Impact of regional digital transformation on public health: an empirical analysis based on 31 provinces in China | BMC Public Health

Descriptive statistical analysis of variables

Descriptive statistics for the core variables are summarized in Table 3. These results indicated obvious differences in digital transformation and life expectancy across provinces. Additionally, some provinces have not yet undergone digital transformation. Among the control variables, we found notable differences in the values of demographic characteristics, medical resources, and health environment across provinces, indicating that these control variables may be affected.

Benchmark regression results

Model (2) was employed to empirically test Hypothesis 1 to verify the impact of digital transformation on the level of public health. Regression result (1) indicated that the coefficient linking digital transformation and public health was 0.0017, excluding control variables, proving significant at the 10% level of significance. By contrast, regression result (2) demonstrated that the inclusion of control variables, remained statistically significant at the 1% level. This suggested a notably heightened impact of digital transformation on public health, with the supplementary control variables reinforcing this effect. Regression results (3) and (4) showed that digital transformation was significantly positive at the 1% significance level, whether province and year were fixed or not, which tentatively suggested that digital transformation may enhance life expectancy. Thus, digital transformation promotes public health. However, the mechanism responsible for this effect requires additional testing. The prior analysis indicated that digital transformation has a positive impact on public health, thereby validating Hypothesis H1 (Table 4).

Mechanism test

This study explored the mediating effect of technological innovation using the stepwise test of regression coefficients method. In Table 5, column (1)(2)(3) constitute a set of mediating effect equations and column (1) is a regression of digital transformation with life expectancy as the explanatory variable. The results showed that digital transformation positively enhanced life expectancy. Column (2) regresses digital transformation with technological innovation as the explanatory variable, and the results showed a significant contribution of digital transformation to technological innovation. Column (3) confirms the mediating effect of technological innovation in the digital transformation process to enhance life expectancy. The regression coefficient for technological innovation was significant at the level of 0.1, indicating a notable mediating effect. Thus, hypothesis H2 and H3 were tentatively verified.

To ensure the robustness and credibility of the mediated effect results, we examined the influence mechanism via the bootstrap method with 1000 iterations. As shown in Table 6, the confidence interval for the indirect effect of digital transformation on public health excluded zero, further confirming that technological innovation played a mediating role in the relationship between digital transformation and public health.

Threshold effect analysis

Threshold value estimates

Threshold effects were confirmed with 500 random repetitions in the study sample. The results are shown in Table 7, in which columns (1) and (2) show the test results of the threshold effect of digital transformation, and columns (3) and (4) are the test results of technological innovation as the threshold. The data in the table showed that digital transformation passed the test of significance at the 10% level in the double-threshold model, and technological innovation passed the test of significance at the 5% level.

Threshold model estimation

According to the above threshold analysis, the digital transformation and technological innovation threshold and the coefficients of each variable were incorporated into Eq. (6) for verification, and the results are shown in Table 8. When digital transformation was below the first threshold, the coefficient estimate of digital transformation was significant at − 0.0324. In the early stage of digital transformation, the digital advantage was not extensively integrated into the countryside, but it could increase the urban-rural development gap to a certain extent, which was detrimental to the overall development of public health. When digital transformation was between the first and second thresholds, the coefficient estimate of digital transformation increased to 0.01013 and passed the significance test at the 1% level. When digital transformation crossed the second threshold, the coefficient decreased to 0.005928. Thus, along with the development of digital transformation, its impact on public health increased first and then decreased. Meanwhile, when technological innovation was less than the threshold, the coefficient estimate of digital transformation was 0.02527, and it was significant at the 1% level. When technological innovation crossed the first and second thresholds, the coefficient estimate of digital transformation decreased. Digital transformation transcends the temporal and spatial barriers of traditional medical care through digital technology, effectively improving the medical conditions in rural areas, promoting the gradual flow of medical resources to rural areas, improving the primary medical and health care service system, and fostering the comprehensive development of public health.

Heterogeneity analysis

This study divided the country into two distinct regions: the eastern region and the central and western regions, and thoroughly examined the effects of digital transformation on the level of public health in different regions. The specific results are presented in Table 9. In the central and western regions, the positive impact of digital transformation on public health was evident and statistically significant. This finding contradicted the views of some scholars, including those who hypothesized that digital transformation has developed more rapidly in the economically advanced eastern regions than in the other regions. These scholars argue that the impact of digital transformation on public health should be pronounced in these regions. The following reasons may explain this phenomenon: (1) Policy support and regional development imbalance. The central and western regions have received increased policy support in recent years, particularly in infrastructure construction and digital transformation. The 13th Five-Year Plan for the Development of Western China, issued by the National Development and Reform Commission, calls for the strengthening of the construction of the medical and healthcare service system, the improvement of the primary medical and healthcare service system, the strengthening of the configuration of mobile platforms such as mobile medical vans, the steady upgrading of the level of the medical and healthcare service system, building a population health information platform in the western region, and encouraging the construction of a telemedicine information network covering hospitals at the county level and above. For example, by 2024, 100% of counties and townships will have 5G signal coverage in Nagchu, Tibet, providing infrastructure support for the whole-area coverage of the 5G + telemedicine system, which can greatly shorten the time of treatment for acute plateau diseases. At the same time, China’s 14th Five-Year Plan for National e-Government Development explicitly advocates implementing the East-West Collaboration Project for Digital Government Construction. This initiative prioritizes establishing public health information platforms in central and western China by developing cross-regional governmental data-sharing mechanisms. The “Thousand Counties Project” launched in 2021 facilitates extending provincial and municipal medical resources to county-level facilities, addresses service deficiencies and enhances management capabilities in county hospitals, while enabling integrated medical resource sharing across county healthcare systems. This policy tilt has effectively narrowed the digital medical divide between regions. (2) The widespread adoption of digital infrastructure. In 2020, the Chinese government identified three major directions for new infrastructure: information infrastructure, convergence infrastructure, and innovation infrastructure. The Midwest prioritizes establishing foundational digital healthcare infrastructure, while Eastern regions concentrate resources on developing advanced AI-integrated healthcare systems. In 2022, China officially launched the “East Counts, West Counts” project, which involved the planning and construction of national arithmetic network hubs in eight nodes in central and western China, such as Guizhou, Inner Mongolia, and Gansu. This initiative entails the establishment of new-generation digital infrastructure, such as data centers, arithmetic networks, and 5G, thereby enhancing digital infrastructure in central and western China and effectively compensating for the deficiencies of traditional healthcare systems. Consequently, the marginal benefits of digital transformation for public health are particularly significant in central and western China. For example, in Guizhou Province, as China’s first national-level big data comprehensive pilot zone, a province-wide telemedicine collaboration network has been constructed, relying on the development of the National Healthcare Big Data Centre. In 2021, 14 5G + healthcare applications in the province were successfully selected as national pilots, the construction of seven Internet hospitals was initiated, the province’s telemedicine service volume reached 771,000 cases, and the cumulative number of telemedicine services exceeded 2.3 million cases. Thus, substantial progress was made in the promotion of health informatization and Internet + healthcare. In addition, the Guiding Opinions on Promoting the Development of “Internet + Medical and Healthcare” specifically emphasizes the need to support impoverished areas in the central and western regions. This policy advocates accelerating the standardized construction of primary medical and healthcare institutions, enhancing equipment maintenance capabilities at the grassroots level, improving mobile broadband network coverage, and prioritizing the development of facilities such as intelligent health management terminals, teleconsultation systems. These measures aim to bridge the infrastructure gap, thereby enabling central and western regions to capitalize on a late-mover advantage. By contrast, the eastern region focused on the intelligent upgrading of existing systems in the layout of new infrastructure. For example, the Health Brain + Intelligent Healthcare construction promoted by Zhejiang Province is mainly targeted at third-level hospitals for the transformation of Internet of Things equipment and the optimization of diagnosis and treatment, resulting in improvements in public health in terms of service efficiency rather than breakthroughs in accessibility.

Robustness tests

Digital transformation lags one phase

Digital transformation is a process that requires joint promotion from the government, industry, and society. The implementation of policies such as electronic health records and e-healthcare insurance usually necessitates advancement through multiple phases, including planning, piloting, and marketing. The complete effect of such policies is not usually evident in the short term. Simultaneously, delayed processing has been demonstrated to be an effective method for alleviating the bias caused by two-way causality. Therefore, this study introduced a one-period lag in the explanatory variable for digital transformation. The detailed results of regression analysis are presented in Table 10, column (1).

Replacement of independent variable

To overcome the possible adverse effects of measurement bias in the independent variable on the empirical results, drawing on the practice of Zhao [63], we used the entropy value method to calculate the provincial digital economy index to represent the digital construction of the region. This index was derived from five indicators, namely, the number of Internet broadband access users per 100 people, the proportion of employees in the computer services and software industry in the urban sector, the total amount of telecommunication services per capita, the number of mobile phone subscribers per 100 people, and the index of digital financial inclusion. Data were obtained from the Digital Finance Research Centre, Peking University ( and the National Bureau of Statistics ( The regression results are provided in column (2) of Table 10.

Replacement of dependent variable

Building on Zhang [64], the total cost of healthcare refers to the total amount of money consumed by the whole society for healthcare services in a certain period (usually 1 year) in a country or region, and it is considered one of the effective ways to understand the state of healthcare in a country. In this paper, we re-estimate the baseline model by using the per capita health care cost instead of the per capita life expectancy constructed in this paper. The data concerning per capita health costs were obtained from the China Health Statistics Yearbook. The regression results are provided in column (3) of Table 10.

1% Winsorization

In order to ensure the generalizability of the results, a 1% winsorized treatment was implemented for robustness testing purposes. The regression results are provided in column (4) of Table 10.

Endogeneity analysis

Despite the study’s effective control for non-time-varying regional characteristics and time-varying common trends through a two-way fixed effect model, the potential impact of omitted variables, particularly the dimensions of local government fiscal investment and population mobility, remains a subject for further discussion. First, based on Grossman’s theoretical framework of health capital [65], local government financial investment may contribute to the improvement of public health through two pathways: improving healthcare infrastructure and expanding public health services. In the course of digital transformation, regions with greater fiscal strength may possess a greater capacity to invest in digital infrastructure and public health concurrently. This could result in a synergistic effect that might lead to an upward bias in the estimates. Furthermore, the impact of population mobility on public health is of particular significance in the context of the COVID-19 epidemic [66]. Digital transformation has the potential to indirectly influence population mobility patterns by facilitating remote working, online healthcare, and other such services. These, in turn, can act as a conduit for the propagation of infectious diseases [67]. Therefore, this paper attempts to apply the following two approaches to the problem of endogeneity.

Dynamic panel fixed effects models

In this paper, the static panel fixed-effects model is extended to a dynamic panel fixed-effects model by introducing one period lag of the explanatory variables in the equation, and the two-step system generalized method of moments (SYS-GMM) is adopted, and the estimation results are shown in column (1) of Table 11. The results of AR(1) and AR(2) tests show that the model accepts the hypothesis of ‘no autocorrelation of the disturbance terms’, and the results of the Hansen test show that the hypothesis of ‘all instrumental variables is valid’ is accepted, that is, the instrumental variables are not over-identified. The above tests show that the estimation results of SYS-GMM are consistent and reliable. The estimated coefficients for digital transformation are significantly positive at the 1% level, indicating that the conclusion that digital transformation contributes to public health is still valid after endogeneity is taken into account.

Instrumental variable approach

Drawing on the methodology of Huang [68], the instrumental variable for this study was the number of post offices per million people in 1984. The number of post offices was utilized as a proxy variable to measure the level of early communication development and informatization infrastructure in a region. This, in turn, indirectly reflects the potential basis for digital transformation in the region. On the one hand, there is a significant positive correlation between a region’s early communication development and its subsequent digital transformation. On the other hand, the number of post offices in the historical record is a relatively fixed and objective indicator, so it is unlikely to directly influence public health outcomes in the region. This meets the conditions for an instrumental variable. Thus, the key criteria for instrumental variables were satisfied, namely, relevance and exogeneity. Furthermore, the number of post offices per million people in 1984 represented cross-sectional data, which were not inherently suited for econometric panel data analysis. In line with the approach by Nunn and Qian [69], a time-varying variable was introduced to construct panel instrumental variables. In particular, construct the interaction term between the number of post offices per million population across cities in 1984 and one-period lagged digital transformation. Column (2) of Table 11 presents the results of the two-stage least squares (2SLS) estimation. The results showed that digital transformation continued to exert a significant impact on public health. In addition, the F-statistic of the first-stage regression was 18.57, indicating the absence of a weak instrumental variable problem. The p-value in the non-identification test was less than 0.01, which indicated that the original hypothesis of the under-identification of instrumental variables was significantly rejected at the 1% level, suggesting the absence of a non-identification problem in this paper. Therefore, after using instrumental variables to deal with the endogeneity problem, the credibility of the findings of this paper was further strengthened.