The Impact of Rural Energy Poverty on Primary Health Services Efficiency: The Case of China

Abstract

Primary healthcare is vital to achieving universal health coverage, as emphasized by Sustainable Development Goal 3 (SDG 3). However, energy poverty remains a critical yet overlooked barrier to the efficiency of primary healthcare services in rural China—precisely the focus of this study. It employs panel regression models and threshold analysis methods using data from 31 Chinese provinces for the period 2014–2021, sourced from the EPSDATA data platform. Robustness checks are performed using bootstrap procedures, accompanied by detailed mechanism analyses. The empirical results demonstrate that rural energy poverty significantly reduces primary healthcare efficiency, particularly in provinces initially characterized by lower healthcare performance. The mechanism analysis identifies four critical transmission channels—off-farm employment, internet intensity, food safety, and health education—through which rural energy poverty undermines healthcare outcomes. Furthermore, threshold regressions uncover nonlinear relationships, indicating that the negative impacts of rural energy poverty intensify when household medical expenditures exceed 10.9%, the old-age dependency ratio surpasses 22.61%, and the rural energy poverty index is higher than 0.641. In theoretical terms, this study identifies rural energy poverty as a critical determinant of primary healthcare efficiency, thereby addressing an important gap in the existing literature. At the policy level, the findings emphasize the necessity for integrated measures targeting both rural energy poverty and primary healthcare inefficiencies to achieve SDG 3 and sustainably promote equitable, high-quality healthcare access in rural China.

1. Introduction

Healthy individuals are the cornerstone of sustainable economic development, as a productive and resilient workforce drives long-term growth. Achieving universal health is a core objective of the United Nations Sustainable Development Goal 3 (SDG 3) [1], which emphasizes the necessity of ensuring healthy lives and fostering well-being for all at all ages. Central to this goal is the performance of primary health services, which often serve as the first point of contact between individuals and the healthcare system [2]. As the most accessible level of care, primary health services play a critical role in delivering timely, cost-effective, and equitable interventions—particularly for vulnerable and underserved populations [3]. Consequently, improving the efficiency of primary health services not only strengthens the responsiveness of health systems and resource allocation but also serves as a pivotal step for advancing population health and achieving the broader targets set under SDG 3.

However, primary health services do not function in isolation; their efficiency is shaped by a complex interplay of contextual and systemic factors [4]. Among these, energy poverty—an often overlooked but critical barrier to sustainable development [5,6]—has received limited attention in health systems research. Defined as the lack of access to modern and reliable energy services, including electricity and clean cooking fuels, energy poverty imposes several constraints on human well-being and socio-economic progress [7,8,9,10,11,12,13,14,15,16]. The importance of energy access is underscored in the United Nations Sustainable Development Goal 7 (SDG 7), which calls for “access to affordable, reliable, sustainable, and modern energy for all,” reflecting the integral role of energy in advancing global development agendas. Currently, efficiency assessments are commonly conducted utilizing data envelopment analysis (DEA). The two primary models of Data Envelopment Analysis (DEA) are the CCR and BCC models [17]. The CCR model operates under the assumption of constant returns to scale, while the BCC model is predicated on the notion of variable returns to scale. The DEA delineates efficiency in three dimensions: comprehensive efficiency (derived from the CCR model), pure efficiency (derived from the BCC model), and scale efficiency [17].

While a growing body of literature has linked energy poverty to adverse health outcomes, such as impaired vision caused by insufficient lighting, increased respiratory diseases from indoor air pollution, and psychological stress associated with energy poverty [18,19,20], much of this research predominantly focuses on individual-level health effects [21,22]. In contrast, little attention has been given to the implications of energy poverty for the performance of health systems, particularly the efficiency of primary health services. As a critical entry point to care and a cornerstone of health system resilience, primary health services may be especially vulnerable to energy constraints. Yet, the potential impact of energy poverty on their efficiency remains insufficiently explored, leaving a significant gap in the literature.

Additionally, much of the existing literature on primary health services efficiency has concentrated on institutional, demographic, economic, or educational determinants [23,24,25]. While these factors are unquestionably significant, they provide only a partial view of the intricate dynamics affecting health system performance. A notable gap in this research is the limited attention given to energy poverty, despite the fact that energy access is a fundamental enabler of effective healthcare delivery. For instance, electricity is essential for powering medical equipment, preserving cold chains for vaccines, and maintaining the continuous operation of health facilities—particularly during emergencies. Without reliable energy access, the capacity of primary healthcare providers to deliver timely, safe, and effective services is severely impaired. This highlights energy poverty as a potentially critical, yet underexplored, determinant of primary health services efficiency. Therefore, exploring this relationship is particularly relevant in the context of sustainable development, as both reliable energy access and efficient healthcare delivery are key to achieving SDG 3 and SDG 7.

Rural energy poverty (REP) must be objectively assessed to systematically examine its impacts on primary health services efficiency. Notably, there is no universally accepted method for calculating REP, particularly in countries such as China, where diverse socioeconomic and infrastructural conditions present unique challenges. Existing evaluation frameworks often fail to capture all pertinent elements unique to China’s setting due to the multifaceted and complicated character of energy poverty [26]. Furthermore, the lack of consistent and comprehensive data further limits the applicability of indicators, underscoring the need for a context-specific and multidimensional evaluation framework that reflects China’s unique rural energy landscape and supports more accurate and policy-relevant assessments of energy poverty.

Many variables from the International Energy Agency’s (IEA) Energy Development Index (EDI), with the exception of the percentage of individuals having access to electricity, are relevant to China [26]. Likewise, specific metrics from the Multidimensional Energy Poverty Index introduced by Nussbaumer et al. (2012) [27], such as household appliances ownership, offer meaningful insights into energy affordability and are suitable for incorporation in this study. Pachauri et al. (2004) emphasized the importance of energy supply investment and household income as key dimensions of energy poverty [28], pertinent for evaluating energy poverty in China. Li and Zhou (2024) highlighted that energy use, energy supply, and energy management are critical components in addressing energy poverty [29]. This study refines and extends the indicator framework developed by Ren et al. (2024) and Zhao et al. (2021) to construct a comprehensive system for rural energy poverty in China [26,30].

We propose a complete evaluation index for rural energy poverty in China, comprising three primary categories: Energy Service Accessibility (ESA), Energy Management Comprehensiveness (EMC), and Energy Affordability and Energy Facility (EAF). Specifically, ESA denotes the capacity and opportunities for residents to utilize modern energy services, evaluated by energy use and supply. EMC reflects the management and investment capabilities of energy organizations. Management capacity influences the planning, construction, and maintenance of energy systems, while investment capacity directly affects the scale and quality of energy infrastructure. Weaknesses in either dimension may lead to high production costs, pricing volatility, and limited affordability, thereby intensifying energy poverty. EAF indicates energy affordability and infrastructure, assessed by household income, energy prices, and the ownership rate of high-energy-consuming appliances (e.g., air conditioners, range hoods). Ownership of such appliances implies access to reliable energy and the financial means to cover ongoing operating costs, thus serving as a proxy for improved energy affordability.

This study makes several unique contributions to the existing literature. First, it offers robust empirical evidence on the correlation between rural energy poverty and primary healthcare services efficiency, employing a range of estimation methods to enhance the credibility of the findings. Specifically, quantile regression is utilized to capture the heterogeneous effects of rural energy poverty across different levels of healthcare efficiency. Second, the study identifies four critical mediating pathways—off-farm work, internet intensity, food safety, and health education—through which rural energy poverty impacts primary health services efficiency. By unpacking these mechanisms, the analysis advances a more comprehensive understanding of the multifaceted channels linking energy deprivation to health system outcomes. Third, the study explores potential nonlinearities by conducting a threshold effect analysis, revealing that the negative impact of energy poverty on primary health services efficiency is significantly intensified under specific conditions—specifically, when household medical expenditure exceeds 10.9% of income, the old-age dependency ratio surpasses 22.61%, and rural energy poverty rises above 0.641. These findings underscore the compounded effects of economic and demographic vulnerabilities in shaping healthcare system efficiency.

The rest of the paper is organized as follows. Section 2 presents the data and the methodology used to construct rural energy poverty and primary health services efficiency indices. Section 3 presents and discusses the empirical results. Finally, Section 4 concludes and proposes the policy implications, limitations, and future directions of the study.

2. Methods

2.1. Rural Energy Poverty

The proposed indicators offered a thorough representation of the status of REP in China while also addressing issues associated with data scarcity. The comprehensive evaluation index for REP comprised three categories and 12 metrics, as detailed in

Table 1. Indicators for measuring rural energy poverty.

Category	Variable	Indicator	Measurement	Average Value	Standard Deviation
Energy Service Accessibility (ESA)	Energy consumption	Rural electricity consumption per capita	Rural electricity consumption/rural population (kWh/person)	2209.917	5657.948
	Energy supply situation	Rural electricity generation per capita	Rural electricity generation/rural population (kWh/person)	453.7145	524.4042
		Gas penetration rate	Gas penetration rate (%)	26.61729	23.71484
		Number of energy supply units	Number of legal entities in the electricity, heat, gas and water production and supply industry (units)	3505.391	2543.621
Energy Management Comprehensiveness (EMC)	Energy management capacity	Employed in energy management organizations	Employed persons in urban units of the electricity, heat, gas and water production and supply industry (10,000 persons)	12.37218	6.924199
	Energy investment capacity	State-owned energy industry inputs	Investment in fixed assets in the production and supply of electricity, steam and hot water in the state economy (billions of dollars)	293.1071	177.3158
		Urban energy investments	Investment in electricity, gas and water production and supply (billions of dollars)	751.5847	516.1026
Energy Affordability and Energy Facility (EAF)	Energy affordability	Rural household income	Per capita disposable income of rural residents (yuan)	14,884.68	5770.829
		Energy price	Retail fuel price index (previous year = 100)	104.1152	8.111185
	energy facilities	Ownership of air conditioners	Average air-conditioning ownership per 100 rural households (units)	53.95854	52.03622
		Ownership of range hoods	Average year-end ownership of range hoods per 100 rural households (units)	24.82298	19.30975
		Ownership of refrigerators	Average year-end ownership of refrigerators (cabinets) per 100 rural households (units)	92.33306	13.76631

. Data for these indicators were derived from EPSDATA, a professional and authoritative data platform that consolidated a wide range of official Chinese statistical yearbooks.

Additionally, the entropy technique was employed to compute the rural energy poverty index (REP) utilizing the indicators presented in Table 1. i represented the province, j signified the measurement indication, x specified the value of the indicator, and x′ referred to the normalization of the measurement. For positive measurements, the normalization was written as Equation (1). For negative measurements, the normalization was written as Equation (2).

$$x_{ij}^{\prime }={\displaystyle \frac{x_{ij}-min(x_{1j} ,\dots , x_{nj})}{max(x_{1j} ,\dots , x_{nj})-min(x_{1j} ,\dots , x_{nj})}}$$

(1)

$$x_{ij}^{\prime }={\displaystyle \frac{max(x_{1j} ,\dots , x_{nj})-x_{ij}}{max(x_{1j} ,\dots , x_{nj})-min(x_{1j} ,\dots , x_{nj})}}$$

(2)

In this paper, the ratio of the jth measurement in the ith province to the jth measurement in each province was calculated using the following equation.

$$p_{ij}={\displaystyle \frac{x_{ij}^{\prime }}{{\sum }_{i=1}^n{x}_{ij}^{\prime }}}$$

(3)

The entropy of the jth measurement was calculated as follows:

$$e_j=-k{\sum }_{i=1}^n{p}_{ij}Ln\left(p_{ij}\right)$$

(4)

where $k=1Ln\left(n\right)>0;e_j\ge 0$.

The method for calculating the information entropy redundancy was as follows:

$$d_j=1-e_j$$

(5)

The weights of each measurement were calculated as follows:

$$w_j={\displaystyle \frac{d_j}{{\sum }_{j=1}^m{d}_j}}$$

(6)

The following steps were used to determine the composite index (REP_i) for the energy poverty calculation:

$$R{EP}_i={\sum }_{j=1}^n{w}_j·x_{ij}^{\prime }$$

(7)

2.2. Primary Health Services Efficiency

This study constructed a comprehensive framework to assess Primary Health Services Efficiency (PHSE), grounded in a review of existing literature and guided by the principles of data accuracy, availability, and consistency in input-output indicators. As shown in

Table 2. Indicators for measuring the efficiency of primary health care services.

Type of Indicator	Indicator Name	Indicator Description
Input indicators	Number of institutions	Number of primary health care organizations, in units. Mainly includes community health centers (stations), township health centers, village health clinics, outpatient clinics and clinics.
	Number of beds	Number of beds in primary health care organizations, in units. They are mainly located in community health service centers and township health centers.
	Number of health personnel	Number of all employees working in primary health care, in 10,000 persons.
Output indicators	Number of consultations	Total number of visits to primary health care services for all consultations, in tens of thousands of persons.
	Number of hospitalizations	All hospitalizations for illness in primary health care services, in 10,000 persons.

, the model included three input indicators and two output indicators.

The input indicators consisted of the following: (1) the number of primary healthcare institutions, (2) the number of beds in primary healthcare institutions, and (3) the number of primary healthcare personnel. These indicators collectively captured the foundational infrastructure and resource capacity of primary healthcare systems. Specifically, the number of institutions and beds reflected the physical infrastructure necessary for service delivery, with higher values generally associated with broader service coverage and improved capacity, especially in rural and underserved regions. The number of healthcare personnel served as a key proxy for human resource input, which was critical for delivering effective and high-quality care. This indicator encapsulated the operational strength and service potential of primary healthcare facilities by reflecting the size and capability of the frontline healthcare workforce.

The output indicators consisted of two components: (1) the number of outpatient visits to primary healthcare institutions and (2) the number of inpatient cases that these institutions manage. Outpatient visits served as a key indicator of service utilization and operational efficiency, reflecting the daily activity levels of primary healthcare providers. A higher volume of outpatient visits typically indicated better accessibility, relevance, and trust in primary care services among the local population. In contrast, the number of inpatient cases reflected the capacity of primary healthcare institutions to manage more complex and severe health conditions. This metric was particularly significant as it demonstrated the institutions’ diagnostic and treatment capabilities. An increase in inpatient case volumes suggested enhanced capacity at the primary care level to address more serious medical needs, thereby offering a more holistic assessment of the efficiency and effectiveness of primary healthcare service delivery.

The selected indicators comprehensively covered key aspects of public health services, including resource inputs, service outputs, and overall service capacity. Their integration adhered to the principles of comprehensiveness and non-redundancy, ensuring that all relevant aspects of PHSE were captured without duplication or omission. This method improved the validity and reliability of the evaluation framework, providing a robust basis for systematically assessing PHSE.

The PHSE employed in this study was a comprehensive efficiency metric that assessed the capacity to produce a specified output with minimal input resources. Therefore, the input-oriented CCR model was utilized. The DEA assigns a score of 1 to all efficient DMUs, complicating the establishment of a hierarchy among them. Consequently, ranking is attainable solely for inefficient DMUs. This undermines the DEA’s efficacy as a metric for efficiency evaluation.

Andersen and Petersen (1993) introduced the concept of “super efficiency” as a method to establish a hierarchy among Decision-Making Units (DMUs) [31]. The core principle of the super-efficiency evaluation method involves excluding the effective evaluation unit from the dataset and conducting a re-evaluation; this preserves the initial non-effective value assessment, and if the original effective value assessment exceeds 1, a comparison can be made. We utilized a super-efficiency Data Envelopment Analysis for PHSE measurement. Let there be n Decision-Making Units (DMUs), m input indices, and q output indices. The subsequent model was employed to ascertain the PHSE (Equation (8)):

$$\begin{array}{c}\min \left(\theta -\epsilon \left({\sum }_{i=1}^m{s}_i^{-}+{\sum }_{j=1}^q{s}_j^{+}\right)\right)\\ \mathrm{s}.\mathrm{t}.\left\{\begin{array}{c}\begin{array}{c} \sum _{\begin{array}{c}k=1\\ k\ne j\end{array}}^n{\lambda }_k{x}_{ik}+s_i^{-}={\theta x}_i i=\mathrm{1,2},\cdots ,m\\ \sum _{\begin{array}{c}k=1\\ k\ne j\end{array}}^n{\lambda }_k{y_{jk}-s}_j^{+}=y_j j=\mathrm{1,2}, \cdots , q \\ {\lambda }_k\ge 0, k=1, \dots , n \end{array}\\ s_i^{-}\ge 0, s_j^{+}\ge 0 \end{array}\right.\end{array}$$

(8)

For the k th DMU, $x_{ik}$ denoted the i th input indicator, $y_{jk}$ represented the j th output indicator. and $s_i^{-}$ and $s_j^{+}$ were input and output slack variables, respectively. ${\lambda }_k$ denoted the weight coefficient. A higher θ value indicated higher PHSE.

2.3. Econometric Model

2.3.1. Baseline Model

This research investigated the impact of REP on PHSE. Consequently, REP served as the primary independent variable, while PHSE functioned as the dependent variable. This study employed dynamic panel methods to examine the potential lag effect of PHSE; the econometric model was structured as follows:

$${lnPHSE}_{it}={\alpha }_0+{\alpha }_1{lnPHSE}_{i,t-1}{+{\alpha }_{2}lnREP}_{it}+{\alpha }_3{Control}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}$$

(9)

in which i represented the province utilized in the analysis and t represented the year to which the variable corresponded. We referred to the random perturbation term as ${\epsilon }_{it}$, which had an independent and identical distribution. ${\nu }_i$ and ${\mu }_t$ were the individual and year fixed effects. ${\alpha }_0$ was the symbol used to represent the intercept term. The estimated coefficients we were most interested in were ${\alpha }_2$, which captured the impact of REP on PHSE. $Control$ denoted the control variables, mainly consisting of infrastructure development level (IDL), financial expenditure on healthcare (FEH), GDP per capita (PGDP) and urbanization (URB). IDL was quantified by road space per capita (square meters), FEH represented government fiscal spending on health care in billions of Yuan (RMB), PGDP denoted GDP per capita (Yuan), and URB indicated the proportion of the urban population.

2.3.2. Mediation Model

To enhance the comprehension of the processes by which PHSE is influenced by REP, the following model was developed to facilitate a mechanism analysis [32].

$${lnM}_{it}={\beta }_0+{\beta }_1{lnM}_{i,t-1}{+{\beta }_{2}lnREP}_{it}+{\beta }_3{Control}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}$$

(10)

$${lnPHSE}_{it}={\delta }_0+{\delta }_1{PHSE}_{i,t-1}{+{\delta }_{2}lnREP}_{it}+{\delta }_3{lnM}_{it}+{\delta }_4{Control}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}$$

(11)

where ${lnM}_{it}$ was the mediating variable, which was indicated by off-farm work (OFW), internet intensity (II), food safety (FS), and health education (HE). If the coefficients ${\beta }_2$ in Equation (10) and ${\delta }_3$ in Equation (11) remained significant, and the absolute value and significance of ${\delta }_2$ in Equation (11) decreased substantially in comparison to those of ${\alpha }_2$ in Equation (9), the variable ${lnM}_{it}$ performed a mediating role in the REP-PHSE nexus.

The ratio of employment in non-agricultural industries to total employment was the metric by which OFW was assessed. II was the ratio of mobile Internet subscribers (10,000) to mobile Internet access traffic (10,000 GB). The annual number of foodborne disease outbreaks was denoted by FS. The annual number of public health education activities was denoted by HE.

2.3.3. Threshold Regression Model

To test the potential nonlinear relationship between REP and PHSE, this study employed a panel threshold regression model, specified as follows:

$$Ln{PHSE}_{it}=\alpha +{\beta }_1{REP}_{it}·I({Thres}_{it}<\eta )+{\beta }_2{REP}_{it}·I({Thres}_{it}\ge \eta )+{\beta }_3{X}_{it}+{\mu }_i+{\epsilon }_{it}$$

(12)

where I (·) was an indicator function that took the value of 1 if the condition in parentheses was met and 0 otherwise. The model introduced a threshold variable (${Thres}_{it}$), including medical expenditure (as a share of household consumption) (ME), the old-age dependency ratio (OLD), and rural energy poverty (REP). The sample was divided into two regimes based on whether the threshold variable exceeded the estimated threshold value $\eta ,\mathrm{w}\mathrm{i}\mathrm{t}\mathrm{h} \mathrm{d}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{c}\mathrm{t} \mathrm{s}\mathrm{l}\mathrm{o}\mathrm{p}\mathrm{e} \mathrm{c}\mathrm{o}\mathrm{e}\mathrm{f}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{i}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{s}{\beta }_1$ and ${\beta }_2$ capturing the differential effects of energy poverty on healthcare efficiency across the two regimes. The empirical estimations were performed using Stata 17.0, a widely used econometric software for panel data analysis.

2.3.4. Descriptive Statistics

The selected variables involved in this paper were panel data of 31 provinces in China from 2014 to 2021. These data were sourced from the EPSDATA data platform, a publicly accessible, authoritative, professional, and accurate data organization that organized and compiled variables from the statistical yearbooks of each Chinese province. This study began in 2014 for three main reasons. First, from a policy perspective, in 2014, the former National Health and Family Planning Commission launched a major initiative to strengthen the service capacity of primary healthcare institutions. The Strategic Action Plan for Energy Development was released in 2014, emphasizing investment in rural energy infrastructure. This policy shift is directly relevant to the issue of rural energy poverty explored in this paper. This ensured that some important external policy factors can be avoided from that period. Third, from the perspective of data availability, the period from 2014–2021 is the most robust in terms of key indicators, and the analysis is more accurate and reliable during this period.

Table 3. Descriptive statistics of variables.

VarName	Obs	Mean	SD
lnREP	248	−0.439	0.372
lnPHSE	248	−0.468	0.167
lnOFW	248	−0.391	0.202
lnII	248	3.253	1.542
lnFS	248	4.355	1.218
lnHE	248	7.475	1.036
lnIDL	248	2.781	0.352
lnFEH	248	5.974	0.674
lnPGDP	248	10.965	0.411
lnURB	248	4.084	0.205
ME	248	8.490	2.009
OLD	248	16.544	5.052

shows the descriptive statistics of the selected variables and indices.

3. Results and Discussion

3.1. Baseline Regression Results

Table 4. Baseline regression results.

	(1)	(2)	(3)	(4)
	OLS	FE	RE	SYS-GMM
L.lnPHSE				1.090 ***
				(0.0161)
lnREP	−0.871 ***	−1.109 ***	−1.025 ***	−0.254 ***
	(0.212)	(0.190)	(0.179)	(0.0432)
lnIDL	−0.247 ***	0.113 *	0.0781	−0.0681 ***
	(0.0585)	(0.0608)	(0.0550)	(0.0223)
lnFEH	0.143 ***	−0.167 **	−0.0189	−0.238 ***
	(0.0400)	(0.0726)	(0.0569)	(0.0212)
lnPGDP	−0.230 **	0.354 ***	0.288 ***	0.263 ***
	(0.104)	(0.0687)	(0.0656)	(0.0290)
lnURB	0.294 *	−0.489 **	−0.324 *	−0.306 ***
	(0.176)	(0.225)	(0.183)	(0.0834)
Constant	0.309	−1.998 **	−2.650 ***	−0.136
	(0.834)	(0.912)	(0.754)	(0.233)
Year Dummy		YES	YES
AR(1)				0.000
AR(2)				0.1575
Sargan test				0.3411
Observations	248	248	248	217
R-squared	0.343	0.743
Number of id		31	31	31

Notes: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

presents the estimation results. In all four estimation approaches, the coefficients for rural energy poverty remain consistently negative and statistically significant, indicating that REP exerts a detrimental effect on PHSE. Specifically, the SYS-GMM estimation results show that a 1% increase in rural energy poverty leads to an estimated 0.254% decline in primary healthcare efficiency. This finding underscores the disruptive role of energy deprivation, which is often linked to infrastructural deficits such as unreliable electricity supply. Inadequate energy access hampers the operation of medical equipment, interrupts routine healthcare delivery, and undermines the overall quality and responsiveness of primary care services [33].

IDL exhibits a significantly negative effect, suggesting that general infrastructure investment may not translate into improved primary healthcare efficiency, likely due to misalignment between infrastructure development and the specific needs of the healthcare sector. FEH also shows a significantly negative coefficient, implying that increased financial input—when not accompanied by effective allocation and oversight—may fail to enhance service efficiency. PGDP exerts a significantly positive impact, implying that economic development facilitates more efficient healthcare delivery through improved resource availability and service accessibility. URB negatively impacts PHSE, as growing urban populations concentrate medical resources in large hospitals, leaving rural primary care underfunded. This imbalance leads to shortages of essential equipment, staff, and medicines, reducing the efficiency of primary healthcare services.

3.2. Quantile Regression Results

To capture the heterogeneity in the effect of REP on PHSE, this study employs quantile regression, with results reported in

Table 5. Quantile regression results.

	(1)	(2)	(3)	(4)	(5)
VARIABLES	q10	q25	q50	q75	q90
lnREP	−1.531 ***	−0.784 ***	−0.374	−0.119	−0.0735
	(0.288)	(0.267)	(0.331)	(0.196)	(0.180)
lnIDL	−0.513 ***	−0.332 ***	−0.206 ***	−0.0416	−0.0832 ***
	(0.0748)	(0.0791)	(0.0566)	(0.0351)	(0.0310)
lnFEH	0.155	0.230 ***	0.221 ***	0.155 ***	0.130 ***
	(0.114)	(0.0562)	(0.0629)	(0.0443)	(0.0273)
lnPGDP	−0.412 **	−0.122	−0.333 **	−0.262 **	−0.188 *
	(0.162)	(0.195)	(0.157)	(0.113)	(0.108)
lnURB	0.114	0.0961	0.691	0.822 ***	0.648 **
	(0.245)	(0.369)	(0.442)	(0.279)	(0.270)
Constant	2.956 **	−0.487	−0.482	−1.541 **	−1.277 **
	(1.394)	(1.333)	(1.014)	(0.756)	(0.509)
Observations	248	248	248	248	248

Notes: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

. Compared to baseline regressions, which merely indicate the average effect, quantile regression allows for a more nuanced understanding of how explanatory variables influence the dependent variable across its conditional distribution.

The findings indicate that the adverse impact of rural energy poverty on the efficiency of primary healthcare services is most pronounced in the lower quantiles. Specifically, the coefficients for rural energy poverty are significant at the 10% and 25% quartiles (−1.531 and −0.784, respectively) and become smaller and statistically insignificant in the higher quantiles. This evidence suggests that diminished healthcare service efficiency correlates with increased vulnerability to energy poverty, necessitating a focus on enhancing efficiency in regions with low healthcare service performance. The coefficients for IDL are significantly negative at the 10%, 25%, and 50% quartiles, indicating that general infrastructure investment may not be aligned with the specific needs of the healthcare sector. FEH shows positive and statistically significant coefficients at the 25%, 50%, 75%, and 90% percentiles, implying that financial inputs are more effective in improving healthcare efficiency in mid- and high-performing regions, where the institutional environment may be more capable of absorbing and utilizing additional resources. The effect of PGDP is significantly negative at the 10%, 50%, 75%, and 50% percentiles but becomes less negative at higher quantiles, reflecting that early stage economic development is not yet accompanied by proportional improvements in healthcare service delivery efficiency. Lastly, URB has an increasingly positive impact on PHSE in the upper quantiles, becoming statistically significant at the 75% and 90% quartiles, indicating that urbanization may enhance efficiency in regions that already have relatively strong healthcare systems, likely due to economies of scale and better access to healthcare infrastructure.

Figure 1 illustrates the quantile regression estimates of the impact of energy poverty on primary healthcare efficiency across different quantiles. The solid purple line represents the estimated coefficients, with the shaded area indicating confidence intervals, while the black dashed line denotes the OLS estimate for comparison.

The results reveal a heterogeneous effect of REP across the PHSE distribution. At lower quantiles (q10–q25), the negative impact is more pronounced, suggesting that regions with lower healthcare efficiency are more vulnerable to energy poverty. Better-performing healthcare systems are more resilient to energy constraints, as evidenced by the effect decreasing as quantiles rise. The OLS estimate appears to understate the severity of energy poverty’s impact in low-efficiency regions while overestimating it in high-efficiency regions, underscoring the advantage of quantile regression in capturing distributional effects.

3.3. Mechanism Analysis

To further explore the underlying mechanisms through which rural energy poverty influences primary health services efficiency, we employ a stepwise regression approach based on Equations (10) and (11) to conduct a mediation effect analysis. The results, presented in Columns (1)–(8) of

Table 6. Mechanism regression results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	OFW Mediating		II Mediating		FS Mediating		HE Mediating
VARIABLES	lnOFW	lnPHSE	lnII	lnPHSE	lnFS	lnPHSE	lnHE	lnPHSE
lnREP	−0.238 ***	−0.561 ***	5.422 ***	0.0802	1.714 ***	−0.140 ***	−2.286 ***	−0.210 ***
	(0.00922)	(0.0583)	(0.189)	(0.0519)	(0.363)	(0.0422)	(0.289)	(0.0475)
Mediator		−0.500 ***		−0.0299 ***		−0.00964 *		0.0267 ***
		(0.0621)		(0.00468)		(0.00569)		(0.00608)
lnIDL	0.123 ***	−0.0176	0.467 ***	0.00991	−0.589 ***	−0.0118	1.028 ***	−0.0811 ***
	(0.0107)	(0.0272)	(0.0631)	(0.0355)	(0.106)	(0.0180)	(0.121)	(0.0201)
lnFEH	0.0442 ***	−0.232 ***	1.152 ***	−0.136 ***	0.793 ***	−0.233 ***	0.0133	−0.252 ***
	(0.00797)	(0.0238)	(0.0720)	(0.0184)	(0.140)	(0.0331)	(0.136)	(0.0204)
lnPGDP	−0.0813 ***	0.299 ***	1.589 ***	0.365 ***	−0.121	0.458 ***	1.417 ***	0.251 ***
	(0.0171)	(0.0349)	(0.0615)	(0.0486)	(0.179)	(0.0525)	(0.154)	(0.0315)
lnURB	0.456 ***	−0.0928	−0.851 ***	−0.326 ***	−1.126 *	−1.058 ***	−3.612 ***	−0.262 ***
	(0.0399)	(0.108)	(0.115)	(0.100)	(0.601)	(0.126)	(0.419)	(0.0919)
Constant	−1.883 ***	−1.930 ***	−18.17 ***	−1.769 ***	5.753 ***	0.700 ***	3.285 *	−0.224
	(0.0831)	(0.491)	(0.642)	(0.513)	(1.563)	(0.246)	(1.757)	(0.261)
L.Mediator	0.487 ***		0.701 ***		0.540 ***		−0.0794 ***
	(0.0188)		(0.0134)		(0.0238)		(0.0270)
L.lnPHSE		1.079 ***		1.040 ***		1.028 ***		1.131 ***
		(0.0196)		(0.0322)		(0.0174)		(0.0180)

Notes: Standard errors in parentheses. *** p < 0.01, * p < 0.1.

, illustrate the mediating roles of off-farm work (OFW), internet intensity (II), food safety (FS), and health education (HE), respectively.

The existing literature suggests that rural energy poverty impacts healthcare efficiency through both direct infrastructural constraints and indirect socioeconomic and behavioral pathways. First, limited access to energy restricts rural residents’ ability to participate in off-farm employment by constraining mobility, digital connectivity, and skill development, thereby reducing household income and limiting access to higher-quality medical services [34]. Second, although digital connectivity can facilitate health information dissemination, over-reliance on the internet in energy-poor settings often exposes individuals to misinformation [35], which may erode trust in local healthcare systems. Third, inadequate refrigeration and food storage—common in energy-deprived regions—can compromise food safety and increase the incidence of foodborne illnesses and nutrition-related disorders, thereby adding pressure to already limited healthcare resources [36]. Finally, energy shortages hinder public health education initiatives by limiting digital infrastructure and outreach capacity, particularly in rural and remote areas. This reduces health literacy and undermines the adoption of preventive health behaviors, both of which are essential for enhancing primary healthcare efficiency [9]. Taken together, these four pathways highlight the multifaceted mechanisms through which rural energy poverty undermines the effectiveness of local health systems.

The results in Column (1) of Table 6 report a significantly negative coefficient (−0.238), indicating that higher levels of rural energy poverty reduce OFW opportunities, likely due to limited electricity access, poorer health, unstable income [37], and reduced exposure to non-agricultural skills and opportunities, which hinder rural laborers from transitioning to non-agricultural employment. Column (2) shows that OFW exerts a negative impact on PHSE, suggesting that off-farm work exacerbates healthcare inefficiencies. As off-farm work rises and rural labor migrates to urban or industrial centers, the resulting underutilization of local primary healthcare resources creates a mismatch between service capacity and actual demand, leading to operational inefficiency [34]. Therefore, a decrease in REP may be accompanied by an increase in OFW, which in turn decreases PHSE. This implies that the reduction of REP does not fully result in an increase in PHSE, which should be considered in the context of the eradication of energy poverty.

The findings presented in Column (3) of Table 6 show a significantly positive coefficient (5.422), indicating that higher levels of energy poverty are associated with greater internet intensity. This reflects a psychological compensation mechanism, where individuals in energy-deprived regions turn to the internet—particularly social media—for emotional support and a sense of participation due to limited physical infrastructure. However, Column (4) reveals a negative impact of II on PHSE, suggesting that increased digital engagement may erode trust in local healthcare. Exposure to unverified or emotionally charged information can amplify negative perceptions of primary healthcare, leading patients to bypass local clinics for tertiary hospitals—reducing primary-level utilization without a corresponding drop in capacity, and thus lowering overall efficiency. Also, the rapid spread of health-related content on social media may increase pressure on public health institutions, triggering resource misallocation and credibility fatigue, thereby hindering the effectiveness of public health communication and governance [38]. Thus, while internet access appears beneficial, it may inadvertently undermine healthcare efficiency in energy-poor areas.

The results in Column (5) of Table 6 show a significantly positive coefficient (1.714), indicating that rural energy poverty deteriorates food safety due to limited refrigeration, inadequate food preservation, and reliance on lower-quality food, increasing the risk of contamination. Column (6) reveals a negative impact of FS on PHSE (−0.00964), suggesting that poor food safety exacerbates healthcare inefficiencies by increasing the prevalence of foodborne illnesses and malnutrition. This heightened disease burden strains limited healthcare resources, leading to longer wait times, overburdened personnel, and reduced service efficiency, particularly in energy-deprived regions.

The results in Column (7) of Table 6 show a significantly negative coefficient (−2.286), indicating that severe energy poverty limits access to health education due to poor infrastructure, unreliable electricity, and restricted digital learning tools [9]. Column (8) reveals a positive effect of HE on PHSE (0.0267), suggesting that better health education can reduce pressure on primary healthcare facilities by promoting preventive care, improving disease awareness, and encouraging timely medical intervention, thereby enhancing healthcare efficiency.

3.4. Threshold Effect Analysis

To further investigate the nonlinear relationship between rural energy poverty and primary health services efficiency, we employ a threshold regression model based on Equation (12).

Table 7. Threshold regression results.

	(1)	(2)	(3)
	d1	d2	d3
VARIABLES	PHSE	PHSE	PHSE
lnIDL	−0.163 ***	−0.152 ***	−0.172 ***
	(0.0283)	(0.0322)	(0.0320)
lnFEH	0.0964 ***	0.0868 ***	0.0641 ***
	(0.0198)	(0.0224)	(0.0225)
lnPGDP	−0.145 ***	−0.0800	−0.0897
	(0.0508)	(0.0573)	(0.0570)
lnURB	0.317 ***	0.110	0.0756
	(0.0881)	(0.0974)	(0.0962)
REP (ME < 10.9)	−0.426 **
	(0.183)
REP (ME ≥ 10.9)	−1.555 **
	(0.719)
REP (OLD < 22.61)		−0.936 ***
		(0.203)
REP (OLD ≥ 22.61)		−1.006 ***
		(0.350)
REP (REP < 0.641)			−0.0823
			(0.271)
REP (REP ≥ 0.641)			−0.853 **
			(0.424)
Threshold Value	10.9	22.61	0.641
SSR	5.1694	6.6208	6.6011
Observations	248	248	248

Notes: Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

presents the estimation results, where Columns (1)–(3) correspond to different threshold variables: medical expenditure (ME), old-age dependency ratio (OLD), and rural energy poverty (REP), respectively.

Column (1) examines the moderating role of ME, identifying a threshold at 10.9. When ME is below this value, rural energy poverty negatively affects PHSE (−0.426), suggesting that in low-expenditure regions, healthcare inefficiencies are primarily driven by structural deficiencies rather than energy poverty alone. However, when ME exceeds 10.9, the negative impact of energy poverty intensifies (1.555), indicating that rising medical spending does not always translate into better service efficiency under energy-constrained conditions. Energy poverty can indirectly worsen health conditions by limiting access to clean cooking, heating, and sanitation, thereby driving households to seek more frequent or urgent care and increase medical expenditure [39]. The surge in demand for medical services raises the burden on medical institutions and limits service capacity, thereby amplifying the negative impact of energy poverty on primary healthcare efficiency.

Column (2) examines the threshold effect of OLD on the REP-PHSE relationship, identifying a threshold at 22.61. When OLD is below this value, REP negatively affects PHSE (−0.936), but when OLD exceeds 22.61, the impact intensifies (−1.006), indicating that aging populations are more vulnerable to energy poverty constraints. This is likely due to higher energy demands for home-based medical devices, heating, and emergency healthcare, making stable energy access essential for maintaining healthcare efficiency in aging regions.

Column (3) examines REP as a threshold variable, identifying a threshold at 0.641. When REP is below this value, its impact on PHSE is insignificant, suggesting that in regions with mild energy poverty, factors such as healthcare funding and infrastructure play a more dominant role. However, when REP exceeds 0.641, its negative impact on PHSE becomes significant (−0.853), indicating that severe energy poverty worsens healthcare efficiency. This effect can be attributed to inadequate power supply in medical facilities and restricted access to telemedicine and digital health services, further exacerbating inefficiencies in energy-deprived regions.

4. Conclusions and Implications

Living in energy poverty presents a distinct societal challenge [40], posing a significant risk to primary healthcare services efficiency, which is essential for public health and the achievement of SDG 3. Based on panel data from 31 provinces in China from 2014 to 2021, this study provides empirical evidence regarding the impact of energy poverty on primary healthcare services efficiency. The findings indicate that energy poverty significantly reduces healthcare efficiency, with its negative effects being particularly pronounced in regions experiencing higher levels of healthcare efficiency. Furthermore, we identify four critical mechanisms—off-farm work, internet intensity, food safety, and health education—through which energy poverty affects healthcare efficiency. Specifically, energy poverty limits off-farm work opportunities, reducing household income and restricting financial access to healthcare services. Although insufficient energy infrastructure appears to be associated with increased reliance on internet services, this unintended shift may divert public resources away from direct healthcare investment, failing to produce tangible improvements in primary healthcare efficiency. Additionally, energy poverty adversely affects food safety, increasing the prevalence of foodborne illnesses and malnutrition, consequently straining primary healthcare resources. Lastly, unreliable electricity supplies in schools and community settings hinder effective health education programs, reducing public awareness of preventive health measures and undermining early disease intervention efforts.

Another significant finding is that the negative impact of energy poverty on primary healthcare services efficiency intensifies under specific conditions. Specifically, its negative effects become significantly more pronounced when household healthcare expenditure surpasses 10.9%, the elderly dependency ratio exceeds 22.61%, and rural energy poverty rises above 0.641, highlighting the compounded effect of demographic and economic vulnerabilities.

Given the significant adverse impacts of energy poverty on primary healthcare services efficiency, targeted policy interventions are essential. Firstly, expanding rural electrification and improving overall energy access should be prioritized, especially in regions with low primary healthcare efficiency. Investments in renewable energy technologies—including solar, wind, and off-grid power systems—could provide stable and sustainable electricity to rural medical facilities, ensuring reliable service delivery. This could be achieved by channeling central and local funding into clean energy infrastructure specifically designed for rural clinics and township hospitals. Secondly, implementing electricity subsidy programs for low-income households alongside progressive healthcare financing models can reduce financial barriers to medical services and mitigate the economic constraints posed by energy poverty. For example, electricity vouchers could be incorporated into existing rural medical assistance schemes, coordinated through local health insurance and poverty alleviation offices. Thirdly, as part of China’s hierarchical medical system reform, strengthening public trust in primary healthcare institutions is essential. Policymakers should enhance the quality, transparency, and digital visibility of primary care while regulating online health content to maintain credibility. This can be supported by launching a national “Trusted Primary Care” campaign that promotes verified health information through official rural doctor platforms and locally embedded community media. In line with prior evidence that digital tools such as rural e-commerce contribute to narrowing income inequality in rural areas [41], integrated strategies that combine energy access with digital inclusion may offer synergistic benefits for improving healthcare efficiency.

Additionally, improving rural cold-chain logistics and upgrading food storage infrastructure could markedly enhance food safety, reducing healthcare burdens associated with foodborne illnesses and nutrition-related conditions. Local governments may consider investing in solar-powered community refrigeration hubs co-located with health centers, especially in agricultural production regions with weak storage facilities. Furthermore, ensuring a stable electricity supply in educational institutions and community health facilities is equally vital, as it supports public health education, preventive care initiatives, and timely medical interventions. This objective may be supported by prioritizing infrastructure improvements in rural schools and health centers—specifically through the deployment of decentralized renewable energy systems, battery storage, and auxiliary power sources—to ensure the uninterrupted delivery of essential services. Special emphasis should also be placed on regions with higher elderly dependency ratios, ensuring an adequate and stable energy supply for elderly care facilities and home-based medical services to address increasing healthcare demands from aging populations. Local energy planning can designate geriatric care infrastructure as a protected load priority, while subsidies are offered for essential home medical devices such as oxygen concentrators. Lastly, investments in energy-efficient medical equipment and resilient power solutions are critical for mitigating healthcare inefficiencies arising from energy shortages, thereby supporting sustainable and equitable healthcare service delivery. This can be achieved by promoting the adoption of low-energy diagnostic technologies and equipping rural health centers with hybrid solar-battery systems to ensure continuity of care during power disruptions.

This study offers several important strengths. First, it provides a rare empirical investigation into the relationship between rural energy poverty and primary healthcare service efficiency, a topic that has received limited attention in the existing literature. Second, the use of a robust statistical framework—including panel regression, threshold analysis, and mediation analysis—enhances the internal validity of the findings. Third, all data were sourced from EPSDATA, a reputable and comprehensive platform that consolidates official Chinese statistical yearbooks, ensuring reliability and consistency.

Nevertheless, like any research, this study has some limitations. First, while this study fills a gap in population-level research, it should be noted that such data lie lower on the evidence hierarchy compared to individual-level studies, which offer stronger causal inference in healthcare research. Also, the use of provincial-level data limits the granularity of the analysis; finer-scale data (e.g., county or village level) could better capture regional heterogeneity and enhance the robustness of the findings. Second, although this study employs a stepwise regression approach to identify key mediating mechanisms, potential endogeneity between energy poverty and healthcare efficiency cannot be fully ruled out. Future research may adopt instrumental variable techniques or quasi-experimental methods to address this concern more robustly. Third, although four mediating channels were examined—off-farm employment, internet intensity, food safety, and health education—other potential mechanisms, such as household behavioral responses and environmental or institutional factors, warrant further exploration. Future research could expand on these dimensions to offer a more comprehensive understanding of energy poverty’s multidimensional impacts on healthcare efficiency.