Malodor as a Trigger of Stress and Negative Mood in Neighbors of Industrial Hog Operations

Study Participants

Eligible participants in the CHEIHO study were nonsmoking adults who lived within 1.5 miles of at least 1 industrial hog operation and were willing to collect data twice daily for approximately 2 weeks. Between September 2003 and September 2005, participants collected data on odor, stress, mood, physical health symptoms, blood pressure, immune function, and lung function; outcomes analyzed in this study are described in more detail in the paragraphs that follow.

At a central location in each neighborhood, research staff set up a monitoring trailer to collect data on hydrogen sulfide (H₂S; MDA Scientific Single Point Monitor, Honeywell Analytics Inc North America, Lincolnshire, IL), particulate matter less than 10 μm in aerodynamic diameter (PM₁₀) and semivolatile PM₁₀ (Tapered Element Oscillating Microbalance Series 1400a Ambient Particulate Monitor with a Series 8500 Filter Dynamics Measurement System, Thermo Fisher Scientific, Waltham, MA), and weather (Vantage Pro Weather Station, Davis Instruments, Hayward, CA, and Young Model 05103VM-42 Wind Monitor, R. M. Young Company, Traverse City, MI).

Selection of the particular pollutants to be monitored was based on previous work that has documented emissions of both H₂S (a product of the anaerobic decomposition of hog waste) and particulate matter from feed, dried feces, skin cells, hair, and bioaerosols, at confinement buildings and waste lagoons.^6,43 Furthermore, we found that H₂S and PM₁₀ were related to reported malodor in the CHEIHO study; H₂S was associated with reported malodor in models that adjusted for the study's longitudinal design, as was PM₁₀ during times when wind speed was greater than 6.75 miles per hour.⁴⁴

The average distance from the monitoring platform to the nearest industrial hog operation in each neighborhood was 0.51 miles; the minimum distance to the nearest industrial hog operation was 0.20 miles and the maximum distance to the nearest industrial hog operation was 1.42 miles. In 2 of the 16 neighborhoods, the platform was located within 2 miles of 1 industrial hog operation; in the other 14 neighborhoods, however, the platform was located within 2 miles of at least 3 industrial hog operations (maximum of 16 industrial hog operations). We therefore calculated, for each neighborhood, the average distance between the platform and the industrial hog operations within 2 miles of the monitoring platform. The average distance across all neighborhoods was 1.10 miles, with a range by neighborhood from 0.56 miles to 1.50 miles. In contrast, the average distance between participant households and the monitoring platform across 15 of the 16 neighborhoods was 0.20 miles, with a range by neighborhood from 0.03 miles to 0.36 miles.

In 1 neighborhood, the average distance between participant households and the platform was 0.95 miles. In this and 3 other neighborhoods where participant homes were more geographically dispersed, we deployed additional H₂S monitors at homes farthest from the platform. All of the data on particulate matter, however, were collected at the platform and assigned to all participants in the neighborhood.

Participants attended a 3-hour training session during which they learned to complete the required data collection activities. They selected a morning time and an evening time at which they would collect data (for example, 6:00 AM and 6:00 PM). In addition, participants completed an assessment of coping style using the John Henryism Active Coping scale^45,46 and an assessment of threshold odor sensitivity using butanol standards.⁴⁷

At the preselected, twice-daily times, participants spent 10 minutes outdoors at home and then returned indoors to rate any odor present during that 10-minute period on a 9-point scale ranging from 0 (no odor) to 8 (very strong odor). Hourly average H₂S, PM₁₀, and semivolatile PM₁₀ values were calculated for the hour immediately preceding the odor rating. Following the odor rating, they responded to 5 mood state questions: “How do you feel now: (a) stressed or annoyed?, (b) nervous or anxious?, (c) gloomy, blue, or unhappy?, (d) angry, grouchy, or bad-tempered?, (e) confused or unable to concentrate?” They rated these mood questions on a 9-point scale ranging from 0 (not at all) to 8 (extremely). The “stressed or annoyed?” question was an ad-hoc single-item measure,^48,49 and the remaining 4 questions came from the Profile of Mood States instrument,^26,50 specifically, from the Tension–Anxiety, Depression–Dejection, Anger–Hostility, and Confusion–Bewilderment subscales. (The Fatigue–Inertia and Vigor–Activity subscales were not used.)

Statistical Analyses

We used logistic mixed models to evaluate malodor, H₂S, PM₁₀, and semivolatile PM₁₀ as predictors of reported stress and negative mood (NLMIXED procedure in SAS version 9.1.3, Cary, NC). We used 2-level (within person and between person) mixed models to take into account the correlated structure of longitudinal data for individuals. The stress and mood variables were recoded as binary; for stressed or annoyed and nervous or anxious, 0 and 1 on the original scale were coded as 0 and 2 to 8 on the original scale were coded as 1. For the remaining 3 mood variables, 0 on the original scale was also coded as 0 and 1 to 8 on the original scale were coded as 1. These coding decisions were based on the distribution of the data such that approximately 90% of the records for each outcome variable were coded as 0 and approximately 10% were coded as 1. We included all predictor variables as linear terms. We conducted all analyses with records for which the ratings of malodor, stress, and mood, and the airborne emissions data, were not missing.

Random intercepts were included in the mixed logistic models to capture the variation between participants in baseline (average) levels of stress and negative mood. Models included the following time-dependent covariates: time of day (morning vs evening), study day (1 to ≥ 14), and study week (first vs second). For analyses of odor as a predictor of stress and mood, the models also included whether participants reported a cold, flu, or stomach virus at any time during data collection (yes or no). We hypothesized that illness could affect a participant's ability to smell or perception of odor and negative mood. We did not consider time-independent confounders, such as age or gender, because their relationship with exposure and outcome did not vary over time. A sample logistic mixed model follows.

Level 1 (time, within person):

where Pr[Stress_ij = 1] is the probability that stress reported by person j at timepoint i equaled 1, b_0j is the person-specific intercept, b_1j is the effect of the time-dependent odor rating, and b_2j is the effect of time of day (morning vs evening).

Level 2 (between person):

where b_0j is the person-specific intercept, γ₀₀ is the mean of the person-specific intercepts (i.e., fixed intercept), γ₀₁(person_j) is the contribution to the overall mean from person j, and μ_0j is the residual between-person variation in the intercept.

We also evaluated several potential modifiers. For analyses of H₂S as a predictor of stress and negative mood, we considered modification by wind speed (low [≤ 0.57 mph], medium [0.58 mph–6.75 mph], and high [> 6.75 mph]) because of previous work that suggested modification of the relationship between H₂S and reported malodor by wind speed.⁴⁴ Based on previous research,^3,29,30,37 we considered age, dichotomized at the median (≤ 53.7 years vs > 53.7 years), and coping style, dichotomized at the median, (John Henryism Active Coping scale score < 52 vs ≥ 52)^46,47 as potential modifiers of any association between reported odor and stress. We also considered threshold odor sensitivity (low or moderate [< 320 ppm] vs high [≥ 320 ppm]) as a potential modifier of the relationships between odor, stress, and mood to evaluate whether more-sensitive individuals responded differently than less-sensitive ones.

Demographics

Table 1 presents demographic information about study participants. The median age was 53.7 years; age ranged from 19.2 years to 89.5 years. Approximately two thirds of the participants were female, and approximately 85% were African American. Seventy-five percent of participants reported that they grew up around livestock. Six neighborhoods were within 2 miles of 1 to 4 industrial hog operations, 4 were within 2 miles of 5 to 9 industrial hog operations, and 6 were within 2 miles of 10 or more industrial hog operations. Average H₂S values in the 16 neighborhoods ranged from less than 0.01 ppb to 1.5 ppb, and the highest measured H₂S values ranged from 2 ppb to 90 ppb. Average PM₁₀ values ranged from 10.8 μg per cubic meter (μg/m³) to 28.7 μg/m³, and average semivolatile PM₁₀ values ranged from −3.2 μg/m³ (negative values occurred because of measurement imprecision at very low concentrations) to 9.2 μg/m³.⁴⁴

The distribution of twice-daily odor ratings during the preselected 10-minute exposure times is presented in Table 2. Of the 2666 odor ratings recorded after participants spent 10 minutes outdoors, approximately 50% equaled zero. An additional 30% were low (a rating of 1 or 2) on the 9-point scale. Approximately 20% were 3 or higher, and 1% of the data were in each of the 2 highest categories. Most of the ratings of stress and mood state also equaled zero. For “stressed or annoyed,” 81% of reports were zero; 87% were zero for “nervous or anxious,” 88% for “gloomy, blue, or unhappy,” 93% for “angry, grouchy, or bad-tempered,” and 95% for “confused or unable to concentrate” (Table 2).

Mixed Models

Table 3 presents parameter estimates, standard errors, t values, odds ratios (ORs), and 95% confidence intervals (CIs) for H₂S, PM₁₀, semivolatile PM₁₀, and reported malodor as predictors of binary stress and negative mood. Variables considered as time-dependent confounders produced little change in the magnitude of the parameter estimates for the independent variables. However, we adjusted all models for time of day (morning vs evening) because time is an important predictor of odor. Reporting stress or annoyance was strongly associated with increasing levels of H₂S; the OR for a 1 ppb change in H₂S was 1.18 (95%CI = 1.08, 1.30). Hydrogen sulfide was also strongly associated with feeling nervous or anxious (OR = 1.12; 95% CI = 1.03, 1.22). Hydrogen sulfide did not appear to be associated with the other 3 mood state variables, and wind speed did not modify any of the relationships between H₂S and stress or mood.

We found that PM₁₀ did not appear to be associated with stress or negative mood, with the exception of a marginal association with feeling confused or unable to concentrate (Table 3). Semivolatile PM₁₀ was most strongly associated with feeling stressed or annoyed and nervous or anxious. Associated ORs for a 1 μg/m³ increase in semivolatile PM₁₀ were small (1.06 and 1.10, respectively), though ORs associated with a 10 μg/m³ increase, for example, were 1.73 and 2.59, respectively. Semivolatile PM₁₀ appeared to be only marginally associated with feeling gloomy, angry, or confused or unable to concentrate.

Table 3 also presents parameter estimates, standard errors, t values, ORs, and 95% CIs for reported malodor as a predictor of binary stress and negative mood. All parameter estimates were large relative to their standard errors. The ratio of the odds of reporting stress for a 1-unit increase in reported odor on a 0-to-8 scale was 1.81 (95% CI = 1.63, 2.00). Consequently, a 4-unit change on the odor scale (from odor = 0 to odor = 4, for example) yielded an OR of 10.6. Odds ratios for feeling nervous, gloomy, angry, and unable to concentrate, associated with a 1-unit change in odor, were 1.60 (95% CI = 1.41, 1.81); 1.43 (95% CI = 1.25, 1.63); 1.52 (95% CI = 1.37, 1.70) and 1.31 (95% CI = 1.16, 1.50), respectively.

Coping, but not age, appeared to modify the relationship between reported odor and stress. The parameter estimate for participants who scored below the median on the John Henryism Active Coping scale was 0.45 (standard error [SE] = 0.07), whereas the parameter estimate for participants who scored above the median was 0.73 (SE = 0.08). Threshold odor sensitivity did not appear to modify the associations between reported malodor and stress or negative mood.

Strengths and Limitations

The longitudinal design was a particular strength of this research. There were approximately 28 repeated measures for each participant. In the analyses, each participant served as her or his own control. Perceptions of stress and adverse mood vary between people, and we were able to statistically model the between-person variation in such perceptions. Physical measures of pollution are an additional strength of this research; previous studies have relied entirely on self-reported measures of exposure and outcome. We did, however, measure only several constituents of a chemically complex odor plume that includes, potentially, hundreds of volatile organic compounds.²³

A further design limitation was the contemporaneous assessment of both exposure and outcome for the analyses of odor as a predictor of stress and negative mood. Because both exposure and outcome were assessed by self-report, it is difficult to determine how the assessment of one affected the assessment of the other. Participants spent 10 minutes outdoors before returning indoors to complete the required data collection activities; they rated the intensity of any malodor present and then rated stress and mood. Rating the odor while stressed or annoyed for reasons unrelated to odor may have induced a higher rating than the participant would have rated in the absence of feeling stressed or annoyed. Though the results of the analyses of odor and stress or mood must be interpreted in light of this design limitation, odor as a marker of exposure is important because it captures information on numerous other pollutants with odorant properties that we were unable to explicitly measure in this study. Furthermore, it permits consideration of the mixture of chemicals emitted from industrial hog operations as opposed to its individual constituent parts.

Conclusions

In a community-based, longitudinal study of neighbors of industrial hog operations, we observed associations among malodor, several airborne emissions, stress, and negative mood. Specifically, we observed increased reporting of stress and negative mood in response to increasing malodor. Additionally, increases in H₂S and semivolatile PM₁₀, both odorous in nature, were associated with reported stress and 1 or more mood variables. Our findings complement a large literature on malodor as an environmental stressor. Malodor and concomitant airborne emissions do appear to trigger stress and negative mood in nearby residents unwillingly exposed at home.

It is important to contextualize the effect of malodor on the lives of nearby residents. People who cannot afford air conditioning, clothes dryers, membership at a gym, and entertaining in restaurants depend on opening their windows for ventilation, drying their clothes outside, exercising in their yards, and entertaining family and friends in and around their homes. In ethnographic interviews, neighbors of industrial hog operations report that they refrain from gardening, walking, chores, and having cook-outs with family and friends because of hog odor, and they report interruption of their sleep because of hog odor inside their homes.⁵⁴ This is significant, because physical activity, social support, and sleep are important for health. Industrial hog operations in North Carolina are located disproportionately in low income, African American communities³⁵ that have limited financial resources to prevent the influx of polluting industries as well as to manage the impacts of uncontrollable environmental malodors on physical and mental health. Recognizing that health is a state of well-being, and not merely the absence of disease,⁴⁰ public health and environmental professionals should consider the impacts of environmental malodor and its potential role in magnifying health disparities.