Research | Open Access | Volume 9 (2): Article 82 | Published: 21 May 2026
Menu, Tables and Figures
Table 2: Distribution of CO concentration (median, IQR) according to structural, domestic and occupational factors in Kamilombe
Hervé Mutombu Kabwit1,2,&, Nuncia Kon Muzang1, Ruth Mwad-I-Yav Chikomb2, Clarence Mukeng à Kaut2, Adrien Malandj Kes3, Macel Kasongo Mutwale1, Patrick Kambwandj Kabwit4, Yanick Useni Sikuzani5, Françoise Malonga Kaj3
1Higher Institute of Medical Techniques of Kolwezi, Epidemiology and Public Health Unit, Kolwezi, Democratic Republic of Congo, 2University of Kolwezi, School of Public Health, Epidemiology and Biostatistics Unit, Kolwezi, Democratic Republic of Congo, 3University of Lubumbashi, Faculty of Medicine, Department of Public Health, Lubumbashi, Democratic Republic of Congo, 4National Commission for Protection against Ionizing Radiation (CNPRI), Kolwezi, Democratic Republic of Congo, 5University of Lubumbashi, Faculty of Agronomic Sciences, Lubumbashi, Democratic Republic of Congo
&Corresponding author: Hervé Mutombu Kabwit, Higher Institute of Medical Techniques of Kolwezi, Democratic Republic of Congo, Email: hervemutombu@gmail.com
Received: 13 Jan 2025, Accepted: 13 May 2026, Published: 21 May 2026
Domain: Environmental Health
Keywords: Carbon monoxide (CO), Indoor air pollution, Artisanal mining community, Quantile regression, Environmental exposure, Democratic Republic of Congo
©Hervé Mutombu Kabwit et al. Journal of Interventional Epidemiology and Public Health (ISSN: 2664-2824). This is an Open Access article distributed under the terms of the Creative Commons Attribution International 4.0 License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Cite this article: Hervé Mutombu Kabwit et al., Indoor carbon monoxide concentrations in artisanal mining households in Kolwezi, Democratic Republic of the Congo. Journal of Interventional Epidemiology and Public Health. 2026; 9(2):82. https://doi.org/10.37432/jieph-d-26-00011
Introduction: Domestic carbon monoxide exposure represents an emerging public health concern in artisanal mining sites, where residential precarity and high-risk household practices promote the accumulation of this toxic gas. This study estimated indoor carbon monoxide concentrations and identified determinants of elevated levels in artisanal mining community dwellings.
Methods: We conducted an analytical cross-sectional study in 155 dwellings located in the Kamilombe artisanal mining site. Carbon monoxide concentrations were measured using a pre-calibrated MSA-ALTAIR 5X multi-gas detector. Statistical analyses employed nonparametric Mann-Whitney and Wilcoxon tests and multivariate quantile regression to identify factors associated with different concentration levels at a significance threshold of α = 5%.
Results: The median carbon monoxide concentration was 8 ppm (IQR: 2–16 ppm), with an asymmetric distribution. Most dwellings exhibited structural precarity characteristics. In multivariate quantile regression analysis, tarpaulin construction was associated with significant increases in carbon monoxide concentration across all percentiles, with more pronounced effects at higher concentrations (β = 9.25; 95% CI: 6.47–13.99 at the 75th percentile). Conversely, window presence was associated with significant reductions at the median percentile, more pronounced at higher percentiles (β = -5.75; 95% CI: -10.13 to -2.03; p < 0.001).
Conclusion: Our findings suggest that precarious structural housing conditions may contribute to elevated carbon monoxide concentration levels in artisanal mining communities. Improving housing conditions, enhancing ventilation, and reducing domestic pollution sources could constitute important pathways for reducing carbon monoxide exposure in these settings.
Each year, several million premature deaths are attributable to indoor air pollution, making this phenomenon one of the greatest global environmental health challenges, with a burden disproportionately borne by low- and middle-income countries [1,2]. Among the most insidious domestic air pollutants is carbon monoxide (CO), a colourless and odourless gas whose toxicity stems from a haemoglobin affinity approximately 200 to 250 times greater than that of oxygen, leading to carboxyhemoglobin formation and progressive tissue hypoxia [3-5].
This hypoxia, depending on its intensity and exposure duration, can manifest as headaches, cognitive impairment, chronic fatigue, exacerbation of cardiovascular diseases, and even death in severe cases [6,7]. Chronic domestic CO exposure thus represents a silent health risk, particularly concerning because it occurs within the home itself, a space presumed to be protective.
In resource-limited countries, particularly in sub-Saharan Africa, this risk is substantially amplified by unfavourable environmental and socioeconomic determinants [8,9]. The Democratic Republic of Congo (DRC) paradigmatically illustrates this reality: the near-systematic reliance on solid fuels, charcoal and firewood as the primary domestic energy source, combined with precarious housing conditions, generates conditions conducive to indoor air pollutant accumulation [1,6].
Despite growing recognition of environmental health risks affecting African populations, current literature indicates that empirical epidemiological data on environmental determinants remain insufficient to guide appropriate public health policies [9,10]. In the DRC, this vulnerability is particularly pronounced within artisanal mining communities, where multiple structural risk factors converge.
In Lualaba Province, dwellings in artisanal mining sites are predominantly constructed from makeshift materials: packed earth, wooden planks, salvaged metal sheets, or plastic tarpaulins, and almost systematically lack functional ventilation systems [11,12]. These unfavourable architectural characteristics are compounded by high-risk domestic practices, including smoking in enclosed spaces, cooking on open fires, and solid fuel use in confined spaces [12-16].
These combined conditions create an indoor environment conducive to CO accumulation at potentially dangerous concentrations [1,17-22]. A recent systematic review conducted in African contexts confirms the determining influence of these environmental and structural factors on population health and underscores the urgency of strengthening epidemiological research in these specific settings [10].
However, despite this convergence of risk factors, empirical data on indoor CO concentrations in artisanal mining community dwellings remain remarkably scarce, particularly in Lualaba Province, renowned for its copper and cobalt mineral wealth and where artisanal mining constitutes a livelihood for the majority of the population. This absence of local evidence-based data constitutes a major obstacle to rigorous assessment of actual health risk and development of contextually adapted prevention strategies.
The present study aims to fill this scientific gap by estimating indoor carbon monoxide concentration levels in dwellings of the Kamilombe artisanal mining community and identifying determinants associated with elevated levels, thereby providing solid empirical foundations for public health action.
Study setting
This study was conducted at the Kamilombe artisanal mining site, located in Dilala commune, Kolwezi city, Lualaba Province, Democratic Republic of Congo. The Kamilombe site constitutes one of the region’s principal artisanal cobalt and copper mining centres and is managed by the Mining Cooperative for Social Development (Coopérative Minière pour le Développement Social, CMDS). At the time of the study, this site had approximately 15,000 active artisanal miners distributed across nearly 1,800 active extraction pits.
Miners’ and their families’ dwellings are in immediate proximity to mining exploitation zones, creating a residential environment characterised by high population density and potential exposure to various air pollution sources. This proximity between mining activities and residential spaces promotes exposure to pollutants from both domestic activities and mining-associated combustion, notably carbon monoxide (CO).
Study design
We conducted an analytical cross-sectional study over five months, from March to August 2024. This study design was chosen in accordance with WHO guidelines regarding indoor air quality assessment in dwellings [23] and STROBE recommendations [24] for evaluating environmental exposures in a defined population at a given time point. This duration allowed consideration of variations in domestic practices.
Study population
The target population consisted of all dwellings located at the Kamilombe artisanal mining site. The source population was defined from a census conducted in February 2024, which identified 647 occupied housing units at the site. A housing unit was defined as any identifiable residential space, including dwellings, mixed-use kiosks, and storage facilities. The study included permanently occupied dwellings, dwellings located within mining site boundaries, dwellings whose residents provided informed consent, and dwellings accessible for measurements.
Sample size
Sample size was determined using the Epitools tool [25] according to the following formula: n ≥ (Zα/2×σ/d)2, where: Zα/2 is the confidence coefficient in the normal distribution or the Z1-α/2 value for a two-tailed test, in our case 1.96 ; σ is the assumed population standard deviation = 8 ppm, derived from the multicenter study conducted in an African context by Bruce et al., similar to our context [6] ; d corresponds to estimation precision; we desired 1.5 ppm precision, corresponding to 20.0% of the theoretical mean of 12 ppm in accordance with WHO recommendations [1]. Application of this formula yields n ≥ (1.96 × 8 / 1.5)2 = 110 dwellings. To improve statistical power and account for field constraints, we increased our sample to 155 dwellings.
Sampling
Systematic random sampling was applied to the list of 647 dwellings. The sampling interval was set at 4 according to the formula k = N/n where N = 647 and n = 155, thus k = 647/155 = 4.17, approximately 4. The starting unit was randomly selected, then every fourth dwelling was included until reaching a sample of 155 units.
Data collection
Data collection was performed by six investigators trained for three days on the study protocol, CO detector use, and ethical principles. Two senior researchers provided supervision and quality control. Two instruments were used: a pre-tested structured questionnaire and an MSA ALTAIR 5X portable multi-gas detector equipped with an electrochemical sensor [26]. The device has a measurement range of 0 to 1500 ppm with ±5% accuracy. The questionnaire addressed structural, behavioural, and occupational characteristics.
For each selected housing unit, the following procedures were applied:
Measurement strategy and justification
The measurement strategy relied on repeated point measurements, performed three times daily for five consecutive days, totalling 15 measurements per dwelling. This approach was adopted to reconcile operational feasibility in an artisanal mining context with scientific rigour. It allowed capture of intra- and inter-daily variability in CO concentrations related to domestic activities and ventilation conditions.
Methodological limitations and interpretation:
It is important to emphasise that point measurements, even repeated, do not allow precise estimation of 8-hour time-weighted averages nor direct assessment of compliance with WHO guideline values for CO (approximately 8.6 to 9 ppm for 8-hour exposure) [1]. CO concentrations measured in this study should therefore be interpreted as indicators of domestic CO exposure risk, rather than as direct estimates of cumulative exposure.
This approach is consistent with methodological recommendations for environmental exposure assessment studies in resource-limited contexts [27-33], where the primary objective is identification of exposure determinants and comparison of exposure levels between different groups rather than absolute assessment of regulatory compliance.
Variables
Dependent variable: The primary dependent variable was indoor air carbon monoxide (CO) concentration in dwellings, measured in parts per million (ppm). This variable was treated as a continuous quantitative variable.
Independent variables: Independent variables were grouped into two main categories:
Architectural and environmental dwelling characteristics:
Household sociodemographic and behavioural characteristics:
Statistical analysis
Sample characteristics and CO concentrations were described using appropriate statistics according to variable nature and distribution: quantitative variables were presented as median and interquartile range. Categorical variables were expressed as frequencies and percentages with 95% CI. Bivariate analyses were performed to examine associations between each independent variable and median CO concentration. The non-parametric Mann-Whitney and Wilcoxon tests were used to compare CO concentration distributions between different groups. The statistical significance threshold was set at α = 0.05 (two-tailed). Variables presenting p-value < 0.20 in bivariate analyses were considered candidates for inclusion in the multivariate regression model, in accordance with methodological recommendations for variable selection in quantile regression.
A quantile regression model was developed to identify independent determinants of CO concentrations while accounting for its asymmetric distribution. Unlike classical linear regression, which models the conditional mean, quantile regression allowed us to model different quantiles of the CO conditional distribution (25th, 50th, and 75th percentiles), thus offering a more complete description of the relationship between different independent variables and the dependent variable (CO), particularly in this heteroscedastic situation.
The Variance Inflation Factor (VIF) was calculated for each independent variable to detect the presence of multicollinearity. VIF > 5 was considered an indicator of problematic multicollinearity requiring the removal of the concerned variable or its re-coding. All statistical analyses were performed using R software version 4.5.1 [33]. Quantile regression was performed using the quantreg package, while multicollinearity assessment was performed using the car package.
Ethical considerations
This study protocol was approved by the Ethics Committee of the School of Public Health, University of Kolwezi. Administrative authorisations were also obtained from the Mining Cooperative for Social Development (CMDS) and the Small-Scale Mining Assistance and Supervision Service (SAEMAPE), competent authorities of the study site. Written informed consent was obtained from all participants before inclusion. The form, available in French and Swahili, explained study objectives, procedures, risks, and the voluntary nature of participation. For illiterate participants, consent was obtained by fingerprint after document reading in the presence of an independent witness. Data confidentiality was guaranteed through participant anonymization using unique numerical codes. Nominative data were stored separately in a secure file, accessible only to the principal investigator, and data access was limited to authorised research team members. No information allowing participant identification was disclosed.
Descriptive characteristics of dwellings and households
The median carbon monoxide concentration was 8 ppm (IQR: 2–16), with a mean of 10.14 ± 8.72 ppm, ranging from 1 to 33 ppm. The distribution exhibited positive skewness (skewness = 0.84). Most dwellings had two rooms or fewer (84.5%). A substantial proportion of dwellings were constructed with precarious materials, notably tarpaulin (39.4%) and brick (42.6%, Table 1). Furthermore, 45.8% of dwellings lacked a functional window.
Regarding domestic behaviours, cooking was performed indoors in 45.8% of dwellings, while 60% of households reported presence of at least one regular smoker. Occupationally, 34.8% of respondents were artisanal miners. Additionally, 20.6% of dwellings stored minerals indoors, and 57.4% reported recent health problems among residents, notably cough (35.5%) and headaches (21.9%).
CO distribution according to dwelling characteristics
Bivariate analysis of carbon monoxide (CO) concentration showed significant variations according to several structural dwelling and resident characteristics. Median CO concentration was significantly higher in tarpaulin-constructed dwellings (17 ppm; IQR: 10–23; p < 0.001). Conversely, metal sheet and brick dwellings exhibited significantly lower median concentrations, 1 ppm (IQR: 1–8; p < 0.001) and 6 ppm (IQR: 3–7.8; p < 0.001), respectively. Functional window presence was associated with significantly lower median CO concentration (4.5 ppm; IQR: 1–8; p < 0.001). Similarly, indoor cooking was associated with significant CO concentration variation (7 ppm; IQR: 1–13; p = 0.021).
Dwellings whose residents’ primary occupation was mineral excavation exhibited significantly higher median CO concentrations (13 ppm; IQR: 6–21.5; p < 0.001), while those whose residents’ primary occupation was other activities exhibited lower concentrations (6 ppm; IQR: 1–7; p < 0.001). Presence of at least one smoker in the dwelling was also associated with significantly higher median CO concentration (13 ppm; IQR: 6–19; p < 0.001, Table 2).
Furthermore, dwellings whose residents reported at least one health problem during the week preceding the survey exhibited significantly higher carbon monoxide concentrations (11 ppm; IQR: 6–19; p < 0.001). Multicollinearity assessment revealed no significant correlation between explanatory variables, with Variance Inflation Factor (VIF) values ranging between 1.20 and 1.82 (Table 3).
Factors associated with elevated carbon monoxide concentration in dwellings
In multivariate quantile regression analysis, factors associated with carbon monoxide concentration varied according to CO distribution levels. At the 25th percentile, tarpaulin housing was associated with a significant increase in CO concentration (β = 5.00; 95% CI: 3.11–8.01; p = 0.0006). Presence of at least one smoker in the dwelling was also associated with a significant CO increase (β = 2.67; 95% CI: 1.52–5.18; p = 0.038). Conversely, window presence was not significantly associated with CO concentration at this percentile (β = -1.67; 95% CI: -3.82 to 0.61; p = 0.14, Table 4).
At the median (50th percentile), tarpaulin housing remained significantly associated with CO concentration increase (β = 6.60; 95% CI: 4.65–11.39; p < 0.001). Smoker presence in the dwelling was also associated with significant CO increase (β = 5.60; 95% CI: 3.57–7.13; p < 0.001). Conversely, window presence was associated with significant CO concentration decrease (β = -4.00; 95% CI: -6.34 to -1.31; p < 0.001). Similarly, higher number of dwelling rooms was associated with significant CO concentration decrease (β = -1.80; 95% CI: -3.08 to -0.16; p = 0.002). Furthermore, households whose primary occupation was other activities also exhibited significantly lower CO concentration (β = -2.20; 95% CI: -4.60 to -0.93; p = 0.016).
At the 75th percentile, tarpaulin housing was strongly associated with CO concentration increase (β = 9.25; 95% CI: 6.47–13.99; p < 0.001). Similarly, smoker presence in the dwelling was associated with significant CO increase (β = 6.00; 95% CI: 2.90–8.53; p < 0.001). Conversely, window presence (β = -5.75; 95% CI: -10.13 to -2.03; p < 0.001) and higher number of rooms (β = -3.75; 95% CI: -5.24 to -0.39; p = 0.016) were associated with significant decreases in elevated CO concentrations.
This analytical cross-sectional study conducted in 155 dwellings of the Kamilombe artisanal mining community reveals that indoor carbon monoxide (CO) exposure warrants attention. Indeed, a median of 8 ppm (IQR: 2–16 ppm) and a mean of 10.14 ± 8.72 ppm were noted. The positive asymmetric distribution (skewness = 0.84) and value range (1–33 ppm) demonstrate marked heterogeneity in exposure conditions within this vulnerable population.
Multivariate quantile regression analysis identified structural determinants whose effects vary substantially according to exposure level: tarpaulin dwellings exhibit increasing effects at higher percentiles (β = 5.00 at P25; β = 6.60 at P50; β = 9.25 at P75); while functional window presence exerts a protective effect that also amplifies at elevated percentiles (β = -1.67 at P25, non-significant; β = -4.00 at P50; β = -5.75 at P75). Domestic smoking and the number of rooms also influence CO concentrations differentially according to quantiles. These results underscore the importance of a methodological approach sensitive to the complete exposure distribution, particularly in environmental precarity contexts where intra-community inequalities are pronounced.
In our study, the observed median of 8 ppm is slightly above the 7 ppm reference threshold for 24-hour exposure mentioned in several African studies [34,35], although it remains markedly lower than mean concentrations reported in charcoal cooking environments in Uganda (41.52 ppm) [36] or Kenya (15.81 ppm) [37]. This intermediate position suggests that Kamilombe dwellings, although exposed to domestic combustion sources and smoking, do not systematically present the extreme conditions observed in closed kitchens using exclusively solid fuels.
However, interpretation of this median must be nuanced by the strong value dispersion: nearly 25% of dwellings exceed 16 ppm, and some reach 33 ppm, levels compatible with increased health risks, notably respiratory symptoms, headaches, and chronic cardiovascular vulnerability [37,38].
In our study, we employed quantile regression, which constitutes a major methodological contribution of this study, allowing us to overcome limitations of classical linear models that capture only the average effect of predictors. By analyzing associations at the 25th, 50th, and 75th percentiles, we demonstrated a differential exposure gradient according to structural determinants, a phenomenon masked by mean-centred approaches. The effect of tarpaulin construction material particularly illustrates this phenomenon: its coefficient increases from 5.00 ppm at P25 to 9.25 ppm at P75, an 85% increase. This amplification suggests that tarpaulin dwellings, already disadvantaged in terms of ventilation and insulation, accumulate additional risk factors (overcrowding, proximity to combustion sources, chimney absence) that manifest disproportionately in the most polluted environments.
This result converges with observations by Mulat et al. in Ethiopia, who documented associations between precarious construction materials and elevated indoor pollutant concentrations [34], and with those of Pollard et al., who emphasise the interaction between housing structural characteristics and domestic practices in generating extreme exposures [39].
We noted in this study that the protective effect of functional window presence exhibited an inverse but equally instructive profile: non-significant at P25 (β = -1.67), it becomes substantial at P50 (β = -4.00) and P75 (β = -5.75). This progression indicates that ventilation exerts its beneficial effect primarily in dwellings where CO sources are substantial, consistent with WHO recommendations identifying ventilation as a key lever for indoor pollution prevention [23]. In low-exposure dwellings (P25), absence of major CO sources renders ventilation effect marginal, while in heavily polluted environments (P75), adequate ventilation can reduce concentrations by nearly 6 ppm, a relative decrease of approximately 30 to 40%.
Regarding domestic smoking, in our study, it exhibited a relatively stable effect between P25 (β = 2.67) and P75 (β = 6.00), with slight amplification at higher percentiles. This relative stability suggests that smoking acts as an additive rather than a multiplicative risk factor, contributing consistently to exposure regardless of background pollution level. With 60% of dwellings having at least one smoker, this determinant represents an accessible public health intervention target, although its modification requires behavioural approaches complementary to structural interventions.
Regarding the effect of the number of rooms, it was found to be significant only at median and upper percentiles (β = -1.80 at P50; β = -3.75 at P75), likely reflecting a volumetric dilution effect: in overcrowded or single-room dwellings, CO sources (cooking, smoking) directly affect living space, while functional room separation allows exposure reduction in rest areas. This result echoes observations by Kansiime et al. in Uganda, who documented the importance of dwelling spatial configuration in modulating biomass fuel exposure [40].
The absence of multicollinearity (VIF between 1.20 and 1.82) (see Table 3) strengthens the validity of our estimates and suggests that determinants identified in our study operate through relatively independent mechanisms, allowing targeted interventions on each factor.
Compared to international literature, CO concentrations observed at Kamilombe fall within the low to intermediate range of values reported in sub-Saharan Africa. Nakora et al. measured means of 41.52 ppm in charcoal kitchens in Uganda, four times our study median [36], while Shupler et al., in the multi-country CLEAN-Air (Africa) study, report means of 15.81 ppm in Kenya and 1.77 ppm in Ghana, illustrating extreme exposure variability according to fuels and cooking practices [35]. Mulat et al. documented a median of 7.9 mg/m³ (approximately 7 ppm) in Jimma, Ethiopia, under precarious housing conditions [34] comparable to those at Kamilombe.
Methodologically, our study is distinguished by quantile regression use, an approach rarely employed in African literature on indoor pollution. Identified studies [34-36,41] primarily used linear or logistic regressions centred on the mean, approaches that do not capture differential effects at distribution extremes. Chronic health effects documented in the literature reinforce the relevance of our results. Mocumbi et al. synthesized evidence of cardiovascular effects of indoor pollution in sub-Saharan Africa [38], while González-Pedraza et al. documented associations between CO/PM2.5 exposure and respiratory symptoms in rural women [37]. Wernecke et al. recently confirmed links between solid fuels and respiratory health in South Africa [41].
Overall, this study’s results call for an integrated intervention strategy differentiated according to three priority axes: First, housing condition improvement, notably through tarpaulin dwelling replacement and functional window installation promoting ventilation, constitutes a major structural lever. The amplifying effect of tarpaulin at higher percentiles (β = 9.25 at P75) indicates that interventions targeting the most precarious dwellings would generate the most substantial health gains, consistent with WHO recommendations [23].
Second, the energy transition toward clean fuels must be promoted, consistent with recommendations by Shupler et al. and Kansiime [35,39]. Third, behavioural interventions targeting domestic smoking reduction are necessary, given high prevalence (60% of dwellings) and smoking’s constant additive effect (β = 2.67 to 6.00 according to percentiles). These interventions should be co-designed with communities, consistent with Pollard et al.’s recommendations on the importance of participatory approaches [38]. The specific vulnerability of artisanal mining communities requires adapted approaches, integrated into broader community development strategies including land tenure regularization, clean energy access, and local health system strengthening [37].
Strengths and limitations
In terms of strengths, quantile regression use, model stability (VIF between 1.20 and 1.82), and direct CO measurement by calibrated instruments constitute the principal methodological strengths of this study. However, the cross-sectional design, which does not allow establishment of causal relationships, and residual confounding factors (cooking practices, occupation duration, measurement season, etc.) may influence observed associations. Furthermore, point CO measurement does not capture temporal exposure variations, and the absence of detailed data on fuel types and the single site (Kamilombe) limits inter-site comparison and external validity of results.
Perspectives
Multi-site longitudinal studies, integrating repeated environmental measurements and health indicators, are necessary to confirm these results and support the development of environmental health policies adapted to artisanal mining communities in Lualaba Province and the DRC generally.
This study highlights substantial heterogeneity in indoor carbon monoxide concentration in dwellings of the Kamilombe artisanal mining community, with more pronounced associations at elevated exposure levels. Tarpaulin dwellings, insufficient ventilation, and residential overcrowding were more frequently observed in upper percentiles of CO distribution, suggesting significant associations. These results underscore the importance of targeted strategies addressing housing condition improvement, ventilation, and domestic pollution source reduction, particularly in dwellings exhibiting the highest exposure levels.
What is already known about the topic
Indoor air pollution by carbon monoxide is strongly associated with:
What this study adds
The research protocol was designed by HMK, who also handled data collection, processing, and analysis, as well as manuscript writing and revision. NKM, RMC, CMK, AMK, MKM, PKK contributed to data collection, processing, and review, as well as manuscript design and revision. YUS, FMK contributed to manuscript revision. The final manuscript was read and approved by all authors.
| Variable | N (%) | 95%CI |
|---|---|---|
| Distance between dwelling and excavation gallery > 1000 | ||
| Yes | 37 (23.9) | 17.8 – 31.2 |
| No | 118 (76.1) | 68.8 – 82.2 |
| Tarpaulin construction | ||
| Yes | 61 (39.4) | 32 – 47.2 |
| No | 94 (60.6) | 52.8 – 68 |
| Sheet metal construction | ||
| Yes | 28 (18.1) | 12.8 – 24.9 |
| No | 127 (81.9) | 75.1 – 87.2 |
| Brick construction | ||
| Yes | 66 (42.6) | 35.1 – 50.5 |
| No | 89 (57.4) | 49.5 – 64.9 |
| Dwelling with ≤ 2 rooms | ||
| Yes | 131 (84.5) | 78 – 89.4 |
| No | 24 (15.5) | 10.6 – 22 |
| Dwelling used as accommodation | ||
| Yes | 79 (51) | 43.2 – 58.7 |
| No | 76 (49) | 41.3 – 56.8 |
| Dwelling used as a kiosk | ||
| Yes | 40 (25.8) | 19.6 – 33.2 |
| No | 115 (74.2) | 66.8 – 80.4 |
| Dwelling used as a storage room | ||
| Yes | 16 (10.3) | 6.5 – 16.1 |
| No | 139 (89.7) | 83.9 – 93.5 |
| Dwelling with more than 2 occupants | ||
| Yes | 78 (50.3) | 42.5 – 58.1 |
| No | 77 (49.7) | 41.9 – 57.5 |
| Dwelling with ventilation system | ||
| Yes | 84 (54.2) | 46.3 – 61.8 |
| No | 71 (45.8) | 38.2 – 53.7 |
| A dwelling where cooking is done inside | ||
| Yes | 71 (45.8) | 38.2 – 53.7 |
| No | 84 (54.2) | 46.3 – 61.8 |
| Dwelling for which the respondent has their main occupation being the mining of ores | ||
| Yes | 54 (34.8) | 27.8 – 42.6 |
| No | 101 (65.2) | 57.4 – 72.2 |
| Dwelling whose main occupation is small-scale commerce | ||
| Yes | 39 (25.2) | 19 – 32.5 |
| No | 116 (74.8) | 67.5 – 81 |
| Dwellings where the respondent’s main occupation is other activities unrelated to artisanal mining | ||
| Yes | 62 (40) | 32.6 – 47.9 |
| No | 93 (60) | 52.1 – 67.4 |
| Dwelling with ores stored inside | ||
| Yes | 32 (20.6) | 15 – 27.7 |
| No | 123 (79.4) | 72.3 – 85 |
| Dwelling with at least one smoker present | ||
| Yes | 93 (60) | 52.1 – 67.4 |
| No | 62 (40) | 32.6 – 47.9 |
| Housing with recent health problems among the occupants | ||
| Yes | 89 (57.4) | 49.5 – 64.9 |
| No | 66 (42.6) | 35.1 – 50.5 |
| A dwelling where the occupants reported headaches as a health problem | ||
| Yes | 34 (21.9) | 16.1 – 29.1 |
| No | 55 (35.5) | 28.4 – 43.3 |
| A dwelling where the occupants reported coughing as a health problem. | ||
| Yes | 55 (35.5) | 28.4 – 43.3 |
| No | 34 (21.9) | 16.1 – 29.1 |
| Features | Modality | N | Median CO (Q1-Q3) | p_value |
|---|---|---|---|---|
| Distance between dwelling and excavation gallery ≤ 1000 | Yes | 118 | 9 (3-17) | 0.080 |
| Sheet metal dwelling | Yes | 28 | 1 (1-8) | <0.001 |
| Brick house | Yes | 66 | 6 (3-7.8) | <0.001 |
| Tarpaulin dwelling | Yes | 61 | 17 (10-23) | <0.001 |
| Number of rooms in the dwelling ≤ 2 | Yes | 131 | 8 (4-16) | 0.053 |
| Presence of a ventilation system | Yes | 84 | 4.5 (1-8) | <0.001 |
| Cooking inside the home | Yes | 71 | 7 (1-13) | 0.021 |
| Number of occupants > 2 | Yes | 78 | 8 (3-16) | 0.732 |
| Dwelling type of accommodation | Yes | 79 | 6 (1.5-18.5) | 0.704 |
| Kiosk-type dwelling | Yes | 40 | 8 (4-12) | 0.893 |
| Storage-type dwelling | Yes | 16 | 8 (4.8-12) | 0.700 |
| Primary occupation: mining | Yes | 54 | 13 (6-21.5) | <0.001 |
| Main occupation of small business | Yes | 39 | 10 (3.5-16.5) | 0.506 |
| Main occupation other | Yes | 62 | 6 (1-7) | <0.001 |
| Dwelling with ore storage | Yes | 32 | 9.5 (4.5-19) | 0.416 |
| Presence of at least one smoker | Yes | 93 | 13 (6-19) | <0.001 |
| Health problem in the last week preceding the investigation | Yes | 89 | 11 (6-19) | <0.001 |
| Headache-type health problem | Yes | 34 | 13 (8.2-21.5) | 0.056 |
| Health problem such as a cough | Yes | 55 | 9 (2.5-16.5) | 0.056 |
Note: 1 Wilcoxon and Mann-Whitney test
| Features | VIF | Interpretation |
|---|---|---|
| Distance between dwelling and excavation gallery ≤ 1000 | 1.40 | Weak |
| Sheet metal dwelling | 1.43 | Weak |
| Tarpaulin dwelling | 1.39 | Weak |
| Number of rooms in the dwelling ≤ 2 | 1.39 | Weak |
| Presence of a ventilation system | 1.28 | Weak |
| Cooking inside the home | 1.20 | Weak |
| Primary occupation: mining | 1.71 | Weak |
| Main occupation other | 1.82 | Weak |
| Presence of at least one smoker | 1.39 | Weak |
| Health problem in the last week preceding the investigation | 1.21 | Weak |
Note: VIF = Variance Inflation Factor. A value greater than 5 indicates potential multicollinearity.
| Variable | β (Q25) 95% CI | β (Q50) 95% CI | β (Q75) 95% CI |
|---|---|---|---|
| Distance between dwelling and excavation gallery > 1000 m | -1.00 (-3.77; 0.93) | -0.80 (-3.19; 0.86) | -1.25 (-4.98; -0.20) |
| Sheet metal dwelling | -1.00 (-3.22; 0.94) | -1.00 (-3.55; 1.54) | 0.50 (-1.74; 3.31) |
| Tarpaulin dwelling | 5.00 (3.11; 8.01) | 6.60 (4.65; 11.39) | 9.25 (6.47; 13.99) |
| Number of rooms in the dwelling ≤ 2 | 0.00 (-1.81; 2.32) | -1.80 (-3.08; 0.16) | -3.75 (-5.25; -0.39) |
| Presence of a ventilation system | -1.67 (-3.82; -0.61) | -4.00 (-6.34; -1.31) | -5.75 (-10.13; -2.03) |
| Cooking inside the home | -0.67 (-1.48; 2.55) | 1.00 (-0.93; 2.19) | 1.00 (-0.93; 4.19) |
| Main occupation of a mining craftsman | 0.00 (-0.70; 2.31) | -0.60 (-2.63; 2.90) | 1.75 (-3.00; 4.88) |
| Main occupation other | -0.67 (-2.43; 1.55) | -2.20 (-4.60; 0.93) | -0.50 (-4.49; 2.49) |
| Presence of at least one smoker | 2.67 (1.52; 5.18) | 5.60 (3.57; 7.13) | 6.00 (2.90; 8.53) |
| Health problem in the last week preceding the investigation | 0.67 (-0.87; 3.53) | 0.80 (-0.80; 5.04) | 1.50 (-0.61; 4.51) |
Note: β : regression coefficient ; 95% CI: 95 % confidence interval ; P25: 25th percentile ; P50: median; P75: 75th percentile.