How health literacy and eHealth literacy influence quality of life in older adults with chronic diseases: a network and path analysis Authors: Hao Zou, Shenglan Zhou, Li Jin, Lijun Xie, Yuchen Wang, Haoyang Shi, Linlin Zhang, Hong Li Published: 2026 ABSTRACT AbstractBackgroundDespite existing research on factors influencing quality of life (QOL) in older adults with chronic diseases, the underlying mechanisms by which health literacy (HL) and e-health literacy (eHL) contribute to QOL improvement remain underexplored. This study aimed to elucidate the complex relationships among HL, eHL, health behaviors, psychosocial factors, and QOL, and to identify the driving pathways of HL and eHL in enhancing QOL.MethodsA cross-sectional study was conducted between March and June 2025 at a tertiary hospital in Xi’an, China. A total of 304 older adults with chronic diseases participated in the study. Participants completed assessments for HL, eHL, cognitive function, frailty, nutrition, physical activity, sleep quality, family function, depression, and QOL (including Physical and Mental Component Summaries, PCS and MCS). Grip strength was also measured. Multiple linear regression, network analysis, and path analysis were employed to determine influencing factors and structural relationships.ResultsHigher HL and eHL levels were associated with younger age, higher socioeconomic status, longer daily smartphone usage, and having personal interests; HL was additionally linked to better family function. Network and path analyses revealed that depression, frailty, physical activity, grip strength, and family function were primary direct predictors of QOL, PCS and MCS. Depression and frailty were identified as key risk factors, while sleep quality and nutritional status served as significant mediators. Although HL and eHL did not directly influence QOL, they functioned as upstream variables that indirectly improved QOL by positively influencing these intermediary health behaviors and psychosocial factors.ConclusionsHL and eHL indirectly enhance QOL in older adults with chronic diseases by driving improvements in health behaviors and psychological status. Interventions should target depression, frailty, sleep quality, and nutrition as critical modifiable factors. Future programs aiming to improve HL and eHL should prioritize older individuals with low socioeconomic status and limited digital experience, incorporating age-friendly designs, social interaction, and family involvement.Keywords: Health literacy, E-health literacy, Quality of life, Chronic diseases Introduction Globally, approximately 80% of older adults suffer from at least one chronic condition, and nearly half experience multimorbidity [1]. Quality of life (QOL) is a crucial indicator for assessing the burden of chronic diseases [2]. Unlike traditional physiological metrics (e.g., blood glucose, blood pressure), QOL focuses on patients’ subjective health experiences, encompassing physical functioning, psychological state, and social relationships [3]. This subjective perception often better reflects the real impact of disease on daily life than objective measures [4, 5]. Poor QOL frequently indicates difficulties in disease management and self-care [6, 7]. Effective management of chronic diseases requires deep patient engagement, which relies on health literacy (HL) — defined as the ability of individuals to access, understand, evaluate, and apply health information to make health decisions [8]. Studies report that patients with higher health literacy exhibit better self-care behaviors and healthier lifestyles, which are associated with improved health outcomes [9, 10]. With the development of the internet and mobile devices, health literacy has extended to include e-health literacy (eHL), which assesses an individual’s ability to access, comprehend, evaluate, and apply health information within digital environments [11]. eHL has been found to be significantly associated with cognitive health in older adults, yet most older adults lack adequate eHL [12]. Both HL and eHL represent essential skills necessary for patients to understand health issues, make informed decisions, and take appropriate actions. Previous research has yielded growing evidence on predictors of QOL in older adults. Identified predictors primarily focus on physiological psychological and behavioral levels, including demographic characteristics, symptom indicators, and lifestyle factors [13–15]. These findings provide a crucial foundation for developing targeted interventions and policies. However, our understanding of the interaction mechanisms among these factors remains limited, particularly concerning the relationships between HL/eHL, these factors, and QOL. Nevertheless, one can envision such a path: higher HL is associated with better QOL because patients with greater HL tend to exhibit healthier behaviors. This could trigger a cascade of effects promoting physical, psychological, and cognitive health. However, this hypothesis requires empirical validation in older adults with chronic conditions. Furthermore, conventional prediction models typically treat variables as independent, additive predictors [16]. Given that QOL is a multidimensional concept involving physical, social, and psychological domains—factors that often interact complexly [17]—previous models, constrained by their statistical frameworks, struggle to capture intricate network relationships among variables [18–20]. Consequently, it remains unclear which variables possess key centrality in linking health literacy to QOL, nor are the structural relationships through which they reciprocally amplify or buffer one another fully elucidated. Baker proposed a conceptual model illustrating the cascading causal processes through which HL influences health outcomes. This model posits health behaviors and self-efficacy as key mediators between HL and health status [21]. The relationship between HL and self-efficacy has been validated in our prior work [22]. In addition to Baker’ s work, other pathway-oriented frameworks have been proposed to explain how HL translates into health outcomes. For example, Paasche-Orlow and Wolf described three major pathways linking HL to health outcomes: navigation of the healthcare system, patient-provider interaction, and self-care [23]. Similarly, von Wagner et al. suggested that HL influences health outcomes through health actions driven by motivational and volitional determinants [24]. These frameworks collectively indicate that the effects of HL on health outcomes do not occur in a vacuum; rather, they operate through intermediate mechanisms embedded within broader behavioral and social contexts. Consistent with this perspective, the health ecology model is widely utilized in chronic disease prevention and health promotion research. It emphasizes that individual and population health are shaped by a combination of personal traits, interpersonal networks, and social environments, as well as the dynamic interactions among these factors [25]. Building upon these perspectives, we constructed the logical framework for this study. We hypothesize that individuals with varying levels of HL/eHL will exhibit differences in health behaviors (e.g., nutrition, sleep quality, and physical activity) and interpersonal environments (e.g., family function), ultimately leading to divergent health outcomes, including frailty, cognitive decline, depressive symptoms, and diminished QOL (Fig. 1). These variables were strategically selected because previous literature has reported their individual associations with QOL, and they represent core factors amenable to intervention that are readily assessable in a clinical hospital setting [15, 16, 26–28]. We acknowledge that not all potentially relevant variables were included in the present model. In particular, factors such as healthcare-system navigation, patient-provider interaction, and broader social participation or social support beyond the family unit were not incorporated, in order to maintain a parsimonious framework centered on the core pathways amenable to intervention most relevant to our research objective. Given the likely interaction among these selected variables, we first used network analysis to characterize their structural relationships and identify central variables on the basis of strength centrality. Because network analysis primarily captures non-directional associations, we subsequently used path analysis, informed by the network findings, to further explore the potential directional relationships and mediating pathways linking HL/eHL to QOL. Together, these complementary analyses were intended to clarify the multi-level mechanisms through which HL/eHL may contribute to QOL among older adults with chronic diseases. Theoretical hypothetical model. Health literacy and eHealth literacy may influence quality of life through health behaviors, interpersonal environment, and subsequent health outcomes Against this background, the present study differs from previous research in two respects. First, rather than focusing on patients with a single specific disease, we examined older adults with chronic diseases, including those with multimorbidity, a more common clinical context in later life. Second, instead of examining isolated predictors or selected mediators in a fragmented manner, we adopted a more integrated analytic framework in which network analysis and path analysis served complementary roles [29]. Accordingly, this study aimed to identify central variables within the HL/eHL–QOL system and to systematically explore the potential pathways through which HL/eHL may be linked to QOL in older adults with chronic diseases. Methods Study design and sample A cross-sectional study was conducted between March and June 2025 at a tertiary Grade A general hospital in Xi’an, China. The hospital primarily serves patients from Baqiao District and the neighboring Lantian County. Its patient population mainly comprises individuals from low- to middle-income households, including a large proportion of retired adults, migrant workers, and rural residents. In terms of healthcare financing, most patients were covered by basic medical insurance, including the urban employee and urban-rural resident insurance schemes, whereas a small proportion paid out of pocket. Eligible participants were adults aged 60 years or older who had been diagnosed with at least one chronic disease in a stable condition and were able to provide informed consent voluntarily. Patients were excluded if they had psychiatric disorders or cognitive impairment that precluded completion of the questionnaires. The minimum sample size in network analysis can be calculated using the formula N + N*(N-1)/2 [30]. In this study, a total of 11 nodes were involved (N = 11). Considering the minimum sample size requirement for structural equation modeling (N ≥ 200) [31] and allowing for a 20% inefficiency rate, a final sample of 304 participants was included. Instrument A self-designed questionnaire was developed to collect general information, including demographic characteristics, personal interests (defined as regular hobbies or leisure activities such as reading, music, exercise, or gardening), smartphone usage, chronic disease status, and handgrip strength (measured on the participant’s dominant hand using a handgrip dynamometer; three consecutive measurements were taken, and the maximum value was recorded). Socioeconomic status (SES) was derived from three indicators: economic status, educational attainment, and occupational rank. Specifically, economic status was assessed using a self-reported single item asking participants to compare their living standard to the local average on a 5-point scale, scored from 1 (very difficult) to 5 (very affluent). Educational attainment was categorized and scored from 1 to 5: illiterate (1), primary school (2), middle school (3), high school (4), and university (5). Occupational rank was classified and scored from 1 to 3 based on skill and administrative level: (1) homemakers, farmers, and unemployed individuals; (2) general staff, service workers, and industrial workers; and (3) professionals and technicians, doctors, teachers, administrators, and military personnel. These three indicators were then integrated using principal component analysis (PCA). Based on the resulting PCA scores and their distribution percentiles (P33 and P67), participants were stratified into high, moderate, and low SES groups. Details of other scales employed in this study are presented in Table 1. Table 1The scales used in this studyScaleDimensionsItemsScoring methodCronbach’α in this studyAPGAR [32]Adaptation, Partnership, Growth, Affection, Resolve.5Total score 0–10; higher scores indicate better family function level, reflecting greater support, cooperation, emotional connection, and problem-solving ability among family members.0.93eHEALS [12, 33]application ability, evaluation ability, decision-making ability.85-point Likert scale (8–40); higher score indicate higher eHealth literacy.0.94FRAIL [34]fatigue, resistance, ambulation, illnesses, and weight loss.5“Yes” = 1 point, “No” = 0; total 0 = non-frail, 1–2 = pre-frail, ≥ 3 = frail, higher score indicate higher the risk of frailty.0.64HeLMS [35]information acquisition ability, communication interaction ability, health improvement willingness, economic support willingness.245-point Likert scale (24–120); higher score indicate higher health literacy.0.87MMSE [36]orientation, registration, attention and calculation, recall, and language.30Each correct answer = 1 point (0–30); higher score indicate better cognitive function.0.63MNA [37]Anthropometry, general assessment, dietary habits, self-assessment.18Total score 0–30; higher scores indicate better nutritional status.0.60PASE [38]Leisure-time, household, work-related physical activity.12Weighted total score (0−500); higher scores indicate higher physical activity0.65PHQ−9 [39]Depressive symptoms.9Total score 0–27; higher scores indicate more severe depressive symptoms.0.83PSQI [40]Subjective sleep quality, latency, duration, efficiency, disturbances, medication use, daytime dysfunction.19Total score 0–21; higher scores indicate more severe sleep problems.0.70SF−12 [41]Physical Component Summary (PCS): Physical function, role physical, bodily pain, general health, Mental Component Summary (MCS): vitality, social function, role emotion, mental health.12PCS and MCS scores both range from 0 to 100, with higher scores indicating better quality of life.0.92Note: APGAR: Family APGAR Index, eHEALS: eHealth Literacy Scale, FRAIL: Frail scale, HeLMS: Health Literacy Management Scale, MMSE: Mini-Mental State Examination, MNA: Mini Nutritional Assessment, PASE: Physical Activity Scale for the Elderly, PHQ−9: Patient Health Questionnaire−9, PSQI: Pittsburgh Sleep Quality Index, SF−12: Short Form 12-Item Health Survey. This study used validated Chinese versions of all the above scales The scales used in this study Note: APGAR: Family APGAR Index, eHEALS: eHealth Literacy Scale, FRAIL: Frail scale, HeLMS: Health Literacy Management Scale, MMSE: Mini-Mental State Examination, MNA: Mini Nutritional Assessment, PASE: Physical Activity Scale for the Elderly, PHQ−9: Patient Health Questionnaire−9, PSQI: Pittsburgh Sleep Quality Index, SF−12: Short Form 12-Item Health Survey. This study used validated Chinese versions of all the above scales Data collection Data were collected through face-to-face interviews using a structured questionnaire. The data collection team consisted of one general practitioner and three undergraduate nursing students. Before the formal survey began, all interviewers completed a standardized 4-hour training session covering interview scripts, questionnaire completion procedures, and ethical principles. Interviews were conducted in inpatient wards, and each questionnaire took approximately 30–40 min to complete. To ensure data quality and methodological rigor, the entire data collection process was supervised by a researcher with a master’s degree in nursing and extensive experience in independently conducting survey research. The supervisor reviewed all completed questionnaires daily for completeness and adherence to study procedures. Therefore, there were no missing data in this study. Data analysis Data analysis was performed using SPSS 26.0 (IBM Corp., Armonk, NY, USA), AMOS 29.0 (IBM Corp., Armonk, NY, USA), and R 4.5.1 (R Foundation for Statistical Computing, Vienna, Austria). Both HL and eHL scores exhibited skewness and kurtosis values within ± 1, indicating approximately normal distributions. Categorical variables were summarized as frequencies and percentages, while continuous variables not following a normal distribution were presented as medians and quartile ranges (P25, P75). First, we employed independent samples t-tests, one-way analysis of variance (ANOVA), Spearman correlation analysis, and multiple linear regression analysis in SPSS to explore the factors associated with HL and eHL. Second, we utilized the qgraph and bootnet packages in R to construct a Gaussian Graphical Model (GGM) for the pre-specified variables included in the network model, which were selected a priori based on the study’s logical framework and prior empirical findings. Partial correlation networks were estimated using the Extended Bayesian Information Criterion graphical least absolute shrinkage and selection operator (EBICglasso) algorithm. Centrality metrics—including strength, closeness, betweenness, and expected influence—were calculated using the centralityPlot function. To assess the stability of the network structure, we performed nonparametric bootstrap analyses for edge weight accuracy and case-dropping bootstrap analyses for the robustness of centrality indices. Correlation stability coefficients (CS-coefficients) were computed to evaluate the reliability of the centrality metrics [42]. Finally, based on the network analysis results and logical framework, path analysis was conducted in AMOS to construct the initial structural model. The model was revised during the fitting process to meet the commonly used fit indices such as x2/df < 3, GFI, AGFI, RFI, IFI, CFI, TLI ≥ 0.9, and RMSEA ≤ 0.06 [43]. The visualization graphics were completed by GraphPad Prism 10 (GraphPad Software, LLC, Boston, MA, USA), R 4.5.1 (R Foundation for Statistical Computing, Vienna, Austria) and Visio (Microsoft Corporation, Redmond, WA, USA). A two-tailed p-value < 0.05 was considered statistically significant in this study. Results Sociodemographic characteristics The median age of participants in this study was 70 years (66, 75). Among them, 55.3% were male and 44.7% were female. The majority were married (81.9%) and residing in rural areas (52.0%). A total of 58.6% of participants were classified as having low socioeconomic status. Regarding smartphone use, 38.8% of older adults reported using a smartphone for 2–4 h per day, while 21.1% reported usage of ≥ 4 h per day. Median scores were 97 (85, 105) for HL and 19 (8, 29) for eHL. The overall median SF-12 score was 100 (89.3, 106.3), with Physical Component Summary (PCS) and Mental Component Summary (MCS) medians of 43 (32.3, 50.8) and 57 (48.7, 62.7), respectively. Additional sociodemographic and health-related characteristics are detailed in Table 2. Table 2Sociodemographic characteristics (N = 304)VariablesCharacteristicsParticipants N (%) / ±SD / M (P25, P75)GenderFemale136 (44.7%)Male168 (55.3%)SESLow178 (58.6%)Middle85 (28.0%)High41 (13.5%)Marital statusMarried249 (81.9%)Unmarried/Divorced/Widowed55 (18.1%)Place of residenceRural areas158 (52.0%)Urban areas146 (48.0%)Daily Smartphone Usage Duration≤ 1 h72 (23.7%)1–2 h50 (16.4%)2–4 h118 (38.8%)≥ 4 h64 (21.1%)InterestYes130 (42.8%)No174 (57.2%)FRNon-frail154 (50.7%)Pre-frail107 (35.2)Frail43 (14.1%)Frailty Score0 (0, 2)No. of Chronic Diseases5 (4, 6)Age70 (66, 75)HL94.8 ± 13.7eHL19.7 ± 10.2PR10 (8, 10)CF26 (23.25, 26)MNA24 (21.5, 25.5)PA93 (52.2, 122.7)SQ7 (5, 10.8)DP3 (2, 6)GS20.2 (14.2, 27.0)QOL100 (89.3, 106.3)PCS43 (32.3, 50.8)MCS57 (48.7, 62.7)Note: SES: Socioeconomic status, HL: health literacy, eHL: e-health literacy, CF: cognitive function, PR: APGAR (family function), PA: physical activity, SQ: sleep quality, DP: depression, MNA: Mini Nutritional Assessment (nutrition status), FR: frail, GS: Grip strength, QOL: quality of life, PCS: Physical Component Summary, MCS: Mental Component Summary Sociodemographic characteristics (N = 304) Note: SES: Socioeconomic status, HL: health literacy, eHL: e-health literacy, CF: cognitive function, PR: APGAR (family function), PA: physical activity, SQ: sleep quality, DP: depression, MNA: Mini Nutritional Assessment (nutrition status), FR: frail, GS: Grip strength, QOL: quality of life, PCS: Physical Component Summary, MCS: Mental Component Summary Univariate analysis and multiple linear regression analysis As shown in Figure. 2, differences in both HL and eHL were observed across various demographic characteristics. Participants who were married (P < 0.01), reported longer daily smartphone use (P < 0.001), had personal interests (P < 0.001), and had higher SES (P < 0.001) exhibited significantly higher levels of both HL and eHL. Regarding gender, a significant difference was found only in HL, with females scoring lower than males (P < 0.05). In contrast, place of residence was significantly associated with eHL, with rural residents showing lower eHL scores than urban residents (P < 0.05), while no significant association was observed between residence and HL. The differences in HL and eHL among different demographic groups. a: Univariate analysis of HL across demographic characteristics, b: Univariate analysis of eHL across demographic characteristics. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 As shown in Figure. 3, age was negatively correlated with both HL (rs = -0.21, P < 0.01) and eHL (rs =-0.22, P < 0.01). Notably, age demonstrated a stronger negative correlation with the information acquisition dimension of health literacy (rs = -0.31, P < 0.01), while no significant associations were found between age and the other HL dimensions. In contrast, age was negatively associated with all dimensions of eHL (P < 0.05). In addition, family function was positively correlated with overall HL (rs = 0.26, P < 0.01), as well as with each of its individual dimensions (P < 0.05). However, no significant associations were observed between family support and eHL. Spearman analysis. ×: Not significant, IAA, CIA, HIW, ESW are the four dimensions of HL, while AA, EA, DA are the three dimensions of eHL. Circles of different colors and sizes represent the magnitude of the correlation coefficients; red indicates positive correlations, while blue indicates negative correlations Taking HL and eHL as dependent variables, and the variables with statistical significance in the univariate analysis as independent variables, a multiple linear regression analysis was conducted. The results (Fig. 4a) showed that family function, age, interest, daily smartphone usage duration, and SES were the influencing factors of HL. Among them, patients without interests had significantly lower HL than those with interests (P < 0.001); patients with daily smartphone usage durations of 1–2 h (P < 0.01), 2–4 h (P < 0.01), and ≥ 4 h (P < 0.001) had significantly higher HL than those with usage durations of ≤ 1 h; patients with high SES had significantly higher HL than those with low SES (P < 0.01); In addition, higher family function was associated with better HL (P < 0.01), while older age was associated with lower HL (P < 0.01). Multivariate Linear Regression Forest Plot, a: Multivariate linear regression forest plot of HL influencing factors, b: Multivariate linear regression forest plot of eHL influencing factors For eHL, significant influencing factors included age (P < 0.01), interests (P < 0.05), daily smartphone usage duration (P < 0.05), and SES (P < 0.05), demonstrating trends similar to those observed for HL. Moreover, participants with moderate SES also had significantly higher eHL than those with low SES. Figure 4b. Network analysis Figure 5 shows that in any of the networks, there are strong edges connections between HL, eHL and cognitive function, as well as between sleep quality and depression. HL, eHL, family function and cognitive function form a cognitive health network; sleep quality, depression, nutrition, frailty, physical activity and QOL form a somatic-mental health network. All three networks consisted of 11 nodes. In the QOL network, the network density was 0.71 (39 / 55), with non-zero edge weights ranging from (-0.41-0.45). As shown in Fig. 5a1, there were strong edge connections between depression, frail, and QOL (-0.41, -0.31). Both frail and depression also showed notable edge connections with nutritional status (-0.23, -0.16), while depression was strongly linked to sleep quality (0.45). These findings suggest that depression and frail serve as key mediators linking other variables to QOL within the network. As shown in Fig. 5a2, depression, frail, and HL exhibited high strength centrality values (1.27, 0.90, and 0.99, respectively). Moreover, HL and eHL demonstrated strong positive expected influence within the network (0.81, 0.60), followed by cognitive function (0.42). In contrast, frail and depression showed strong negative expected influence (-0.84, -0.31). The network structure, node centrality, and strength centrality stability. a1, b1, c1: Estimated network structures for QOL, PCS, and MCS, respectively. Edge thickness reflects the strength of associations, with blue lines representing positive correlations and red lines representing negative correlations. a2, b2, c2: Centrality indices (strength, closeness, betweenness, and expected influence) for each node in the QOL, PCS, and MCS networks. a3, b3, c3: Correlation stability (CS) analysis for node strength in the QOL, PCS, and MCS networks, showing average correlation with the original sample across case-dropping bootstraps In the PCS network, the network density was 0.74 (41 / 55), with non-zero edge weights ranging from (-0.37-0.53). As shown in Fig. 5b1, frail and physical activity exhibited strong edge connections with PCS (-0.37, 0.17). In addition, physical activity and frail were also strongly connected to nutritional status (0.15, -0.23). Similar to the QOL network, depression, frail, and HL showed high strength centrality in the PCS network (1.17, 1.04, and 1.00, respectively). HL and eHL demonstrated strong positive expected influence (0.85, 0.63), whereas frail and family function showed strong negative expected influence (-0.60, -0.32) within the network. In the MCS network, the network density was consistent with that of the QOL network, with non-zero edge weights ranging from (-0.44-0.44). As shown in Fig. 5c1, depression, family function, and sleep quality exhibited edge connections with MCS (-0.44, 0.11, and − 0.10, respectively). Additionally, frail, nutritional status, and family function showed strong edge connections with depression (0.18, -0.16, and − 0.12, respectively), suggesting that depression serves as a key mediating node within this network. Figure 5c2 shows that depression, HL, and cognitive function had high strength centrality (1.41, 0.96, and 0.91, respectively). Frail and depression demonstrated strong negative expected influence (-0.46 and − 0.17), while HL and eHL exhibited strong positive expected influence (0.76, 0.59). Network stability was assessed by calculating the correlation stability (CS) coefficient for strength centrality. The CS coefficients for the three networks were 0.60, 0.52, and 0.60, respectively, all above the recommended threshold of 0.5 [44], indicating that the centrality estimates were robust across subsamples Fig. 5a3, b3, c3. Path analysis Figure 6a1 showed that depression had the strongest direct effect on QOL (β= -0.52), followed by frail (β= -0.29). Physical activity contributed a positive direct effect of 0.11. Therefore, depression, frail, and physical activity were identified as the main direct predictors of overall QOL (R²= 0.53). Regarding total effects size, depression exerted the strongest negative impact on quality of life (ES= -0.59), followed by sleep quality (ES= -0.38) and frail (ES= -0.29). In contrast, HL showed the strongest positive total effect on QOL (ES = 0.25), followed by physical activity (ES = 0.26) and nutritional status (ES = 0.23) (Fig. 6a2). As shown in Fig. 6b1, frail had the strongest direct negative effect on the PCS (β= -0.44). Physical activity and grip strength showed direct positive effects on PCS, with coefficients of 0.17 and 0.12, respectively. Regarding total effects, frail and depression had the strongest negative impacts on PCS (ES= -0.44 and − 0.18), while physical activity, HL, and nutritional status had the strongest positive effects (ES = 0.30, 0.17, and 0.15, respectively) (Fig. 6b2). Figure 6c1 showed that depression had a strong direct negative effect on the MCS (β= -0.61), whereas family function exhibited a direct positive effect (β = 0.10). In terms of total effects, depression remained the strongest negative impact on MCS (ES= -0.61), followed by sleep quality (ES= -0.32). Family function and nutritional status had relatively strong positive effects on MCS (ES = 0.28, 0.26), followed by health literacy (ES = 0.14) (Fig. 6c2). The model fit indices for all three path models met the required statistical thresholds (detailed fit indices including x2/df, GFI, AGFI, RFI, IFI, CFI, TLI, and RMSEA are presented within Fig. 6). Potential pathways linking HL and eHL to QOL. a1, b1, c1: Path analysis of HL/eHL on QOL, PCS, and MCS, a2, b2, c2: Standardized direct, indirect, and total effects of each variable on QOL components. The numbers on the arrows represent standardized path coefficients. Blue arrows indicate positive effects, while red arrows indicate negative effects. Solid lines denote statistically significant paths, and dashed lines represent non-significant paths. The width of the arrows reflects the magnitude of the effects. *P < 0.05, **P < 0.01, ***P < 0.001 Discussion Our model results indicate that depression, frailty, physical activity, grip strength, and family function are the primary direct predictors of overall QOL, PCS, and MCS. HL and eHL serve as key driving factors that indirectly improve QOL/PCS/MCS by positively influencing various health behaviors and psychological factors, consistent with our initial hypotheses. Notably, depression and frailty exhibit high strength centrality and strong negative total effects across all three network and path models, identifying them as key risk factors impacting patients’ QOL. Depression and frailty: key risk factors Depression exhibited the highest strength centrality in the network, followed by frailty, with both showing negative effects, indicating that they are the most influential risk factors within the network structure. Patients with chronic diseases are at an increased risk of developing depression due to persistent physical symptoms (e.g., pain, functional limitations) and adverse effects of medications [45–47]. In the present study, 36% of participants reported depressive symptoms (PHQ-9 ≥ 5) [48], which was higher than in previous national surveys (the discrepancy might be attributable to research tools and population characteristics) [49]. Patients with depression often experience persistent low mood and diminished interest in daily activities. Such emotional distress impairs their ability to experience pleasure and satisfaction, leading to a marked reduction in subjective well-being [50, 51]. Depressive individuals may also withdraw from social interactions, exhibit reduced engagement with family members, and experience intensified feelings of isolation—all of which contribute to a lower QOL [52, 53]. Furthermore, depressive symptoms may exacerbate the severity of chronic diseases, increase treatment challenges, and ultimately compromise QOL [54]. Our findings align with a previous network analysis focusing on quality of life in patients with diabetes, which also identified depression as a central variable highly correlated with poorer QOL [55]. Compared to that study, our research extends the literature by examining a broader population of older adults with chronic diseases (including those with multimorbidity) and by explicitly integrating HL and eHL into the QOL network alongside a comprehensive set of behavioral and psychosocial factors. In this study, 49% of participants were identified as either pre-frail or frail (Frail ≥ 1). Frailty increases vulnerability to stressors, thereby elevating the risk of falls, disability, and long-term care dependency [56]. It severely limits an individual’s physical, physiological, and social functioning, further deteriorating QOL [57, 58]. Depression and frailty frequently co-occur and may reinforce each other, thereby intensifying the overall disease burden [59]. Consistent with prior research, 27% of the participants in our study presented with both depression symptom and pre-frailty/frailty. These individuals reported significantly lower QOL compared to those without such comorbidity. Path analyses of PCS and MCS also suggested that depression and frailty can interact with each other, leading older adults patients into a predicament of dual deterioration in both physical and mental health. Therefore, the key to improving the QOL of patients lies in early identification and interruption of the vicious cycle between depression and frailty. HL and eHL: cognitive empowerment drivers HL and eHL demonstrated strong positive influence within the network, indicating that they act as active “empowering agents” that positively drive other nodes. Across all three networks, HL, eHL, and cognitive function formed a stable cognitive health cluster. Consistent with previous findings, eHL was associated with higher levels of cognitive function [12]. Individuals with high eHL are more likely to actively search for, review and apply health information from digital channels such as the Internet [60]. During this process, they continuously activate cognitive functions such as learning memory, executive function, verbal organization strategies, attention, and decision-making [61, 62]. This continuous cognitive engagement is similar to “exercise” for the brain, which helps to enhance or maintain cognitive reserves and thereby promotes the sustained improvement of cognitive functions [62, 63]. HL was positively associated with both eHL and cognitive function. This finding is conceptually plausible, because eHL can be understood as an extension of HL into digital contexts through the incorporation of digital skills [64]. Therefore, people with higher HL are more likely to master and utilize online health resources, thereby improving their cognitive performance. Importantly, the application of HL itself also depends on a collaboration of varying cognitive processes, including memory, numeracy and executive functions. For example, calculating calorie intake, remembering the foods that cause allergies, organizing and dosing of medication, etc. [65]. Furthermore, the decline in cognitive functions may also cause the older adults to feel ashamed and embarrassed, thereby reducing interpersonal communication and leading to lower HL [66]. Although cognitive function was not directly associated with QOL in this study, it may exert an indirect influence through frailty and depression. Cognitive impairment is common among older adults and easily leads to difficulties in daily activities [67, 68]. Lower cognitive function may also increase the risk of psychiatric disorders [69]or dementia [70], thereby negatively affecting QOL. Moreover, a bidirectional relationship exists between cognitive decline and frailty, forming a vicious cycle that exacerbates adverse health outcomes among older adults with chronic conditions [71, 72]. Taken together, HL, eHL, and cognitive function appear to reinforce each other and collectively form a cognitive-empowerment support system. This system may help disrupt the negative cycle of cognitive decline and frailty, ultimately improving QOL. Path analysis revealed that, in addition to its effects on eHL and cognitive function, HL had significant positive effect on physical activity, grip strength, and family function. A meta-analysis indicated that older adults with adequate HL were approximately 61% more likely to engage in physical activity at least five days per week than those with inadequate HL [73]. As conceptualized by Nutbeam, HL comprises three domains—functional, communicative, and critical literacy [74] (all of which are captured by the HL instrument used in this study); individuals with high HL are better able to obtain, understand, and apply health information, which not only strengthens their sense of control over health decisions but also, through the development of communicative and critical literacy, enhances their motivation, confidence, and critical thinking skills, enabling them to make informed choices, adapt their lifestyle, and sustain health-promoting behaviors such as regular physical activity [75]. Regular physical activity, in turn, increases the demand for energy and nutrients [76], which, with adequate dietary intake, may enhance nutritional status, supporting muscle protein synthesis and functional maintenance [77]. This process strengthens muscle, as evidenced by higher grip strength levels [78], reduces frailty risk [79], and thereby improves patients’ QOL. The communicative literacy of HL emphasizes social skills [74]. Due to aging and chronic disease-related limitations, older adults often experience a shrinking social network, making family the primary source of emotional and practical support [22, 80]. Higher HL is associated with stronger communication skills, which may foster more effective family health communication. This can include receiving advice, sharing health-related experiences, and collaboratively solving health-related problems—all of which enhance perceived familial support [81, 82]. Moreover, effective family function has been shown to significantly improve emotional well-being, reduce the occurrence of diseases, and enhance health outcomes [83]. This was verified in the MCS pathway, where greater family function was associated with lower depression, thereby indirectly promoting MCS. Sleep quality and nutritional status: critical mediating factor Across all three path models, sleep quality consistently showed a significant negative effect on depression, indicating that the poorer the sleep quality, the more obvious the depressive symptoms—a finding corroborated by our previous research [84]. Mechanistically, sleep disturbances may contribute to depression by activating the hypothalamic-pituitary-adrenal (HPA) axis, elevating proinflammatory cytokine levels (e.g., IL-6 and TNF), disrupting monoamine neurotransmitter systems (e.g., serotonin and dopamine), and perturbing circadian rhythm gene expression, thereby impairing brain emotion regulation and increasing depressive symptoms [85, 86]. In both the overall QOL and MCS pathways, sleep quality contributed a relatively high negative total and mediating effects, suggesting that it serves as a key mediator linking various upstream factors to psychological well-being and overall QOL. Notably, findings from the MCS path, HL can influence sleep quality by enhancing family function. On the one hand, it is because good family function can actively facilitate the older adults to share their emotions and pains, and obtain emotional support through effective communication with family members to improve sleep quality [83]. On the other hand a good family function helps maintain family relationships, provide life support and create a comfortable internal and external environment for better sleep quality [87]. This pathway highlights how HL, particularly its communicative dimensions, can be translated into concrete, supportive interpersonal environments. These environments, in turn, facilitate healthier sleep behaviors, mitigate depressive symptoms, and ultimately improve MCS. Another noteworthy finding is that nutritional status consistently contributed a relatively high positive total and mediating effect across all three path models. Clearly, in addition to its associations with physical activity and frailty, HL can directly influence nutritional status, thereby alleviating depression and enhancing QOL. Previous studies have reported that higher levels of literacy, numeracy, and HL are associated with the use of food labels and good portion-size estimation skills. This may promote better dietary behaviors, such as choosing nutrient-rich foods, leading to a more balanced overall nutrient intake and thereby improving dietary quality [88–90]. HL also plays a signiicant role in improving dietary adherence, helping patients with chronic diseases follow high-quality dietary patterns and reducing the risks of disease deterioration and death [91]. Improved nutritional levels can also reduce the risk of depression through various biological pathways, such as reducing chronic inflammation, promoting beneficial gut bacteria, and enhancing neuroprotection [92–94]. In conclusion, HL and eHL indirectly improve the QOL through a series of intermediate factors such as sleep quality, nutritional status, physical activity, depression, family function, and so on, rather than acting directly. On the one hand, it might be that these mediating variables are themselves strongly correlated with and explanatory for QOL, thereby absorbing the effect of HL in the model and attenuating or even masking its direct impact on QOL. On the other hand, this is also in line with the theoretical mechanism: HL primarily acts as a health resource or capability, exerting its influence on QOL gradually by promoting the adoption of healthy behaviors and improvements in health status, rather than producing an immediate direct effect, highlighting its role as a “driver” in improving QOL. Policy and clinical recommendations Policy recommendations This study further identified several factors associated with HL and eHL, including younger age, higher SES, longer daily smartphone usage, and having personal interests. These findings provide important implications for policy development aimed at improving HL/eHL and, ultimately, QOL among older adults with chronic diseases. First, aging is associated with declines in learning ability and information processing capacity, alongside the presence of a digital divide [95, 96]. Therefore, age-friendly and intuitive digital literacy training programs should be developed to reduce barriers to accessing, understanding, and using health information among older adults. Second, increasing the frequency of digital engagement may help improve HL and eHL. Policymakers and health systems should therefore promote the integration of smart devices with accessible health content, such as personalized disease management apps, WeChat mini-programs, and online consultation platforms, in order to enhance older adults’ familiarity with and trust in digital health resources. Third, because individuals with higher SES tend to have better HL/eHL, reflecting the influence of resource accessibility and educational attainment on health capabilities [97], it is essential to provide more inclusive and equitable health education resources for low-SES populations, with particular attention to vulnerable groups. In addition, older adults with hobbies or personal interests often show better cognitive functioning, psychological well-being, and social interaction. This suggests that community-based programs that encourage participation in meaningful activities and social engagement may serve as effective policy approaches to improving HL/eHL and promoting better QOL. Clinical recommendations Our network and path analyses indicated that depression and frailty are central risk factors for diminished QOL, whereas sleep quality and nutritional status act as crucial mediating pathways. Consequently, clinical interventions to improve QOL should extend beyond merely enhancing HL/eHL to actively target these interconnected, modifiable factors. Specifically, the central positions of depression and frailty in the network highlight them as high-priority targets for early identification and multidisciplinary management. Routine screening for these two conditions, paired with psychoeducation and exercise-based rehabilitation, can help interrupt the cycle of physical and psychological problems, yielding cascading benefits across multiple domains of QOL. Furthermore, as sleep and nutrition serve as key mediators, incorporating sleep hygiene education and individualized dietary counseling is essential to effectively translate HL/eHL improvements into tangible health benefits. Finally, the mutually reinforcing relationship observed between family function and HL suggests that interventions should involve the family unit. Family-based health education and joint digital training sessions can strengthen emotional support and ensure the sustainability of HL/eHL programs, particularly for older adults with limited digital experience or socioeconomic resources. Limitations This study has several limitations. First, due to its cross-sectional design and relatively limited sample size, causal relationships among the variables cannot be conclusively established. Second, all data were collected through questionnaires, which inevitably introduces the potential for self-report bias. Furthermore, because the survey was conducted in inpatient wards, some respondents completed it in the presence of others. This may have compromised privacy and affected the authenticity of responses to sensitive questions, potentially leading to social desirability bias. Third, the data were sourced from a single hospital, which may limit the generalizability and representativeness of the findings due to regional and sample-specific characteristics. Conclusions In conclusion, this study emphasizes that HL and eHL serve as core upstream factors that promote cognitive empowerment and drive health-related behaviors, indirectly enhance QOL by influencing multiple mediating pathways, including depression, frailty, sleep quality, physical activity, nutrition, and family function. Network and path analyses identified several key modifiable targets, such as depression, frailty, nutritional status and sleep quality, offering empirical support for the development of multi-level and coordinated intervention strategies. Acknowledgements We sincerely thank all the participants who took part in this study. An unauthorized version of the Chinese MMSE was used by the study team without permission, however this has now been rectified with PAR. The MMSE is a copyrighted instrument and may not be used or reproduced in whole or in part, in any form or language, or by any means without written permission of PAR (www.parinc.com). Abbreviations Health Literacy Electronic Health Literacy Quality of Life Physical Component Summary Mental Component Summary Cognitive function Family function Physical activity Sleep Quality Depression Mini Nutritional Assessment Grip Strength Author contributions Hao Zou and Hong Li contributed to the conceptualization and methodology of the study. Li Jin, Lijun Xie, Yuchen Wang, and Haoyang Shi were responsible for data curation. Formal analysis and visualization was carried out by Hao Zou, and Shenglan Zhou. Hao Zou drafted the original manuscript, while Linlin Zhang and Hong Li critically reviewed and edited the manuscript. Hong Li provided supervision and served as the corresponding author. Funding acquisition was supported by Hao Zou and Hong Li. Funding This work was funded by the Shaanxi Provincial Natural Science Foundation Program (General Project, Grant No. 2025JC-YBQN-989) and the Fujian Province Science and Technology-Economy Integration Service Platform (Grant No. FJKX-2024XRH04). Data availability The datasets generated or analyzed during this study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study was conducted in accordance with the Declaration of Helsinki and received approval from the Ethics Committee of the Second Affiliated Hospital of Xi’an Medical University (Approval No.: S-X2Y2024-043). Written informed consent was obtained from all participants prior to their inclusion in the study. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets generated or analyzed during this study are available from the corresponding author on reasonable request.