internet-access-linked-to-better-self-rated-health-alongside-increased-depressive-symptoms
Internet Access Linked to Better Self-Rated Health Alongside Increased Depressive Symptoms

Internet Access Linked to Better Self-Rated Health Alongside Increased Depressive Symptoms

A comprehensive longitudinal investigation tracking tens of thousands of adults across China reveals that engagement with digital technologies carries a distinct double-edged consequence for personal well-being. Analyzing five biennial survey waves from the China Family Panel Studies collected between 2014 and 2022, researchers found that basic internet access was positively associated with general self-rated health, yielding an ordered logit regression coefficient of 0.097 and an odds ratio of 1.102. However, this same connectivity was simultaneously associated with a higher reported frequency of feeling depressed, showing a coefficient of 0.069 and an odds ratio of 1.071. These concurrent outcomes highlight the coexistence of empowering and burdening mechanisms within modern digital engagement. Rather than operating as an unmitigated social benefit or a uniform public health hazard, digital technology use demonstrates complex, divergent associations that depend heavily on weekly duration, device configurations, the functional nature of specific online tasks, and the demographic environments of individual users across contemporary society.

The theoretical motivation behind this inquiry stems from the rapid expansion of digital infrastructure and national digital literacy initiatives, such as China’s Action Outline for Improving Digital Literacy and Skills for All. While previous scholarship frequently treated technology adoption as a simplistic binary measure of access, this study conceptualizes digital engagement as a multidimensional behavioral construct. Drawing upon an adapted formulation of Ronald Andersen’s Behavioral Model of health service utilization, the authors hypothesized that health-related outcomes are jointly shaped by behavioral predisposing characteristics, enabling resources, and broader environmental contexts. Observable behavioral dimensions—including extensive connectivity, weekly usage hours, the technical platforms employed, and the functional purposes of online activities—capture fundamentally distinct modes of social interaction. By distinguishing simple access from the intensity and diversity of use, the investigation sought to explain conflicting findings in previous literature, where digital tools have variously correlated with improved physical vitality, reduced social isolation, heightened anxiety, or negligible psychological changes.

To rigorously evaluate these behavioral dimensions over time, the research team assembled an analytical panel dataset drawn from the China Family Panel Studies, a nationally representative survey covering twenty-five provinces, autonomous regions, and municipalities. Restricting the target population to adult respondents aged eighteen and older and eliminating records with missing entries across relevant indicators resulted in an unbalanced panel containing 129,098 person-wave observations from 42,952 unique individuals. Descriptive comparisons between internet users and non-users revealed pronounced structural contrasts across the sample. Internet users were substantially younger on average, with a mean age of 37.15 years compared to 55.91 years among non-users. Connected individuals also demonstrated higher average educational attainment, greater household and personal income, lower rates of chronic disease, and a higher likelihood of non-agricultural employment. These systematic baseline differences underscored the necessity of employing analytical models that control for time-invariant individual selection bias and observed confounding social attributes.

Methodologically, the authors addressed unobserved individual heterogeneity and the ordinal structure of the survey responses by employing a fixed-effects ordered logit model with year fixed effects. Implemented through the blow-up and cluster estimator developed by Baetschmann and colleagues, this econometric technique dichotomizes the ordered categorical outcome at every cumulative threshold and combines the conditional likelihoods. This strategy conditions out unobserved, time-invariant individual characteristics while retaining all threshold transitions, with standard errors clustered at the individual level to correct for repeated observations over time. The primary dependent variables comprised self-rated health, graded on a five-point scale from unhealthy to extremely healthy, and depressive symptom frequency, recorded on a four-point scale ranging from never to often or daily. Time-varying covariates integrated into the specifications included marital status, formal educational attainment, employment categorization, log-transformed personal income, chronic disease diagnosis, medical insurance coverage, pension enrollment, and residential registration status.

Evaluating the intensive margin of engagement demonstrated that prolonged connectivity introduces significant diminishing returns for physical well-being. When weekly internet hours were evaluated linearly, longer usage duration was negatively associated with self-rated health, exhibiting a statistically significant coefficient of negative 0.003 and an odds ratio of 0.997. Conversely, weekly duration showed a positive linear association with depressive symptom frequency, with a coefficient of 0.001. When testing for non-linear dynamics by adding a quadratic term for weekly hours, the linear duration coefficient for self-rated health remained negative at negative 0.002 but lost statistical significance, while the quadratic term was negligible. For depressive symptoms, incorporating the quadratic term shifted the linear coefficient to negative 0.001 without statistical significance, alongside a marginally positive quadratic term of 0.000.

Examining the hardware platforms through which individuals connected to the internet revealed further behavioral differentiation. In non-exclusive specifications, accessing the internet via mobile phones and computers both demonstrated positive, statistically significant associations with self-rated health, with coefficients of 0.106 and 0.136, respectively. Neither device type exhibited a statistically significant link to depressive symptoms in baseline models. A direct Wald test comparing mobile phones to personal computers indicated no significant difference in their relative associations with self-rated health. However, evaluating mutually exclusive device configurations illustrated substantial advantages for multi-device users. Individuals relying exclusively on mobile devices reported a positive self-rated health coefficient of 0.113, whereas those utilizing both mobile phones and personal computers demonstrated a coefficient of 0.240. Pairwise coefficient comparisons confirmed that utilizing both platforms yielded a significantly stronger association with self-rated health than relying on mobile phones alone, reflecting how complementary hardware ecosystems expand capabilities for health information retrieval and complex communication.

Disaggregating online engagement into four distinct activity domains—learning, socializing, entertainment, and shopping—demonstrated that the purpose of digital use strongly governs its health correlates. Participation in online educational activities and social networking was positively and significantly associated with self-rated health, producing coefficients of 0.103 and 0.113, respectively. In contrast, online entertainment and commercial shopping showed no statistically significant relationship with subjective physical health. For psychological outcomes, online learning was positively associated with the frequency of feeling depressed, yielding a coefficient of 0.068, whereas social networking, entertainment, and shopping showed no significant association. Pairwise comparisons confirmed that learning and socializing had significantly stronger positive associations with self-rated health than entertainment or shopping. Strikingly, when examining overall activity breadth among active internet users, the aggregate count of online activities was negatively associated with self-rated health, with a coefficient of negative 0.059, while showing no significant association with depressive symptoms, countering assumptions that broader digital activity portfolios are inherently advantageous.

Subgroup analyses revealed that these health associations diverge sharply across demographic life stages. Stratifying the population using a threshold of forty-five years established that the negative association between weekly usage time and self-rated health was concentrated entirely among younger adults aged eighteen to forty-five, who registered a coefficient of negative 0.005, whereas middle-aged and older adults exhibited an insignificant coefficient of 0.001. Even more pronounced generational disparities emerged regarding psychological distress. Internet access and weekly duration were positively and significantly linked to depressive symptoms among younger adults, showing coefficients of 0.165 and 0.002, respectively, but were completely non-significant among adults older than forty-five. Furthermore, engaging in a wider diversity of online activities was positively associated with depressive symptoms exclusively within the younger cohort, showing a coefficient of 0.044, compared to an insignificant negative 0.002 in older adults.

Geographic and residential stratification highlighted substantial structural disparities between rural and urban populations, as well as across China’s major geographic macro-regions. While internet access supported self-rated health across both rural and urban areas with comparable coefficients of 0.097 and 0.096, weekly duration imposed a significantly heavier penalty on self-rated health in urban centers than in rural settlements. Regarding mental health, internet access and activity diversity were positively associated with depressive symptoms among urban residents, recording coefficients of 0.101 and 0.032, but showed no significant relationship among rural dwellers. Regionally, internet connectivity was positively linked to self-rated health in the western and central provinces, with coefficients of 0.106 and 0.121, but failed to reach significance in the more economically developed eastern region. Conversely, internet access was positively associated with depressive symptoms in both the western and eastern zones, while device diversity exhibited opposing associations, correlating positively with depressive feelings in the west but negatively in central China.

Investigation of theoretical boundary conditions indicated that cognitive capacities and material resources moderate these empirical relationships in unexpected ways. Standardized cognitive test scores significantly enhanced the positive association between internet use and self-rated health, showing an interaction term of 0.044; predictive margins indicated that the difference in reporting excellent health between users and non-users rose from 1.438 percentage points at the twenty-fifth cognitive percentile to 2.150 percentage points at the seventy-fifth percentile. Conversely, higher formal education attenuated the positive association with self-rated health while amplifying depressive symptom risks. Economic standing exhibited comparable dampening patterns: personal income and per capita household income both negatively moderated the association between internet use and self-rated health, yielding interaction coefficients of negative 0.007 and negative 0.076.

Physical functioning and community engagement further conditioned the consequences of technology adoption. Greater limitations in activities of daily living significantly reduced the positive association of internet use with self-rated health, with an interaction coefficient of negative 0.015, while also attenuating the positive link to depressive symptoms. Active participation in formal social organizations exerted the opposite effect, reinforcing the positive relationship between connectivity and physical health evaluations, with a significant interaction coefficient of 0.098. The robustness of the core baseline associations was substantiated across multiple sensitivity procedures, including trimming extreme age brackets between twenty-two and ninety years, controlling for additional confounders such as medical spending and physical exercise, and applying propensity score matching. Under nearest-neighbor matching, the average treatment effect on the treated remained positive and significant at 0.210 for self-rated health and 0.090 for depressive symptoms, confirming that observed demographic imbalances do not drive the baseline findings.

Despite these extensive econometric controls, several limitations contextualize the scope of these findings. Because the China Family Panel Studies relies on broad observational questionnaires, it lacks granular indicators to measure advanced digital literacy competencies, such as critical information appraisal or algorithmic safety skills. Additionally, while individual fixed effects absorb time-invariant unobserved characteristics and matching algorithms minimize observed selection bias, observational panel models cannot completely eliminate the risk of reverse causality or time-varying confounding. For instance, individuals developing chronic illnesses or psychological distress might independently alter their screen time or online communication habits. Consequently, the authors caution against interpreting the observed coefficients as definitive causal mechanisms. Instead, the empirical evidence demonstrates that digital transformation functions simultaneously as an empowering resource and a behavioral burden, indicating that public health policies must advance beyond promoting sheer technical connectivity to cultivate balanced, context-aware digital engagement across diverse demographic populations.

Subject of Research: Multidimensional associations of digital technology use with self-rated health and depressive symptoms

Article Title: Digital empowerment or digital burden? Heterogeneous associations between digital technology use and health outcomes

Article References: Zhang, K., Hao, Y., & Wang, W. (2026). Digital empowerment or digital burden? Heterogeneous associations between digital technology use and health outcomes. Frontiers in Public Health, 14, Article 1883525. https://doi.org/10.3389/fpubh.2026.1883525

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1883525

Keywords: digital technology use, self-rated health, depressive symptoms, China Family Panel Studies, fixed-effects ordered logit, blow-up and cluster estimator, usage duration, device diversity, online activity breadth, digital empowerment, digital burden, cognitive capacity, economic resources, social embeddedness, rural-urban disparities, regional health variation, Andersen behavioral model, panel data econometrics