From February 2021 to March 2022, we conducted a longitudinal surveillance study of miners employed at 2 rural surface coal mines operated by the same company in Campbell County, Wyoming (300 miners employed, of whom 61 were recruited) and McKinley County, New Mexico (230 miners employed, of whom 169 were recruited; Figure 1). The main reason for refusal to participate was participants’ unwillingness to consent to filling out the questionnaire and/or to provide biological samples. Both surface coal mines had similar engineering, administrative, and personal protective controls in place. Both sites conducted daily SARS-CoV-2 symptom and fever screening and required the use of face masks on-site as part of standard mine safety operations during the COVID-19 pandemic. Workplace policy at both sites precluded miners with COVID-19 symptoms from reporting for duty. The New Mexico site also implemented frequent point-of-care molecular workplace testing on self-administered nasal swabs for workers [15]. Inclusion criteria were as follows: male or female miners currently employed at the abovementioned mines, aged at least 18 years, willing and able to consent to study participation, and willing and able to comply with study procedures. Exclusion criteria were as follows: unable or unwilling to consent and aged <18 years.

At the start of the study, all workers self-reported demographic and exposure information. At the start of each work shift, trained mine safety personnel collected a 24-hour history of SARS-CoV-2 symptoms and the health care professional’s diagnosis of infection. Temperature was measured using a forehead thermometer. At the start of every alternate work shift, staff at the New Mexico site collected a self-administered nasal swab from each participating worker for point-of-care surveillance testing for the SARS-CoV-2 antigen, as described in detail in our previous report [15]. Surveillance on alternating shifts was chosen to balance practical considerations, costs, and test availability. Positive surveillance test results were confirmed by reverse transcription polymerase chain reaction and followed by isolation and health care professional referral, consistent with the US Centers for Disease Control and Prevention guidance [15].

Venipuncture blood samples of 8 mL were collected at baseline and at 3, 6, and 12 months from all participants to measure SARS-CoV-2 serological outcomes. These longitudinal sampling intervals were chosen to characterize and evaluate the natural history of SARS-CoV-2 infections across the study cohorts, given the stability of seroconversion status in 3 to 6 months [16,17]. Venipuncture blood was centrifuged at the mine site, and plasma was aspirated, aliquoted, and frozen by trained nurse technicians. Plasma samples were tested to determine the seroprevalence of COVID-19 using the immunoglobulin G antibody to the nucleocapsid protein of SARS-CoV-2. Tests were performed using a qualitative chemiluminescent immunoassay (ARUP Laboratory) under the Emergency Use Authorization. Unlike antibodies to the spike protein, antibodies to the nucleocapsid protein develop in response to natural infection and not due to COVID-19 vaccination. The assay demonstrated a reported sensitivity of 100% in samples collected 14 days or more after symptom onset and a specificity of 99.6% [18].

Data were transferred manually by the miner coinvestigators to the university-based investigators, who entered the information into a secure REDCap (Vanderbilt University) database. All information entered was double-checked for accuracy.

Descriptive statistics were used to summarize miner characteristics and exposures. Characteristics of miners with low vs high occupational dust exposure were summarized as frequencies and percentages. Frequencies of categorical variables for the 2 groups were compared using χ² tests and Fisher exact tests.

Analyses separately focused on 3 predictor variables (occupational dust exposure, residential wood burning, and shared transportation) and their respective relationships with the outcome. The associations between predictor and outcome variables were evaluated using logistic regression models. We used the Firth penalized likelihood technique to address 0-event cells and reduce small-sample bias during logistic regression modeling [20]. For multivariable models, the least absolute shrinkage and selection operator model selection method [21] was used to improve model parsimony and reduce overfitting. Potential confounders, each coded as dummy variables, included race and ethnicity; mine site (state of employment: New Mexico or Wyoming); COVID-19 vaccination status; and baseline (prevalent) seropositivity, defined as a positive antibody result at enrollment. Although some covariates, such as COVID-19 vaccination status, were not strictly confounders but independent predictors of the outcome, we included them in the multivariable analyses to account for their potential effects. Odds ratios (ORs) and 95% CIs were computed to describe association strength. SAS (version 9.4; SAS Institute Inc) was used for statistical analyses, and a P value of less than .05 was considered statistically significant. With a sample size of 230, our estimated statistical power for a 2-group comparison of seropositivity was at least 80% if the groups differed by at least 21%.

The overall missing rates for predictor variables were all below the cutoff point of 5%. Given the low proportion of missing data, complete case analysis was used [22].

The University of New Mexico Health Sciences Center’s Human Research Protections Office (HRPO 20-680), our institutional review board, approved the study. All participants provided informed consent and could opt out. Participant privacy and confidentiality were protected during the study. The data were deidentified for analysis. Participants were compensated with a US $10 electronic merchandise card for each specimen and survey collected.

A total of 230 miners were studied, including 169 (73.5%) from New Mexico and 61 (26.5%) from Wyoming. At baseline, COVID-19 vaccination was reported by 22.6% (n=52) of participants; 25.2% (n=58) of participants reported high baseline dust exposure, 17.8% (n=41) reported burning wood very often, and 23.5% (n=54) reported sharing transportation very often.

The mean self-reported dust exposure intensity score was 3.49 (SD 2.18), and the median was 3.0 (IQR 2.0-5.0). There were no differences in demographic characteristics between the dust exposure groups, except for a greater frequency of men in the high–dust-exposure group than in the low–dust-exposure group (Table 1). SARS-CoV-2 baseline (prevalent) seropositivity status upon study enrollment was not associated with dust exposure (Table 1).

Lower postbaseline incident seropositivity was associated with greater dust exposure (unadjusted OR 0.39, 95% CI 0.16-0.92; P=.03; Table 2). This association was not significant in the adjusted model. The frequencies of residential wood burning and carpooling were not significantly associated with postbaseline incident seropositivity (Tables 3 and 4, respectively).

In this longitudinal cohort of surface coal miners in New Mexico and Wyoming, self-reported occupational dust exposure was associated with lower odds of incident seroconversion, although this association was not statistically significant after adjustment. Residential wood burning was not associated with incident seroconversion. Shared transportation was also not associated with incident SARS-CoV-2 infection. These findings indicate that environmental and occupational factors may not consistently affect the risk of SARS-CoV-2 infection in this cohort of rural essential workers.

Exposure to PM has been related to new cases and clinical severity of SARS-CoV-2 infections in the published literature, suggesting that PM may be a vector for viral transmission or generate oxidative stress and inflammatory responses associated with viral infection [23-27]. Therefore, we expected a positive association between occupational dust exposure and SARS-CoV-2 infection but unexpectedly found a nonsignificant inverse adjusted association with incident seroconversion in our cohort of miners. This finding appears biologically counterintuitive, as several lines of evidence suggest that particulate exposures may increase susceptibility to COVID-19. For example, a large US cohort study reported that higher historical PM2.5 exposure was significantly associated with increased county-level COVID-19 death rates after adjustment for multiple confounders [28]. Similarly, PM2.5 exposure was a significant risk factor for COVID-19 infection in low- and middle-income Asian countries [29]. Moreover, experimental studies reveal that inhaled PM enhances the expression of angiotensin-converting enzyme 2 and transmembrane protease serine 2, facilitators of SARS-CoV-2 entry into airway epithelial cells [30,31]. Exposure to mineral and silica dust, prevalent in mining environments, has long been known to impair macrophage function and increase susceptibility to respiratory infections [8,32]. The inverse association in our study most likely reflects unmeasured workplace confounding rather than a protective effect of dust. Miners in dustier roles may be more likely to wear respirators (N95 filtering facepiece respirators); to operate haul trucks and other heavy equipment from enclosed, filtered cabs; or to work alone or outdoors; each of these would reduce opportunities for person-to-person SARS-CoV-2 transmission regardless of dust exposure. Because we did not collect data on individual respirator use or personal protective equipment compliance, equipment cab enclosure and filtration, or other work microenvironment characteristics, we could not evaluate these explanations, and our findings do not permit any causal interpretation. Health-related selection into or out of high-dust jobs (the “healthy worker effect”) may also contribute [33]; however, we did not collect data on job tenure, prior job transitions, or health status at job assignment, so this explanation remains speculative and less likely. Differential access to or use of SARS-CoV-2 testing outside the workplace by population subgroups may also have influenced our observed seropositivity patterns and differential ascertainment of infection [34].

We found no association between residential wood burning and incident seroconversion; no incident seroconversions occurred among the 35 participants who reported burning wood often or very often (0/4 and 0/31, respectively). This finding contrasts with population-based studies showing that wood smoke PM2.5 exposure is associated with increased COVID-19 incidence, severity, and mortality [35,36] and with experimental evidence that household air pollution from wood combustion can impair mucociliary clearance and respiratory defense [13,37]. However, the small number of participants in the frequent wood-burning categories limited our ability to detect an association, and our results neither support nor exclude residential wood smoke as a risk factor for SARS-CoV-2.

We found no significant associations between shared transportation and seropositivity. Shared transportation has been implicated in transmission in other occupational settings. Evidence from large urban transit systems shows that closed, crowded transport can facilitate SARS-CoV-2 transmission. For instance, studies from New York City found that neighborhoods with higher subway use saw higher COVID-19 rates and mortality [38,39]. Other studies suggest that poor ventilation in shared vehicles raises aerosol buildup and transmission risk [40]. The lack of a significant association in our study suggests that shared transportation in rural settings, such as consistent passenger groups, open-window ventilation, adherence to mask guidelines, and the possibility of symptomatic or infected miners voluntarily reducing their use of shared transportation, may have reduced the potential for transmission compared with crowded public transport settings.

The strengths of our study include its prospective design, multiple serological tests, assessment of both home and workplace exposures, a study population of rural essential workers, and robust participation by racial and ethnic minority groups. In addition, participation took place during active mining operations, providing real-world applicability during the pandemic. A major limitation is reliance on self-reported, nonvalidated exposure measures (eg, the job-exposure matrix), which are subject to measurement error and recall bias, compared with objective environmental monitoring; this bias is estimated to be nondifferential. Other limitations include potential residual confounding by unmeasured workplace-specific factors (eg, ventilation quality, respirator and mask use and adherence, work in enclosed equipment cabs, indoor vs outdoor work, and shift patterns) and personal factors (eg, socioeconomic status, housing quality, and household density). Additionally, small subgroup sizes, particularly for frequent wood burning and shared transportation, may have limited statistical power and increased the risk of type II error. In particular, no incident seroconversions occurred among participants reporting frequent wood burning, precluding estimation of prospective associations for this exposure. Furthermore, not adjusting for multiple comparisons may increase type I error; this adjustment would have required an ɑ value of .017 (.05/3) for the 3 predictor variables. The associations described in our study findings do not imply causal relationships and should be interpreted cautiously. Declining antinucleocapsid antibody levels over time may also have affected the detection of past infections, with underascertainment of cases contributing to misclassification bias. Studies report a median seroreversion time of approximately 8 months and substantial declines in detectability within the first year after infection [16,17]. Because the study spanned 12 months, seroreversion could differentially affect detection across exposure groups if infection timing varied, introducing potential differential misclassification bias. Given these limitations, the study findings should be considered preliminary, and additional confirmation may be helpful to establish internal and external validity.

Future studies should incorporate objective environmental monitoring (particulate sensors), viral genomic analyses, and larger sample sizes across multiple mining sectors. In addition, further research should examine behavioral factors, household ventilation, and specific workplace practices to understand the pathways of risk better. Exploring how environmental exposures interact with infectious disease dynamics may help develop targeted interventions for essential rural workforces.

In conclusion, increased occupational dust exposure may be associated with lower postbaseline incident seropositivity among miners. These results highlight the need to consider both occupational and household environmental exposures in pandemic response planning and underline the importance of ongoing research to improve respiratory protection strategies for rural mining communities. The implications of our findings extend beyond this study, providing valuable insights for designing interventions to maintain employees’ safety in other essential workplaces during outbreaks of viral respiratory pathogens.