Assessing the Association of Cooking Fuel Type on chronic respiratory diseases among middle-aged and older adults in China: Insights from residential area and self-care capability

Abstract

Objective

This study aimed to evaluate the association between clean versus solid cooking fuels and chronic respiratory diseases (CRD) in adults aged 45 and older, focusing on the moderating effects of urban-rural location and self-care ability (independent, partially independent, or dependent).

Methods

Using cross-sectional data from the 2020 China Health and Retirement Longitudinal Study, logistic regression was applied to examine the relationship between cooking fuel types and CRD, while exploring the moderating effects of residential area and self-care ability. Robustness checks were conducted to confirm the findings.

Results

The use of clean cooking fuels was linked to lower CRD prevalence across all groups. While urban residents had a higher baseline CRD risk, the relationship between clean fuels and CRD was similar in both urban and rural areas. Individuals with poorer self-care abilities had a higher CRD risk, regardless of fuel type, indicating that clean fuels alone may not fully mitigate health risks for this group.

Conclusions

Clean cooking fuels are linked to lower CRD prevalence in both urban and rural populations, and their protective effect is similar across these groups. Poorer self-care ability is associated with higher CRD risk, highlighting that interventions beyond fuel switching are necessary for vulnerable individuals with limited self-care abilities. These findings underscore the need for comprehensive public health strategies.

Highlights

• •Clean cooking fuels are linked to lower chronic respiratory disease risk in all groups.
• •Urban residents tend to have a higher chronic respiratory disease risk than rural residents.
• •The association between clean fuel use and chronic respiratory disease is observed in all areas.
• •Individuals with limited self-care ability show a higher chronic respiratory disease risk.
• •Additional measures may be needed for high-risk groups beyond fuel use changes.

1. Introduction

Chronic Respiratory Diseases (CRD) pose a major global public health challenge, especially in low- and middle-income countries (X. Li et al., 2022; Meghji et al., 2021; Zar and Ferkol, 2014). The growing aging population further increases the burden of these conditions, particularly among individuals with limited self-care abilities (Fang et al., 2020). Solid fuel use for cooking is a well-documented risk factor for respiratory diseases, including chronic bronchitis and emphysema (Amadu et al., 2023; Pathak et al., 2020; Regalado et al., 2006).

Hu et al. (Hu et al., 2014) found that burning wood and coal significantly raises exposure to fine particulate matter <2.5 μm (PM2.5), which is closely linked to lung cancer. Proper ventilation, such as chimney-equipped stoves, can reduce this risk. Studies from the China Kadoorie Biobank and the Prospective Urban and Rural Epidemiology study also confirm a strong link between solid fuel use and respiratory diseases (Chan et al., 2019; Hystad et al., 2019; Li et al., 2019). Similarly, Guercio et al. (Guercio et al., 2022) reported that prolonged exposure to solid fuel emissions increases both incidence and mortality rates of CRD. Meanwhile, Yang and Chen (Yang and Chen, 2024) found that using clean fuels reduces the risk of chronic lung diseases, whereas reliance on solid fuels significantly raises respiratory disease prevalence (Qiu et al., 2023).

Urbanization and the use of solid biomass fuels are strongly linked to a higher risk of acute respiratory infections in children (Amadu et al., 2023; Yu et al., 2020). (Odo et al., 2022) found that prolonged PM2.5 exposure in 35 low- and middle-income countries significantly increased acute respiratory infection incidence, particularly among boys, younger children, and rural residents, highlighting the severe impact of biomass fuel smoke on vulnerable groups.

Household cooking fuel is a key factor affecting respiratory health (Gordon et al., 2014; Siddharthan et al., 2018; Yang and Chen, 2024). Pollutants from cooking are more concentrated in enclosed spaces, heightening health risks (Hou et al., 2022). Urban housing, with limited ventilation, exacerbates this issue by trapping cooking fumes, and increasing exposure (Zhou et al., 2021). Additionally, elderly individuals with limited self-care abilities often rely on solid fuels, leading to prolonged exposure to indoor pollutants and greater health risks.

While previous studies have linked fuel type to respiratory health, they often focus on specific populations. Limited research has examined how this relationship varies across urban and rural areas or among individuals with different self-care abilities. The moderating role of self-care ability, particularly in enclosed urban environments where elderly individuals with poor self-care face greater risks, remains largely unexplored. As the aging population grows, understanding the association between cooking fuel on CRD becomes increasingly important (Connelly and Maurer-Fazio, 2016; Guo et al., 2022).

This study evaluates the link between cooking fuel types on CRD risk among individuals aged 45 and above, using prospective data from China Health and Retirement Longitudinal Study (CHARLS). It specifically examines how urban-rural differences and self-care ability moderate this relationship. A key innovation is the systematic analysis of these moderating effects, which have been largely overlooked. By filling this gap, the study contributes to epidemiological research and provides empirical evidence for targeted public health policies, particularly for urban health challenges and elderly individuals with limited self-care ability. The findings aim to guide policymakers in designing effective interventions to improve health outcomes for high-risk populations.

2. Methods

2.1. Study design

This study was based on a secondary analysis of cross-sectional population-level data, aiming to explore the link between clean and solid cooking fuels on CRD among adults aged 45 and above, with a particular focus on the moderating effects of urban-rural differences and self-care ability.

2.2. Data sources and sample selection

The CHARLS is a nationally representative survey of middle-aged and older adults in China, modeled after the US Health and Retirement Study and other international aging studies. It assesses aging-related health factors, including the association of social policies and major public health events (Guo et al., 2022). This study used the 2020 CHARLS cross-sectional dataset, which includes demographic, environmental, and health information. The dataset is publicly available and anonymized (http://charls.pku.edu.cn/en). Ethical approval for the use of this dataset was granted by the Peking University Institutional Review Board (IRB00001052–11015), and the study adhered to institutional guidelines for the protection of human subjects, privacy, and data security. We excluded households that did not cook, used other fuel types, or had missing data, resulting in a final sample of 18,715 cases (4927 using solid fuels and 13,788 using clean fuels).

2.3. Variables

2.3.1. CRD

CRD, the primary dependent variable in this study, was measured based on respondents' self-report of a doctor's diagnosis of CRD, including chronic bronchitis, emphysema, pulmonary heart disease, or asthma (excluding tumors or cancers). Respondents were classified into two groups: those with and without a doctor-diagnosed CRD.

2.3.2. Fuel type

Fuel type is the primary independent variable in this study. The classification was determined based on responses to the question: “What is the primary source of cooking fuel in your household?” The responses were categorized into two groups: solid fuels including coal, briquettes, straw, and firewood. And clean fuels including piped natural gas, liquefied petroleum gas, electricity, solar energy, and biogas.

The binary classification method (solid fuels vs. clean fuels) is widely supported by existing literature. Previous studies have demonstrated that this classification effectively differentiates fuel-related health risks and aligns with public health and environmental research standards (Chan et al., 2019; Hu et al., 2014; Yang and Chen, 2024) Specifically, solid fuels are associated with higher levels of indoor air pollution, which significantly increases the risk of respiratory diseases and other adverse health effects. In contrast, clean fuel sources produce fewer pollutants and are generally considered safer for household use.

By employing this classification method, comparability with previous research is ensured, enabling standardized evaluations of fuel-related health risks. It also provides valuable insights for policymakers to reduce household air pollution and promote cleaner energy alternatives for vulnerable populations.

2.3.3. Residential type

Residential Type is used as a moderating variable in this study and is derived from a multiple-choice question in the survey asking respondents to indicate their current place of residence. The available options were: the city center, suburban area, rural area, and others. For analytical purposes, we categorized responses into two groups: urban (including both city center and suburban area) and rural (rural area). Responses marked as “others” were excluded to maintain clarity and consistency in classification.

Importantly, this classification is based on self-reported actual residence, rather than household registration status (Hukou), to better reflect the respondents' living environment and its potential influence on health outcomes. It enhances the accuracy of assessing the relationship between residential context and health and contributes to a more nuanced understanding relevant to urban-rural public health interventions.

2.3.4. Self-care ability

Another moderating variable is self-care ability, which was measured using respondents' performance in activities of daily living and instrumental activities of daily living. Respondents who reported difficulty in activities such as dressing, bathing, eating, getting in and out of bed, or controlling bladder or bowel functions, and required assistance or could not complete these tasks, were classified as “dependent”. Respondents who had no difficulties in these activities and were able to complete other daily activities independently (such as cooking, shopping, making phone calls, taking medication, and managing finances) were categorized as “independent”. Those between these two categories were labeled as “partially independent”.

Covariates were included to control for confounding factors. These variables include age, gender, education level (treated as a continuous variable where higher values indicate higher levels of education), marital status (married and living with a spouse vs. other), annual household income, smoking status, and drinking status.

2.4. Statistical approach

In the statistical analysis, we initially employed a hierarchical logistic regression model with the occurrence of CRD as the dependent variable. Control variables were introduced progressively to assess the robustness of the results.

Subsequently, we conducted heterogeneity analyses using moderation effects to examine differences by residential area (urban vs. rural) and different levels of self-care ability. Following the marginality principle (Nelder, 1977), all moderation models explicitly included the respective main effects of fuel type and the moderator variables (e.g., residential area or self-care ability) to ensure the correct interpretation of interaction terms. However, to comprehensively assess subgroup differences even in the absence of significant interactions, we further performed marginal effects analysis by calculating predicted probabilities across subgroups (e.g., urban vs. rural). This approach aligns with recommendations to explore potential heterogeneity in observational studies, particularly when prior evidence suggests plausible subgroup variations (VanderWeele and Knol, 2014). At last, we applied the Bootstrap method for robustness checks (Efron and Tibshirani, 1994) to verify the stability of the model estimates. To address potential biases in the sample, we performed weighted analyses and conducted Firth's Logistic Regression (Rudolph et al., 2013; Firth, 1993). All statistical analyses were performed using Stata 17 software.

3. Results

3.1. Participant characteristics

This study utilized data from the 2020 CHARLS, including 18,715 participants aged 45 years or older, with 5.54 % diagnosed with CRD. Regarding cooking fuels, 26.33 % of households used solid fuels, while 73.67 % used clean fuels. The average age of participants was 61.81 years (SD = 9.81), with a gender distribution of 52.74 % male and 47.26 % female. Education levels were generally low, with a mean score of 0.323 (SD = 1.271). In terms of marital status, 75.48 % were married and living with their spouse. The average annual household income was 15,404.61 RMB (∼2234 USD, based on the 2020 exchange rate of 6.90 RMB/USD) (SD = 44,791.15 RMB, ∼6491 USD), reflecting significant income variation. Most participants did not smoke (74.93 %) or drink alcohol (73.48 %). Geographically, 36.38 % lived in urban areas, while 63.62 % resided in rural areas. Regarding self-care ability, 67.07 % could fully care for themselves, 26.19 % partially cared for themselves, and 6.73 % were completely unable to care for themselves (see Table 1).

Table 1.

Demographic and Lifestyle Characteristics of Participants (Middle-Aged and Older Adults in China, 2020).

Characteristic	Mean(SD)	n/ (%)
Chronic respiratory diseases
No		17,679 (94.46)
Yes		1036 (5.54)
Fuel type
Solid Fuels		4927(26.33)
Clean Fuels		13,788(73.67)
Age	61.81(9.81)
Gender
Male		8844(47.26)
Female		9871(52.74)
Education	0.32(1.27)
Marital Status
Married and Living Together		14,126(75.48)
Other		4589(24.52)
Annual Household income	(RMB): 15,404.61 (44,791.15) Equivalent in USD: 2234.00 (6491.00)
Smoke
No		14,024(74.93)
Yes		4691(25.07)
Drinking
No		13,751(73.48)
Yes		4964(26.52)
Self-reported Residential Type
Urban		6809(36.38)
Rural		11,906 (63.62)
Self-reported Self-care Ability
Independent		12,553(67.07)
Partially Independent		4902(26.19)
Dependent		1260(6.73)

Note: Solid fuels include coal, briquettes, straw, and firewood. Clean fuels refer to energy sources such as piped natural gas, liquefied petroleum gas, electricity, solar energy, and biogas.

3.2. Associations between clean fuels and the risk of CRD

To examine the association between cooking fuel types and CRD risk in the elderly and assess the robustness of this relationship, two regression models were applied, as presented in Table 2.

Table 2.

Association Between Cooking Fuel Type and Chronic Respiratory Diseases Among Middle-aged and Older Adults in China, 2020.

Variables	Model 1		Model 2
	OR(SE)	95 % CI	OR(SE)	95 % CI
Fuel type
Solid Fuels	1		1
Clean Fuels	0.79(0.06)	(0.69,0.90)	0.79(0.06)	(0.68,0.92)
Self-reported Residential Type
Urban			1
Rural			0.82(0.06)	(0.71,0.95)
Self-reported Self-care Ability
Independent			1
Partially Independent			1.77(0.13)	(1.52,2.05)
Dependent			2.53(0.27)	(2.05,3.12)

Note: Solid fuels include coal, briquettes, straw, and firewood. Clean fuels refer to energy sources such as piped natural gas, liquefied petroleum gas, electricity, solar energy, and biogas. Model 1 is unadjusted, while Model 2 is adjusted for age, gender, smoking, drinking, education level, annual household income, and marital status.

Model 1 assessed the crude association between cooking fuel type and CRD. The results indicated that clean fuel use was associated with a lower CRD risk compared to solid fuels (OR = 0.79, 95 % CI: 0.69–0.90). Model 2 adjusted for age, gender, education, marital status, annual household income, smoking, alcohol consumption, residential area, and self-care ability. The association between clean fuel use and lower CRD risk remained significant (OR = 0.79, 95 % CI: 0.68–0.92). Additionally, rural residents had a lower probability of CRD compared to urban residents (OR = 0.82, 95 % CI: 0.71–0.95). Self-care ability was strongly related to CRD risk, with partially dependent (OR = 1.77, 95 % CI: 1.52–2.05) and dependent individuals (OR = 2.53, 95 % CI: 2.05–3.12) showing higher odds.

3.3. Heterogeneity analysis based on residential area and self-care ability

To examine whether the relationship between clean fuel use and CRD risk varies by residential area or self-care ability, we conducted moderation analyses by including interaction terms (fuel type × residential area and fuel type × self-care ability) in regression models, the results are shown in Table 3.

Table 3.

Effect Modification of Cooking Fuel Type on Chronic Respiratory Diseases by Geographic Location and Self-Care Ability Among Middle-Aged and Older Adults in China, 2020.

Variables	Residential Area		Self-care Ability
	OR(SE)	(95 %CI)	OR(SE)	(95 %CI)
Fuel type
Solid fuels	1		1
Clean fuels	0.65 (0.12)	(0.46, 0.92)	0.73(0.08)	(0.59, 0.90)
Self-reported Residential Type
Urban			1
Rural	0.67 (0.12)	(0.47, 0.95)	0.82(0.06)	(0.71, 0.95)
Self-reported Self-care Ability
Independent	1
Partially Independent	1.76(0.13)	(1.52, 2.04)	1.62 (0.21)	(1.27,2.08)
Dependent	2.52(0.27)	(2.04, 3.12)	2.11 (0.39)	(1.47,3.03)
Fuel type$\mathit{\times }$Self-reported Residential Type
Clean fuels $\mathit{\times }$ Rural	1.27 (0.25)	(0.87, 1.86)
Fuel type$\mathit{\times }$Self-reported Self-care Ability
Clean fuels $\mathit{\times }$Partially Independent			1.13 (0.17)	(0.84,1.52)
Clean fuels $\mathit{\times }$ Dependent			1.30 (0.28)	(0.85,2.00)

Note: Solid fuels include coal, briquettes, straw, and firewood. Clean fuels refer to energy sources such as piped natural gas, liquefied petroleum gas, electricity, solar energy, and biogas. All models are adjusted for age, gender, smoking, drinking, education level, annual household income, and marital status.

The results indicate that rural residents had lower odds of CRD compared to urban residents, independent of fuel type (OR = 0.67, 95 % CI: 0.47–0.95). Clean fuel use was linked to a lower CRD risk (OR = 0.65, 95 % CI: 0.46–0.92), but the interaction term was not statistically significant (OR = 1.27, 95 % CI: 0.87–1.86), suggesting that the association between fuel type and CRD risk does not significantly differ between urban and rural areas.

As shown in Fig. 1, the X-axis represents the type of cooking fuel used (1 = Solid Fuels, 2 = Clean Fuels), while the Y-axis represents the marginal effects on CRD probability. The legend distinguishes between urban and rural areas, with the blue line representing urban residents and the red line representing rural residents. The CRD risk curve for urban residents remains consistently higher than that for rural residents, indicating that urban residents tend to have a greater probability of CRD regardless of fuel type. The urban curve in the figure appears steeper, suggesting that the difference in CRD probability between solid and clean fuel users is more pronounced among urban residents. However, since the interaction term is not statistically significant, this trend does not have statistical confirmation, meaning that the association between fuel type and CRD risk does not substantially differ between urban and rural areas.

Fig. 1.

Marginal Effects of Cooking Fuel Type and Self-reported Residential Area Interaction on Chronic Respiratory Diseases Among Middle-aged and Older Adults in China, 2020. (Note: Solid fuels include coal, briquettes, straw, and firewood. Clean fuels refer to energy sources such as piped natural gas, liquefied petroleum gas, electricity, solar energy, and biogas.)

Regarding self-care ability, individuals with partial or complete dependence were more likely to have CRD compared to independent individuals (OR = 1.76 and 2.52). Clean fuel use was consistently associated with a lower CRD risk across all self-care ability groups (OR = 0.73, 95 % CI: 0.59–0.90). However, similar to the residential area model, the interaction term was not statistically significant, indicating that the association between fuel type and CRD risk does not significantly vary across self-care ability levels.

Although the interaction term was not significant, Fig. 2 still offers key insights. The X-axis represents the type of cooking fuel used (1 = Solid Fuels, 2 = Clean Fuels), while the Y-axis represents the marginal effects on CRD probability, with the same axis definitions as in Fig. 1. The CRD risk curve for dependent individuals remains the highest, suggesting that they have the greatest probability of CRD regardless of fuel type. The curve for dependent individuals is relatively flat, implying that differences in fuel type may not be strongly related to changes in CRD probability for this group. However, as the interaction term was not significant, this trend lacks statistical confirmation.

Fig. 2.

Marginal Effects of Cooking Fuel Type and Self-reported Self-Care Ability Interaction on Chronic Respiratory Diseases Among Middle-aged and Older Adults in China, 2020. (Note: Solid fuels include coal, briquettes, straw, and firewood. Clean fuels refer to energy sources such as piped natural gas, liquefied petroleum gas, electricity, solar energy, and biogas.)

In contrast, individuals with greater self-care ability not only showed lower CRD risk but also appeared to experience a stronger association between clean fuel use and reduced CRD risk, where the difference in CRD probability between solid and clean fuel users is more pronounced in these groups.

These findings are consistent with Table 2, reinforcing the potential benefits of clean fuel use and supporting its promotion across the general population. However, the higher likelihood of CRD among certain high-risk groups (e.g., urban residents and dependent individuals) suggests that fuel switching alone may not fully address their risks.

3.4. Robust test

To ensure the robustness of the findings, we primarily report results based on maximum likelihood estimation. To validate these estimates, we conducted additional robustness checks, including Bootstrap resampling (5000 replications), Weighted Regression, and Firth's Logistic Regression(Firth, 1993). These alternative modeling approaches confirmed the stability of our results. For clarity, we focus on the maximum likelihood estimation model in our main analysis and provide the detailed results of Weighted Regression and Firth's Logistic Regression in the Appendix.

4. Discussion

This study identifies an association between cooking fuel type and CRD among middle-aged and older adults, demonstrating a lower likelihood of CRD among those using clean fuels. The findings suggest that clean fuel use is associated with a lower CRD risk. This result aligns with previous studies (Amadu et al., 2023; Gordon et al., 2014; Regalado et al., 2006). Promoting clean fuels may help reduce exposure to harmful pollutants and could be beneficial for respiratory health across different population groups,

However, the study also reveals differences in the relationship between clean fuel use and CRD risk across different population groups. Specifically, while clean fuel use was consistently linked to lower CRD risk, there was no significant difference in this association between urban and rural areas. This suggests that clean fuel use is beneficial across all groups, but its relative association with CRD risk does not vary substantially between urban and rural populations.

Moreover, regardless of the fuel type used, urban residents exhibit a higher CRD risk. This elevated risk is a significant public health concern and may be attributed to poorer indoor ventilation and higher baseline levels of ambient pollution in urban environments (Shen et al., 2019). The findings imply that, in addition to promoting clean fuel use, further measures—such as improving urban air quality—may be necessary to reduce health risks in urban populations.

Another key finding of this study is that individuals with impaired self-care ability are more likely to experience health challenges. Despite using clean fuels, these individuals still had a higher probability of CRD, which may reflect prolonged indoor pollution exposure, reduced mobility, and increased susceptibility to respiratory illnesses (Sun et al., 2023). As such, this study contributes to the literature by highlighting the association between clean fuel use and respiratory health in vulnerable populations with impaired self-care ability, suggesting that even with clean fuel adoption, they remain more susceptible to CRD.

These findings further emphasize that clean fuel use is associated with health benefits across all populations, supporting its broader promotion. However, the greater observed CRD risk among vulnerable groups (e.g., urban residents and individuals with impaired self-care ability) suggests that additional interventions beyond fuel switching may be needed, such as improving ventilation or enhancing healthcare access.

This study offers insights for public health policy, particularly for the design of clean fuel promotion strategies that prioritize high-risk groups. While clean fuel adoption is associated with benefits for all groups, the findings suggest that high-risk groups (e.g., urban residents and individuals with impaired self-care ability) may require additional interventions beyond fuel switching.

Promoting clean fuels should remain a priority for public health efforts across all population groups. However, policy interventions tailored to high-risk groups may be beneficial. Improving indoor air quality by isolating cooking emissions is particularly relevant for urban residents, who may be exposed to higher levels of pollutants due to denser living conditions. Enhancing healthcare access for individuals with impaired self-care ability is also essential, as they continue to face higher CRD risks even with clean fuel use. Providing targeted energy subsidies and environmental improvements in the homes of individuals with limited mobility or self-care ability could further support this vulnerable population. Raising awareness about the benefits of clean fuels and reducing barriers to access are critical steps in improving health outcomes among these groups. Finally, expanding access to clean fuels infrastructure may not only help address health disparities but also contribute to broader environmental sustainability.

This study has several limitations. Self-reported data may be subject to recall or social desirability biases. The cross-sectional design limits causal inference, and unmeasured confounders may remain despite adjustments for known factors. The absence of individual-level air quality data restricts precise exposure assessments. Additionally, the binary classification of CRD in the CHARLS dataset precludes disease-specific analyses. Future work should prioritize larger datasets, objective health metrics, and individual-level air quality monitoring to advance understanding of clean fuels's role in respiratory health.

5. Conclusion

This study highlights the association between clean fuel use and reduced chronic respiratory disease (CRD) risk among middle-aged and older adults. However, vulnerable populations—particularly those with impaired self-care ability—remain disproportionately affected, underscoring the need for complementary interventions (e.g., improved ventilation, and healthcare access) alongside clean fuels adoption. While promoting clean fuels is critical for respiratory health equity, addressing socioeconomic and functional vulnerabilities is equally essential. Future research should integrate longitudinal designs and individualized air quality data to refine environmental health strategies.

CRediT authorship contribution statement

Junzhou Xu: Writing – review & editing, Writing – original draft, Validation, Supervision, Software, Resources, Project administration, Methodology, Data curation, Conceptualization. Ling Zhang: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Project administration, Methodology, Investigation, Formal analysis, Data curation.

Declaration of competing interest

The authors declare that they have no conflicts of interest related to this work.

Appendix A. Robustness Checks using Weighted Regression and Firth's Logistic Regression.

The results of these analyses, presented in Table A1, Table A2, demonstrate that the relationships between fuel types, age, gender, and CRD remain consistent and statistically significant after adjusting for covariates， further confirming the robustness and reliability of our key estimates.

Table A1.

Weighted Regression Analysis for Robustness Check: Main Model and Moderation Effects of Residential Area and Self-Care Ability in China, 2020.

Variables	Main Effects Model		Residential Area		Self-care Ability
	OR(SE)	(95 %CI)	OR(SE)	(95 %CI)	OR(SE)	(95 % CI)
Fuel type
Solid fuels	1		1		1
Clean fuels	0.78(0.06)	(0.67, 0.92)	0.65(0.12)	(0.46, 0.93)	0.73(0.09)	(0.59, 0.90)
Fuel type$\mathit{\times }$Self-reported Residential Type
Clean Fuels & Rural			1.23(0.25)	(0.83, 1.82)
Fuel type$\mathit{\times }$Self-reported Self-care Ability
Clean fuels$\mathit{\times }$Partially Independent					1.13(0.17)	(0.83,1.52)
Clean fuels$\mathit{\times }$Dependent					1.30(0.29)	(0.84,2.01)
Self-reported Residential Type
Urban	1				1
Rural	0.83(0.06)	(0.72, 0.96)	0.69(0.13)	(0.49, 1.00)	0.83(0.06)	(0.72, 0.96)
Self-reported Self-care Ability
Independent	1		1
Partially Independent	1.79(0.14)	(1.54, 2.08)	1.79(0.14)	(1.54, 2.08)	1.65(0.21)	(1.28,2.12)
Dependent	2.54(0.28)	(2.05, 3.14)	2.54(0.28)	(2.05, 3.14)	2.12(0.40)	(1.47,3.06)
Note: Solid fuels include coal, briquettes, straw, and firewood. Clean fuels refer to energy sources such as piped natural gas, liquefied petroleum gas, electricity, solar energy, and biogas. All models are adjusted for age, gender, smoking, drinking, education level, annual household income, and marital status.

Table A2.

Firth's Logistic Regression for Robustness Check: Main Model and Moderation Effects of Residential Area and Self-Care Ability in China, 2020.

Variables	Main Effects Model		Residential Area		Self-care Ability
	OR(SE)	(95 %CI)	OR(SE)	(95 %CI)	OR(SE)	(95 % CI)
Fuel type
Solid fuels	1				1
Clean fuels	0.79 (0.06)	(0.68, 0.92)	0.65 (0.11)	(0.46, 0.91)	0.73 (0.08)	(0.59, 0.90)
Fuel type$\mathit{\times }$Self-reported Residential Type
Clean fuels & rural			1.28 (0.25)	(0.88, 1.87)
Fuel type$\mathit{\times }$Self-reported Self-care Ability
Clean fuels$\mathit{\times }$Partially Independent					1.13(0.17)	(0.85, 1.52)
Clean fuels$\mathit{\times }$Dependent					1.30 (0.28)	(0.85, 1.99)
Self-reported Residential Type
Urban	1				1
Rural	0.82 (0.06)	(0.71, 0.95)	0.67 (0.12)	(0.47, 0.94)	0.82 (0.06)	(0.71, 0.95)
Self-reported Self-care ability
Independent	1
Partially Independent	1.77 (0.13)	(1.53, 2.04)	1.76 (0.13)	(1.53, 2.04)	1.62(0.20)	(1.27,2.08)
Dependent	2.53 (0.27)	(0.01, 0.04)	2.53 (0.27)	(2.05, 3.12)	2.12(0.39)	(1.48,3.04)
Note: Solid fuels include coal, briquettes, straw, and firewood. Clean fuels refer to energy sources such as piped natural gas, liquefied petroleum gas, electricity, solar energy, and biogas. All models are adjusted for age, gender, smoking, drinking, education level, annual household income, and marital status.

Data availability

Data will be made available on request.