildings: a qualitative comparison of odour complaint statistics with model calculations. Agric Ecosyst Environ 2006;114:185–94. Schiffman SS, Williams CM. Science of odor as a potential health issue. J Environ Qual 2005;34:129–38. Shrout PE, Bolger N. Mediation in experimental and nonexperimental studies: new procedures and recommendations. Psychol Methods 2002;7:422–45. Sobel ME. Asymptotic intervals for indirect effects in structural equations models. In: Leinhart S, editor. Sociological methodology. San Francisco: Jossey-Bass; 1982. p. 290–312. Steinheider B, Winneke G. Industrial odours as environmental stressors: exposure– annoyance-associations and their modification by coping, age and perceived health. J Environ Psychol 1993;13:353–63.

Stenlund T, Lidén E, Andersson K, Garvill J, Nordin S. Annoyance and health symptoms and their influencing factors: a population-based air pollution intervention study. Public Health 2009;123:339–45. Sucker K, Both R, Winneke G. Review of adverse health effects of odours in field studies. Water Sci Technol 2009;59:1281–9. Tajik M, Muhammad N, Lowman A, Thu K, Wing S, Grant G. Impact of odor from industrial hog operations on daily living activities. New Solut 2008;18:193–205. Thu K. Public health concerns for neighbors of large-scale swine production. J Agric Saf Health 2002;8:175–84. WHO. Air quality guidelines for Europe. WHO Regional Publication, European Series No. 23. Copenhagen: World Health Organization; 1987. Wing S, Wolf S. Intensive livestock operations, health, and quality of life among eastern North Carolina residents. Environ Health Perspect 2000;108:233–8. Wing S, Horton RA, Marshall SW, Thu K, Tajik M, Schinasi L. Air pollution and odor in communities near industrial swine operations. Environ Health Perspect 2008;116: 1362–8.