Determinants of Oral Health Literacy Among Older Chinese Adults: A Path Analysis Based on Andersen’s Behavioral Model Authors: Linxin Jiang, Simin Li, Zhihao Liao, Daniel R Reissmann, Gerhard Schmalz, Xiangyu Huang Published: 2026 ABSTRACT PurposeThis study aims to identify the direct and indirect pathways associated with Oral health literacy (OHL) among Chinese older adults and to inform targeted oral health promotion strategies for older adults.Patients and MethodsThis cross-sectional study included 423 community-dwelling adults aged 65–74 years from a nationally representative key population surveillance project in China. Standardized clinical dental examinations and validated questionnaires were conducted. OHL was assessed using oral health belief (OHB) scores (0–4) and oral disease awareness (ODA) scores (0–8). Variables were categorized according to Andersen’s behavioral model into predisposing, enabling/behavioral, and need factors. Poisson regression identified associated factors, followed by path analysis, mediation decomposition, and exploratory multi-group analyses.ResultsMedian OHB and ODA scores were 3 (IQR: 3–4) and 5 (IQR: 3–6), respectively. In the path analysis, gender and education level were indirectly associated with OHL through mediating factors; root decayed and filled teeth (Root DFT) were directly and positively associated with both OHB and ODA; dental prosthesis use was directly associated with higher OHB, whereas dental visit was directly associated with higher ODA. Sweetened beverage consumption ≥ 1/day showed a negative direct association with OHB, although this estimate was based on only 10 participants and should be interpreted with caution. Three total indirect effects and five specific indirect pathways reached (marginal) significance. In exploratory subgroup analyses, the patterns of path coefficients varied descriptively across education and residence subgroups, but no interaction term was statistically significant.ConclusionAmong older Chinese adults, OHL was associated with a combination of predisposing, need, and enabling/behavioral factors. Dental treatment encounters may provide opportunities for reinforcing oral health education. Subgroup-tailored oral health education strategies may be beneficial. Introduction The global population is aging at an unprecedented pace, with older adults representing an increasingly substantial demographic segment worldwide.1 In China, the proportion of individuals aged 65 years and above has risen sharply, and this demographic transition poses significant challenges for geriatric healthcare systems.2 Among the multifaceted health concerns confronting older adults, oral health remains a critical yet frequently underemphasized domain. Epidemiological evidence consistently demonstrates that older adults bear a disproportionately high burden of dental caries, periodontal diseases, and tooth loss, all of which substantially impair quality of life, nutritional status, and systemic health.3,4 Oral health literacy (OHL), defined as the degree to which individuals can obtain, process, and understand basic oral health information needed to make appropriate health decisions,5 has emerged as a key determinant of oral health outcomes.6 Individuals with higher OHL are more likely to engage in preventive behaviors, seek timely dental care, and demonstrate better clinical oral health indicators.7,8 Conversely, low OHL has been associated with poorer self-care practices, delayed treatment-seeking, and worse oral health status, particularly among vulnerable populations such as older adults.9 Despite the recognized importance of OHL, research on its determinants among older populations, especially within the Chinese context, remains limited. The Andersen’s behavioral model of health services utilization provides a well-established theoretical framework for understanding the determinants of health-related outcomes.10 This model posits that health outcomes are shaped by the interplay of three categories of factors: predisposing factors, enabling/behavioral factors, and need factors.11 The Andersen model has been widely applied in oral health research to examine patterns of dental service utilization, oral health behaviors, and health outcomes.12 However, few studies have adopted this framework to systematically investigate the complex associated pathways of OHL. Previous studies on OHL have primarily focused on establishing bivariate associations between sociodemographic variables and OHL levels, without exploring the underlying mechanisms through which these factors operate.9 Sociodemographic characteristics may not only directly affect OHL but also indirectly shape it through their influence on clinical oral conditions and health behaviors. For instance, education level may enhance OHL both directly by improving cognitive capacity for health information processing and indirectly by promoting dental service utilization, which in turn provides exposure to oral health education.13 Similarly, gender and residential location may influence OHL through differential patterns of disease experience and healthcare access.14–16 Understanding these mediating pathways is essential for designing effective, targeted interventions to improve OHL among older adults. Furthermore, the clinical oral conditions such as experience with dental diseases may serve as a catalyst for acquiring oral health knowledge, as patients who have undergone treatment or experienced disease may develop heightened awareness of oral health issues.17 This hypothesis, however, has not been formally tested within a structural framework that simultaneously accounts for the roles of health behaviors and access to care. Additionally, the moderating roles of education level and residence in shaping the pathways to OHL remain largely unexplored. Therefore, the present study aimed to identify associated factors of OHL among Chinese older adults, to construct and validate a path model to delineate the direct and indirect pathways, to decompose the mediation mechanisms of predisposing factors, and to examine whether education level and residential location moderate the identified pathways. The findings aim to provide theoretical and practical foundations for developing targeted oral health promotion strategies for the aging Chinese population. Methods Study Design and Ethical Approval This cross-sectional study was part of a nationally representative key population surveillance project conducted in China in 2021. The study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for cross-sectional studies.18 This study was performed in accordance with the Declaration of Helsinki and was approved by the Stomatological Ethics Committee of the China Oral Health Foundation (permit numbers: 2021–001). All participants provided written informed consent for voluntary participation and data use prior to the survey. Participants and Sampling Procedure The target population comprised adults aged 65–74 years who had been long-term residents (six months or more) of the surveillance sites. Eligible participants were those who had signed informed consent, were willing to participate, and were capable of completing the full clinical dental examinations and questionnaire procedures. Individuals were excluded if they had conditions that could prevent completion of the survey or affect data reliability, including cognitive impairment due to mental illness, acute illness states, oral surgery within the preceding month, unwillingness to participate, or withdrawal during the survey process. A stratified two-stage random sampling procedure was employed (Figure 1). After excluding 6 participants due to missing data, the final analytical sample comprised 423 participants. Missing data were handled by complete-case analysis; the 6 excluded participants represented <2% of the enrolled sample. The required minimum sample size was calculated a priori using G Power (version 3.1.9.7), specifying a medium effect size, a significance level of 0.05, and a power of 0.90. The final sample of 423 participants provided adequate statistical power for Poisson regression analysis. For path analysis, the sample size of 423 exceeded the commonly recommended minimum of 200 for structural equation modeling and provided an adequate participant-to-estimated-parameter ratio.Figure 1Flowchart of the stratified two-stage random sampling procedure.Flowchart of a stratified two-stage random sampling procedure for disease surveillance sites.The flowchart illustrates a stratified two-stage random sampling procedure. Stage 1 involves stratified sampling from existing disease surveillance sites, where 13 sites are randomly selected. Stage 2 involves random sampling, selecting 3 communities per site, resulting in 39 residential communities. Each community enrolls 11 participants, totaling 429 participants. Due to missing data, 6 participants are excluded, leaving a final analytical sample of 423 participants. The process ensures a structured approach to sampling for disease surveillance. Flowchart of the stratified two-stage random sampling procedure. Clinical Dental Examinations Standardized dental examinations were performed by calibrated examiners in accordance with the criteria and methods described in the World Health Organization’s Basic Methods for Oral Health Surveys (5th edition).19 Intra-examiner and inter-examiner reliability were assessed by the Kappa coefficient, which was acceptable (Kappa > 0.8). All permanent teeth except third molars (28 teeth) were examined in a fixed sequence. The following clinical parameters were recorded as need factors within the Andersen model framework: Dental caries status: crown decayed, missing, and filled teeth (Crown DMFT) and root decayed and filled teeth (Root DFT).Number of functional teeth: teeth with intact crowns capable of normal masticatory function, including caries-free teeth, filled teeth without recurrent caries, sealed teeth, abutment teeth, and teeth with crowns or veneers.Periodontal status: number of teeth with bleeding on probing (BOP), number of teeth with dental calculus, number of teeth with periodontal pockets (probing depth ≥ 4 mm), and number of teeth with clinical attachment loss (CAL) ≥ 4 mm. Dental caries status: crown decayed, missing, and filled teeth (Crown DMFT) and root decayed and filled teeth (Root DFT). Dental caries status: crown decayed, missing, and filled teeth (Crown DMFT) and root decayed and filled teeth (Root DFT). Number of functional teeth: teeth with intact crowns capable of normal masticatory function, including caries-free teeth, filled teeth without recurrent caries, sealed teeth, abutment teeth, and teeth with crowns or veneers. Number of functional teeth: teeth with intact crowns capable of normal masticatory function, including caries-free teeth, filled teeth without recurrent caries, sealed teeth, abutment teeth, and teeth with crowns or veneers. Periodontal status: number of teeth with bleeding on probing (BOP), number of teeth with dental calculus, number of teeth with periodontal pockets (probing depth ≥ 4 mm), and number of teeth with clinical attachment loss (CAL) ≥ 4 mm. Periodontal status: number of teeth with bleeding on probing (BOP), number of teeth with dental calculus, number of teeth with periodontal pockets (probing depth ≥ 4 mm), and number of teeth with clinical attachment loss (CAL) ≥ 4 mm. Questionnaire Survey A structured questionnaire adapted from a validated instrument used in the 4th National Oral Health Survey of China20 was administered by trained interviewers (Supplementary Figure S1). The interviewers underwent standardized training (agreement rate > 90% was acceptable). The questionnaire collected information on predisposing factors and enabling/behavioral factors according to the Andersen model: Predisposing factors included: age (years), gender (male/female), education level (low education level, defined as primary school or below; and high education level, defined as junior high school or above), and residence (classified as urban area or rural area according to participants’ registered identity information).Enabling/Behavioral factors included: dental prosthesis (no/yes, defined as having any implant-supported, fixed, removable partial, or complete denture), dental visit (no/yes within the past 12 months), consumption frequencies of desserts or candy, sweetened beverages, and sweetened milk (never / < 1/day / ≥ 1/day), smoking (no/yes), toothpaste type (none/fluoride-free/fluoride), tooth brushing frequency (< 1/day / 1/day / ≥ 2/day), interdental cleaning frequency (never / < 1/day / ≥ 1/day; including dental floss, interdental brushes, or water flossers), and systemic disease (absence/presence, based on self-reported diagnosis of stroke, diabetes, hypertension, or coronary heart disease).Outcome factors: OHL was assessed using two complementary domains (Supplementary Table 1), oral health belief (OHB) and oral disease awareness (ODA), which capture attitudinal and knowledge-related aspects of oral health literacy. These two domains were selected because oral health literacy is conceptualized as a multidimensional construct encompassing both health-related beliefs/attitudes and functional knowledge required to make appropriate oral health decisions.5,6 The OHB and ODA items were adapted from the validated instrument used in the 4th National Oral Health Survey of China.20 The OHB scale comprised 4 items. The oral disease awareness ODA scale comprised 8 items. Each item was scored dichotomously: 1 for a correct response and 0 for an incorrect or “don’t know” response, yielding score ranges of 0–4 for OHB and 0–8 for ODA. To ensure instrument reliability, internal consistency was assessed using Cronbach’s α, composite reliability (CR), and average variance extracted (AVE). All indices exceeded established thresholds (Cronbach’s α > 0.7, CR > 0.7, AVE > 0.5), confirming adequate psychometric properties. Predisposing factors included: age (years), gender (male/female), education level (low education level, defined as primary school or below; and high education level, defined as junior high school or above), and residence (classified as urban area or rural area according to participants’ registered identity information). Predisposing factors included: age (years), gender (male/female), education level (low education level, defined as primary school or below; and high education level, defined as junior high school or above), and residence (classified as urban area or rural area according to participants’ registered identity information). Enabling/Behavioral factors included: dental prosthesis (no/yes, defined as having any implant-supported, fixed, removable partial, or complete denture), dental visit (no/yes within the past 12 months), consumption frequencies of desserts or candy, sweetened beverages, and sweetened milk (never / < 1/day / ≥ 1/day), smoking (no/yes), toothpaste type (none/fluoride-free/fluoride), tooth brushing frequency (< 1/day / 1/day / ≥ 2/day), interdental cleaning frequency (never / < 1/day / ≥ 1/day; including dental floss, interdental brushes, or water flossers), and systemic disease (absence/presence, based on self-reported diagnosis of stroke, diabetes, hypertension, or coronary heart disease). Enabling/Behavioral factors included: dental prosthesis (no/yes, defined as having any implant-supported, fixed, removable partial, or complete denture), dental visit (no/yes within the past 12 months), consumption frequencies of desserts or candy, sweetened beverages, and sweetened milk (never / < 1/day / ≥ 1/day), smoking (no/yes), toothpaste type (none/fluoride-free/fluoride), tooth brushing frequency (< 1/day / 1/day / ≥ 2/day), interdental cleaning frequency (never / < 1/day / ≥ 1/day; including dental floss, interdental brushes, or water flossers), and systemic disease (absence/presence, based on self-reported diagnosis of stroke, diabetes, hypertension, or coronary heart disease). Outcome factors: OHL was assessed using two complementary domains (Supplementary Table 1), oral health belief (OHB) and oral disease awareness (ODA), which capture attitudinal and knowledge-related aspects of oral health literacy. These two domains were selected because oral health literacy is conceptualized as a multidimensional construct encompassing both health-related beliefs/attitudes and functional knowledge required to make appropriate oral health decisions.5,6 The OHB and ODA items were adapted from the validated instrument used in the 4th National Oral Health Survey of China.20 The OHB scale comprised 4 items. The oral disease awareness ODA scale comprised 8 items. Each item was scored dichotomously: 1 for a correct response and 0 for an incorrect or “don’t know” response, yielding score ranges of 0–4 for OHB and 0–8 for ODA. To ensure instrument reliability, internal consistency was assessed using Cronbach’s α, composite reliability (CR), and average variance extracted (AVE). All indices exceeded established thresholds (Cronbach’s α > 0.7, CR > 0.7, AVE > 0.5), confirming adequate psychometric properties. Outcome factors: OHL was assessed using two complementary domains (Supplementary Table 1), oral health belief (OHB) and oral disease awareness (ODA), which capture attitudinal and knowledge-related aspects of oral health literacy. These two domains were selected because oral health literacy is conceptualized as a multidimensional construct encompassing both health-related beliefs/attitudes and functional knowledge required to make appropriate oral health decisions.5,6 The OHB and ODA items were adapted from the validated instrument used in the 4th National Oral Health Survey of China.20 The OHB scale comprised 4 items. The oral disease awareness ODA scale comprised 8 items. Each item was scored dichotomously: 1 for a correct response and 0 for an incorrect or “don’t know” response, yielding score ranges of 0–4 for OHB and 0–8 for ODA. To ensure instrument reliability, internal consistency was assessed using Cronbach’s α, composite reliability (CR), and average variance extracted (AVE). All indices exceeded established thresholds (Cronbach’s α > 0.7, CR > 0.7, AVE > 0.5), confirming adequate psychometric properties. Statistical Analysis Normality of continuous variables was assessed using the Shapiro–Wilk test. Continuous variables were presented as mean ± standard deviation (SD) or median (interquartile range, IQR). Categorical variables were presented as frequency (%). Between-group comparisons were performed using the Mann–Whitney U-test for continuous variables and the chi-square (χ2) test or Fisher’s exact test for categorical variables. Multicollinearity among all independent variables was evaluated using variance inflation factor (VIF) diagnostics. All VIF values were below 5, indicating no multicollinearity concerns (Supplementary Table 2). Poisson regression was employed to identify factors associated with OHL. Prior to model fitting, data dispersion characteristics were evaluated using the variance-to-mean ratio and the Cameron-Trivedi dispersion test. The variance-to-mean ratio was 0.905 for ODA scores and 0.317 for OHB scores. Neither outcome exhibited overdispersion (Cameron-Trivedi test: P = 0.9997 for ODA; P = 1.000 for OHB). Accordingly, Poisson regression with Huber-White robust (sandwich) standard errors was utilized. Variables with P < 0.10 in univariable analyses were entered into the multivariable model. Results were expressed as incidence rate ratios (IRR) with 95% confidence intervals (CI). Path analysis was then conducted using the lavaan package in R. An initial conceptual model was constructed based on the modified Andersen’s behavioral model, hypothesizing that predisposing factors are associated with OHL both directly and indirectly through need and enabling/behavioral factors. The structure of the path model was specified primarily on the basis of Andersen’s behavioral framework. To preserve parsimony given the sample size, a relaxed statistical screen (P < 0.10 in Poisson regression) was then applied only to retain candidate indicators within each theoretically specified domain. We acknowledge that this hybrid theory- and data-informed strategy may have excluded variables of theoretical relevance that did not reach the screening threshold. Model fit was evaluated using the following indices: χ2/df < 3, comparative fit index (CFI) > 0.90, Tucker-Lewis index (TLI) > 0.90, root mean square error of approximation (RMSEA) < 0.06, and standardized root mean square residual (SRMR) < 0.08.21,22 The robust maximum likelihood (MLR) estimator was employed for path coefficient estimation to account for non-normality. Direct, indirect, and total effects were estimated and reported as both unstandardized (β) and standardized (Std. β) coefficients with standard errors. Because the data were cross-sectional, all estimated paths represent statistical associations rather than causal effects; the directionality specified in the model reflects the hypothesized theoretical structure of Andersen’s model and not a demonstrated temporal sequence. For mediation analysis, the maximum likelihood (ML) estimator was employed in combination with the nonparametric bootstrap method (5,000 resamples) to construct bias-corrected 95% confidence intervals (BC 95% CI) for indirect effects. This approach was adopted because bootstrap-based inference provides more accurate confidence intervals for indirect effects than normal-theory methods, particularly when the sampling distribution of the product of coefficients is non-normal. The significance of total indirect effects was determined based on whether the bootstrap bias-corrected 95% CI excluded zero. The proportion mediated was calculated as the absolute value of the indirect effect divided by the absolute value of the total effect, multiplied by 100%. The potential moderating effects of education level and residence were first tested by introducing interaction terms into the path model (Supplementary Table 3). Subsequently, configural multi-group path analysis was conducted, allowing path coefficients to vary freely across subgroups stratified by education level (high vs low) and residence (urban vs rural). The scaled chi-square difference test (Satorra-Bentler) was used for model comparison. All statistical analyses were conducted using R software (version 4.4.2). All statistical tests were two-sided with a significance level of P < 0.05. Results Participant Characteristics The characteristics of the 423 participants stratified by residence are presented in Table 1. The overall median age was 68.9 years (IQR: 66.5–71.4), with 52.0% being female. The median OHB scores were significantly higher among urban than rural residents (3.5 vs 3, P = 0.026), whereas ODA scores did not differ significantly between groups (P = 0.795).Table 1Demographic, Behavioral, and Clinical Characteristics of the Study Participants Stratified by ResidenceVariablesTotal(N = 423)Urban Area(n = 192)Rural Area(n = 231)PAge, median (IQR)68.9 (66.5–71.4)68.6 (66.4–71.4)69.0 (67.0–71.2)0.502Gender, n (%)0.032Male203 (48.0)81 (42.2)122 (52.8)Female220 (52.0)111 (57.8)109 (47.2)Education level, n (%)<0.001Low education level250 (59.1)90 (46.9)160 (69.3)High education level173 (40.9)102 (53.1)71 (30.7)Dental prosthesis, n (%)<0.001No214 (50.6)78 (40.6)136 (58.9)Yes209 (49.4)114 (59.4)95 (41.1)Dental visit, n (%)<0.001No160 (37.8)52 (27.1)108 (46.8)Yes263 (62.2)140 (72.9)123 (53.2)Dessert or candy consumption frequency, n (%)<0.001Never131 (31.0)38 (19.8)93 (40.3)<1/day259 (61.2)134 (69.8)125 (54.1)≥1/day33 (7.8)20 (10.4)13 (5.6)Sweetened beverage consumption frequency, n (%)0.446Never259 (61.2)112 (58.3)147 (63.6)<1/day154 (36.4)76 (39.6)78 (33.8)≥1/day10 (2.4)4 (2.1)6 (2.6)Sweetened milk consumption frequency, n (%)0.007Never166 (39.2)65 (33.9)101 (43.7)<1/day168 (39.7)74 (38.5)94 (40.7)≥1/day89 (21.0)53 (27.6)36 (15.6)Smoking, n (%)0.004No315 (74.5)156 (81.3)159 (66.8)Yes108 (25.5)36 (18.8)72 (31.2)Toothpaste type, n (%)0.974None6 (1.4)3 (1.6)3 (1.3)Fluoride-free288 (68.1)130 (67.7)158 (68.4)Fluoride129 (30.5)59 (30.7)70 (30.3)Tooth brushing frequency, n (%)<0.001<1/day12 (2.8)5 (2.6)7 (3.0)1/day219 (51.8)70 (36.5)149 (64.5)≥2/day192 (45.4)117 (60.9)75 (32.5)Interdental cleaning frequency, n (%)0.032Never94 (22.2)32 (16.7)62 (26.8)<1/day66 (15.6)35 (18.2)31 (13.4)≥1/day263 (62.2)125 (65.1)138 (59.7)Crown DMFT, mean ± SD4.601 ± 5.2043.786 ± 4.0745.277 ± 5.9070.021Root DFT, mean ± SD0.504 ± 1.3200.682 ± 1.5680.355 ± 1.053<0.001Number of teeth with BOP, mean ± SD6.823 ± 7.7406.094 ± 7.1747.429 ± 8.1470.262Number of teeth with dental calculus, mean ± SD11.868 ± 8.69011.615 ± 8.71412.078 ± 8.6830.542Number of teeth with periodontal pockets, mean ± SD4.168 ± 5.6443.141 ± 4.1335.022 ± 6.5300.098Number of teeth with CAL ≥ 4 mm, mean ± SD1.473 ± 1.7331.068 ± 1.3731.810 ± 1.922<0.001Number of functional teeth, mean ± SD20.411 ± 6.99121.130 ± 6.39519.814 ± 7.4110.139Systemic disease, n (%)0.097Absence226 (53.4)94 (49.0)132 (57.1)Presence197 (46.6)98 (51.0)99 (42.9)OHB scores, median (IQR)3 (3–4)3.5 (3–4)3 (3–4)0.026ODA scores, median (IQR)5 (3–6)5 (4–6)5 (3–6)0.795Notes: All continuous variables were assessed for normality using the Shapiro–Wilk test and were found to be non-normally distributed. Nevertheless, continuous variables under Need factors are presented as mean ± standard deviation (SD) to facilitate comparability with previous studies, while non-parametric tests were applied for statistical comparisons. Other continuous variables with substantial skewness are presented as median (interquartile range, IQR). Categorical variables are presented as frequency (percentage). P values were derived from the Mann–Whitney U-test for continuous variables and the chi-square (χ2) test or Fisher’s exact test for categorical variables. Bold P values indicate statistical significance at the 0.05 level.Abbreviations: SD, standard deviation; IQR, interquartile range; DMFT, decayed, missing, and filled teeth; DFT, decayed and filled teeth; BOP, bleeding on probing; CAL, clinical attachment loss; ODA, oral disease awareness; OHB, oral health belief. Table 1Demographic, Behavioral, and Clinical Characteristics of the Study Participants Stratified by ResidenceVariablesTotal(N = 423)Urban Area(n = 192)Rural Area(n = 231)PAge, median (IQR)68.9 (66.5–71.4)68.6 (66.4–71.4)69.0 (67.0–71.2)0.502Gender, n (%)0.032Male203 (48.0)81 (42.2)122 (52.8)Female220 (52.0)111 (57.8)109 (47.2)Education level, n (%)<0.001Low education level250 (59.1)90 (46.9)160 (69.3)High education level173 (40.9)102 (53.1)71 (30.7)Dental prosthesis, n (%)<0.001No214 (50.6)78 (40.6)136 (58.9)Yes209 (49.4)114 (59.4)95 (41.1)Dental visit, n (%)<0.001No160 (37.8)52 (27.1)108 (46.8)Yes263 (62.2)140 (72.9)123 (53.2)Dessert or candy consumption frequency, n (%)<0.001Never131 (31.0)38 (19.8)93 (40.3)<1/day259 (61.2)134 (69.8)125 (54.1)≥1/day33 (7.8)20 (10.4)13 (5.6)Sweetened beverage consumption frequency, n (%)0.446Never259 (61.2)112 (58.3)147 (63.6)<1/day154 (36.4)76 (39.6)78 (33.8)≥1/day10 (2.4)4 (2.1)6 (2.6)Sweetened milk consumption frequency, n (%)0.007Never166 (39.2)65 (33.9)101 (43.7)<1/day168 (39.7)74 (38.5)94 (40.7)≥1/day89 (21.0)53 (27.6)36 (15.6)Smoking, n (%)0.004No315 (74.5)156 (81.3)159 (66.8)Yes108 (25.5)36 (18.8)72 (31.2)Toothpaste type, n (%)0.974None6 (1.4)3 (1.6)3 (1.3)Fluoride-free288 (68.1)130 (67.7)158 (68.4)Fluoride129 (30.5)59 (30.7)70 (30.3)Tooth brushing frequency, n (%)<0.001<1/day12 (2.8)5 (2.6)7 (3.0)1/day219 (51.8)70 (36.5)149 (64.5)≥2/day192 (45.4)117 (60.9)75 (32.5)Interdental cleaning frequency, n (%)0.032Never94 (22.2)32 (16.7)62 (26.8)<1/day66 (15.6)35 (18.2)31 (13.4)≥1/day263 (62.2)125 (65.1)138 (59.7)Crown DMFT, mean ± SD4.601 ± 5.2043.786 ± 4.0745.277 ± 5.9070.021Root DFT, mean ± SD0.504 ± 1.3200.682 ± 1.5680.355 ± 1.053<0.001Number of teeth with BOP, mean ± SD6.823 ± 7.7406.094 ± 7.1747.429 ± 8.1470.262Number of teeth with dental calculus, mean ± SD11.868 ± 8.69011.615 ± 8.71412.078 ± 8.6830.542Number of teeth with periodontal pockets, mean ± SD4.168 ± 5.6443.141 ± 4.1335.022 ± 6.5300.098Number of teeth with CAL ≥ 4 mm, mean ± SD1.473 ± 1.7331.068 ± 1.3731.810 ± 1.922<0.001Number of functional teeth, mean ± SD20.411 ± 6.99121.130 ± 6.39519.814 ± 7.4110.139Systemic disease, n (%)0.097Absence226 (53.4)94 (49.0)132 (57.1)Presence197 (46.6)98 (51.0)99 (42.9)OHB scores, median (IQR)3 (3–4)3.5 (3–4)3 (3–4)0.026ODA scores, median (IQR)5 (3–6)5 (4–6)5 (3–6)0.795Notes: All continuous variables were assessed for normality using the Shapiro–Wilk test and were found to be non-normally distributed. Nevertheless, continuous variables under Need factors are presented as mean ± standard deviation (SD) to facilitate comparability with previous studies, while non-parametric tests were applied for statistical comparisons. Other continuous variables with substantial skewness are presented as median (interquartile range, IQR). Categorical variables are presented as frequency (percentage). P values were derived from the Mann–Whitney U-test for continuous variables and the chi-square (χ2) test or Fisher’s exact test for categorical variables. Bold P values indicate statistical significance at the 0.05 level.Abbreviations: SD, standard deviation; IQR, interquartile range; DMFT, decayed, missing, and filled teeth; DFT, decayed and filled teeth; BOP, bleeding on probing; CAL, clinical attachment loss; ODA, oral disease awareness; OHB, oral health belief. Demographic, Behavioral, and Clinical Characteristics of the Study Participants Stratified by Residence Notes: All continuous variables were assessed for normality using the Shapiro–Wilk test and were found to be non-normally distributed. Nevertheless, continuous variables under Need factors are presented as mean ± standard deviation (SD) to facilitate comparability with previous studies, while non-parametric tests were applied for statistical comparisons. Other continuous variables with substantial skewness are presented as median (interquartile range, IQR). Categorical variables are presented as frequency (percentage). P values were derived from the Mann–Whitney U-test for continuous variables and the chi-square (χ2) test or Fisher’s exact test for categorical variables. Bold P values indicate statistical significance at the 0.05 level. Abbreviations: SD, standard deviation; IQR, interquartile range; DMFT, decayed, missing, and filled teeth; DFT, decayed and filled teeth; BOP, bleeding on probing; CAL, clinical attachment loss; ODA, oral disease awareness; OHB, oral health belief. Poisson Regression Analysis The results of univariable and multivariable Poisson regression analyses are displayed in Table 2. Female (IRR = 0.905, 95% CI 0.829–0.989) and interdental cleaning frequency < 1/day (IRR = 0.796, 0.672–0.943) were negatively associated with OHB, whereas dental visit (IRR = 1.094, 1.004–1.193), sweetened beverage consumption ≥ 1/day (IRR = 1.288, 1.037–1.600), and Root DFT (IRR = 1.027, 1.005–1.050) were positively associated with OHB (all P < 0.05). For ODA, dental prosthesis use (IRR = 1.073, 1.012–1.138), dessert or candy consumption ≥ 1/day (IRR = 1.131, 1.023–1.251), and Root DFT (IRR = 1.023, 1.008–1.038) were positively associated (all P < 0.05). Model diagnostics indicated no overdispersion and satisfactory fit.Table 2Univariable and Multivariable Poisson Regression Analysis of Factors Associated with Oral Health Literacy Among Older AdultsOHB ScoresODA ScoresUnivariable AnalysisMultivariable AnalysisUnivariable AnalysisMultivariable AnalysisIRR (95% CI)PIRRa (95% CI)PIRR (95% CI)PIRRa (95% CI)PAge0.995 (0.979–1.011)0.5221.005 (0.993–1.016)0.422GenderMale†––––––Female0.905 (0.832–0.984)0.0200.905 (0.829–0.989)0.0270.974 (0.917–1.035)0.392Education levelLow education level†––––––––High education level1.127 (1.038–1.224)0.0051.070 (0.982–1.167)0.1211.055 (0.993–1.120)0.0811.036 (0.977–1.098)0.240Residence––––––Urban areaRural area0.975 (0.896–1.060)0.5500.932 (0.878–0.989)0.0200.984 (0.927–1.046)0.611Dental prosthesisNo†––––––––Yes1.101 (1.013–1.198)0.0241.066 (0.983–1.157)0.1231.094 (1.030–1.162)0.0041.073 (1.012–1.138)0.019Dental visitNo†––––––Yes1.125 (1.029–1.231)0.0101.094 (1.004–1.193)0.0411.049 (0.983–1.119)0.151Dessert or candy consumption frequencyNever†––––––<1/day0.991 (0.901–1.091)0.8611.058 (0.984–1.137)0.1281.043 (0.970–1.123)0.257≥1/day1.058 (0.900–1.244)0.4931.149 (1.040–1.268)0.0061.131 (1.023–1.251)0.016Sweetened beverage consumption frequencyNever†––––––<1/day1.021 (0.935–1.114)0.6441.003 (0.922–1.092)0.9411.032 (0.972–1.096)0.297≥1/day1.231 (1.007–1.505)0.0421.288 (1.037–1.600)0.0220.898 (0.737–1.093)0.283Sweetened milk consumption frequencyNever†––––<1/day1.071 (0.974–1.178)0.1561.044 (0.975–1.118)0.216≥1/day1.045 (0.931–1.173)0.4531.016 (0.935–1.105)0.704SmokingNo†––––Yes1.036 (0.944–1.138)0.4541.008 (0.937–1.084)0.839Toothpaste typeNone†––––Fluoride-free1.212 (0.645–2.280)0.5501.089 (0.843–1.408)0.512Fluoride1.564 (0.830–2.945)0.1661.166 (0.900–1.509)0.245Tooth brushing frequency<1/day†----1/day0.909 (0.711–1.164)0.4501.000 (0.837–1.195)1.000≥2/day1.037 (0.811–1.327)0.7711.109 (0.930–1.323)0.249Interdental cleaning frequencyNever†------<1/day0.778 (0.655–0.925)0.0050.796 (0.672–0.943)0.0080.912 (0.810–1.027)0.127≥1/day0.952 (0.861–1.052)0.3320.944 (0.853–1.045)0.2690.985 (0.922–1.052)0.652Crown DMFT1.002 (0.994–1.010)0.6170.999 (0.993–1.004)0.605Root DFT1.034 (1.013–1.055)0.0011.027 (1.005–1.050)0.0141.027 (1.012–1.041)<0.0011.023 (1.008–1.038)0.003Number of teeth with BOP1.003 (0.998–1.009)0.2530.998 (0.994–1.001)0.224Number of teeth with dental calculus0.999 (0.994–1.004)0.7620.999 (0.996–1.003)0.716Number of teeth with periodontal pockets1.004 (0.996–1.011)0.3420.993 (0.986–1.000)0.0530.995 (0.988–1.002)0.129Number of teeth with CAL ≥ 4 mm1.004 (0.981–1.028)0.7480.993 (0.976–1.011)0.456Number of functional teeth0.999 (0.993–1.005)0.6971.000 (0.996–1.005)0.910Systemic diseaseAbsence†------Presence1.047 (0.963–1.138)0.2821.055 (0.993–1.120)0.0831.054 (0.993–1.119)0.083Notes: The multivariable model included variables with P < 0.10 in univariable analysis. Poisson regression with robust (sandwich) standard errors was used for both outcomes. Model fit was assessed using the Pearson χ2/df statistic (ODA = 0.839; OHB = 0.315) and deviance/df (ODA = 1.079; OHB = 0.412), indicating no evidence of overdispersion. Bold P values indicate statistical significance at the 0.05 level.Abbreviations: IRR, incidence rate ratio; IRRa, adjusted incidence rate ratio; CI, confidence interval; †, reference variable; DMFT, decayed, missing, and filled teeth; DFT, decayed and filled teeth; BOP, bleeding on probing; CAL, clinical attachment loss. Table 2Univariable and Multivariable Poisson Regression Analysis of Factors Associated with Oral Health Literacy Among Older AdultsOHB ScoresODA ScoresUnivariable AnalysisMultivariable AnalysisUnivariable AnalysisMultivariable AnalysisIRR (95% CI)PIRRa (95% CI)PIRR (95% CI)PIRRa (95% CI)PAge0.995 (0.979–1.011)0.5221.005 (0.993–1.016)0.422GenderMale†––––––Female0.905 (0.832–0.984)0.0200.905 (0.829–0.989)0.0270.974 (0.917–1.035)0.392Education levelLow education level†––––––––High education level1.127 (1.038–1.224)0.0051.070 (0.982–1.167)0.1211.055 (0.993–1.120)0.0811.036 (0.977–1.098)0.240Residence––––––Urban areaRural area0.975 (0.896–1.060)0.5500.932 (0.878–0.989)0.0200.984 (0.927–1.046)0.611Dental prosthesisNo†––––––––Yes1.101 (1.013–1.198)0.0241.066 (0.983–1.157)0.1231.094 (1.030–1.162)0.0041.073 (1.012–1.138)0.019Dental visitNo†––––––Yes1.125 (1.029–1.231)0.0101.094 (1.004–1.193)0.0411.049 (0.983–1.119)0.151Dessert or candy consumption frequencyNever†––––––<1/day0.991 (0.901–1.091)0.8611.058 (0.984–1.137)0.1281.043 (0.970–1.123)0.257≥1/day1.058 (0.900–1.244)0.4931.149 (1.040–1.268)0.0061.131 (1.023–1.251)0.016Sweetened beverage consumption frequencyNever†––––––<1/day1.021 (0.935–1.114)0.6441.003 (0.922–1.092)0.9411.032 (0.972–1.096)0.297≥1/day1.231 (1.007–1.505)0.0421.288 (1.037–1.600)0.0220.898 (0.737–1.093)0.283Sweetened milk consumption frequencyNever†––––<1/day1.071 (0.974–1.178)0.1561.044 (0.975–1.118)0.216≥1/day1.045 (0.931–1.173)0.4531.016 (0.935–1.105)0.704SmokingNo†––––Yes1.036 (0.944–1.138)0.4541.008 (0.937–1.084)0.839Toothpaste typeNone†––––Fluoride-free1.212 (0.645–2.280)0.5501.089 (0.843–1.408)0.512Fluoride1.564 (0.830–2.945)0.1661.166 (0.900–1.509)0.245Tooth brushing frequency<1/day†----1/day0.909 (0.711–1.164)0.4501.000 (0.837–1.195)1.000≥2/day1.037 (0.811–1.327)0.7711.109 (0.930–1.323)0.249Interdental cleaning frequencyNever†------<1/day0.778 (0.655–0.925)0.0050.796 (0.672–0.943)0.0080.912 (0.810–1.027)0.127≥1/day0.952 (0.861–1.052)0.3320.944 (0.853–1.045)0.2690.985 (0.922–1.052)0.652Crown DMFT1.002 (0.994–1.010)0.6170.999 (0.993–1.004)0.605Root DFT1.034 (1.013–1.055)0.0011.027 (1.005–1.050)0.0141.027 (1.012–1.041)<0.0011.023 (1.008–1.038)0.003Number of teeth with BOP1.003 (0.998–1.009)0.2530.998 (0.994–1.001)0.224Number of teeth with dental calculus0.999 (0.994–1.004)0.7620.999 (0.996–1.003)0.716Number of teeth with periodontal pockets1.004 (0.996–1.011)0.3420.993 (0.986–1.000)0.0530.995 (0.988–1.002)0.129Number of teeth with CAL ≥ 4 mm1.004 (0.981–1.028)0.7480.993 (0.976–1.011)0.456Number of functional teeth0.999 (0.993–1.005)0.6971.000 (0.996–1.005)0.910Systemic diseaseAbsence†------Presence1.047 (0.963–1.138)0.2821.055 (0.993–1.120)0.0831.054 (0.993–1.119)0.083Notes: The multivariable model included variables with P < 0.10 in univariable analysis. Poisson regression with robust (sandwich) standard errors was used for both outcomes. Model fit was assessed using the Pearson χ2/df statistic (ODA = 0.839; OHB = 0.315) and deviance/df (ODA = 1.079; OHB = 0.412), indicating no evidence of overdispersion. Bold P values indicate statistical significance at the 0.05 level.Abbreviations: IRR, incidence rate ratio; IRRa, adjusted incidence rate ratio; CI, confidence interval; †, reference variable; DMFT, decayed, missing, and filled teeth; DFT, decayed and filled teeth; BOP, bleeding on probing; CAL, clinical attachment loss. Univariable and Multivariable Poisson Regression Analysis of Factors Associated with Oral Health Literacy Among Older Adults Notes: The multivariable model included variables with P < 0.10 in univariable analysis. Poisson regression with robust (sandwich) standard errors was used for both outcomes. Model fit was assessed using the Pearson χ2/df statistic (ODA = 0.839; OHB = 0.315) and deviance/df (ODA = 1.079; OHB = 0.412), indicating no evidence of overdispersion. Bold P values indicate statistical significance at the 0.05 level. Abbreviations: IRR, incidence rate ratio; IRRa, adjusted incidence rate ratio; CI, confidence interval; †, reference variable; DMFT, decayed, missing, and filled teeth; DFT, decayed and filled teeth; BOP, bleeding on probing; CAL, clinical attachment loss. Path Analysis The conceptual model (Figure 2A) and the final path model (Figure 2B) are presented. The final model showed acceptable fit (χ2/df = 1.457, CFI = 0.960, TLI = 0.879, RMSEA = 0.033 (90% CI: 0.000–0.057), SRMR = 0.031); the TLI was marginally below the conventional 0.90 threshold, whereas all other indices met conventional cut-offs. The explained variance was modest (R2 = 0.064 for ODA; R2 = 0.056 for OHB).Figure 2Path analysis of factors associated with oral health literacy among older adults based on the modified Andersen’s behavioral model. (A) Conceptual model illustrating the hypothesized pathways. (B) Final path model with standardized path coefficients (β). Only statistically significant paths (P < 0.05) are shown with solid lines; non-significant paths are displayed with gray dashed lines. Model fit: χ2/df = 1.457, CFI = 0.960, TLI = 0.879, RMSEA = 0.033 (90% CI: 0.000–0.057), SRMR = 0.031. R2 (ODA scores) = 0.064; R2 (OHB scores) = 0.056.Path analysis of oral health literacy factors among older adults using Andersen’s model.Two panels illustrate factors affecting oral health literacy. Panel A shows a conceptual model with predisposing factors (gender, education level, residence), need factors (root DFT, number of teeth with periodontal pockets) and enabling/behavioral factors (dental prosthesis, dental visit, dessert or candy consumption frequency, sweetened beverage consumption frequency, interdental cleaning frequency) leading to outcomes (OHB score, ODA score). Panel B presents a path model with standardized coefficients. Predisposing factors connect to need factors and enabling/behavioral factors, which influence outcomes. Significant paths include root DFT to ODA scores (0.097), number of teeth with periodontal pockets to ODA scores (0.116), dental visit to OHB scores (0.109) and dessert or candy consumption frequency to OHB scores (-0.078).Abbreviations: DFT, decayed and filled teeth; OHB, oral health belief; ODA, oral disease awareness. Path analysis of factors associated with oral health literacy among older adults based on the modified Andersen’s behavioral model. (A) Conceptual model illustrating the hypothesized pathways. (B) Final path model with standardized path coefficients (β). Only statistically significant paths (P < 0.05) are shown with solid lines; non-significant paths are displayed with gray dashed lines. Model fit: χ2/df = 1.457, CFI = 0.960, TLI = 0.879, RMSEA = 0.033 (90% CI: 0.000–0.057), SRMR = 0.031. R2 (ODA scores) = 0.064; R2 (OHB scores) = 0.056. The direct, indirect, and total effects are summarized in Table 3. Among predisposing factors, female had an indirect effect on OHB and ODA, and education level had an indirect effect on ODA (all P < 0.05). Among need factors, Root DFT had significant direct positive effects on both OHB (Std. β = 0.109, P = 0.002) and ODA (Std. β = 0.097, P = 0.009). Among enabling/behavioral factors, dental visit directly enhanced ODA and dental prosthesis use directly enhanced OHB. In contrast to the positive association observed in the Poisson regression, sweetened beverage consumption ≥ 1/day showed a negative direct association with OHB in the path model (Std. β = −0.078, P = 0.043). Because only 10 participants reported this consumption level, this estimate was unstable across model specifications.Table 3Direct, Indirect, and Total Effects of Predisposing, Need, and Enabling/Behavioral Factors on Oral Health Literacy Based on Path AnalysisPredictorOutcomeDirect EffectIndirect EffectTotal EffectβSEPStd. ββSEPStd. ββSEPStd. βPredisposing factorsGender (Female)OHB−0.1310.1030.202−0.0660.0480.0240.0440.024−0.0830.1020.418−0.041ODA−0.4710.2150.028−0.1160.1370.0540.0110.034−0.3340.2130.118−0.082Education level (High education level)OHB0.0560.1050.5910.0280.0430.0300.1540.0210.0990.1050.3480.049ODA0.3110.2090.1370.0750.1380.0610.0230.0330.4490.2080.0310.108Residence (Rural area)OHB−0.1260.0930.175−0.063−0.0460.0260.075−0.023−0.1720.0970.074−0.086ODA0.0170.1980.9320.004−0.1030.0580.076−0.025−0.0860.2010.668−0.021Need factorsRoot DFTOHB0.0820.0270.0020.109––––0.0820.0270.0020.109ODA0.1500.0580.0090.097––––0.1500.0580.0090.097Number of teeth with periodontal pocketsOHB−0.0180.0100.088−0.100––––−0.0180.0100.088−0.100ODA0.0200.0190.2900.054––––0.0200.0190.2900.054Enabling/Behavioral factorsDental visit (Yes)OHB0.0320.1050.7590.016––––0.0320.1050.7590.016ODA0.4270.2000.0330.102––––0.4270.2000.0330.102Dental prosthesis (Yes)OHB0.2060.0980.0350.103––––0.2060.0980.0350.103ODA0.3080.1910.1060.076––––0.3080.1910.1060.076Dessert or candy consumption frequency (≥1/day)ODA0.1280.3120.6810.017––––0.1280.3120.6810.017Sweetened beverage consumption frequency (≥1/day)OHB−0.5130.2540.043−0.078––––−0.5130.2540.043−0.078Interdental cleaning frequency (≥1/day)OHB−0.0840.1070.434−0.035––––−0.0840.1070.434−0.035ODA−0.4290.2420.076−0.088––––−0.4290.2420.076−0.088Note: Model fit: χ2/df = 1.457; CFI = 0.960; TLI = 0.879; RMSEA = 0.033 (90% CI: 0.000–0.057); SRMR = 0.031. Bold P values indicate statistical significance at the 0.05 level. Indirect effects represent the total indirect effect of each predisposing factor through all mediating variables (need factors and enabling/behavioral factors combined). Indirect effects are not applicable (-) for need and enabling/behavioral factors as these are not hypothesized to operate through further mediators in the model. R2 (ODA scores) = 0.064; R2 (OHB scores) = 0.056.Abbreviations: β, unstandardized coefficient; SE, standard error; Std. β, standardized coefficient; DFT, decayed and filled teeth; ODA, oral disease awareness; OHB, oral health belief; CFI, comparative fit index; TLI, Tucker-Lewis index; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual. Table 3Direct, Indirect, and Total Effects of Predisposing, Need, and Enabling/Behavioral Factors on Oral Health Literacy Based on Path AnalysisPredictorOutcomeDirect EffectIndirect EffectTotal EffectβSEPStd. ββSEPStd. ββSEPStd. βPredisposing factorsGender (Female)OHB−0.1310.1030.202−0.0660.0480.0240.0440.024−0.0830.1020.418−0.041ODA−0.4710.2150.028−0.1160.1370.0540.0110.034−0.3340.2130.118−0.082Education level (High education level)OHB0.0560.1050.5910.0280.0430.0300.1540.0210.0990.1050.3480.049ODA0.3110.2090.1370.0750.1380.0610.0230.0330.4490.2080.0310.108Residence (Rural area)OHB−0.1260.0930.175−0.063−0.0460.0260.075−0.023−0.1720.0970.074−0.086ODA0.0170.1980.9320.004−0.1030.0580.076−0.025−0.0860.2010.668−0.021Need factorsRoot DFTOHB0.0820.0270.0020.109––––0.0820.0270.0020.109ODA0.1500.0580.0090.097––––0.1500.0580.0090.097Number of teeth with periodontal pocketsOHB−0.0180.0100.088−0.100––––−0.0180.0100.088−0.100ODA0.0200.0190.2900.054––––0.0200.0190.2900.054Enabling/Behavioral factorsDental visit (Yes)OHB0.0320.1050.7590.016––––0.0320.1050.7590.016ODA0.4270.2000.0330.102––––0.4270.2000.0330.102Dental prosthesis (Yes)OHB0.2060.0980.0350.103––––0.2060.0980.0350.103ODA0.3080.1910.1060.076––––0.3080.1910.1060.076Dessert or candy consumption frequency (≥1/day)ODA0.1280.3120.6810.017––––0.1280.3120.6810.017Sweetened beverage consumption frequency (≥1/day)OHB−0.5130.2540.043−0.078––––−0.5130.2540.043−0.078Interdental cleaning frequency (≥1/day)OHB−0.0840.1070.434−0.035––––−0.0840.1070.434−0.035ODA−0.4290.2420.076−0.088––––−0.4290.2420.076−0.088Note: Model fit: χ2/df = 1.457; CFI = 0.960; TLI = 0.879; RMSEA = 0.033 (90% CI: 0.000–0.057); SRMR = 0.031. Bold P values indicate statistical significance at the 0.05 level. Indirect effects represent the total indirect effect of each predisposing factor through all mediating variables (need factors and enabling/behavioral factors combined). Indirect effects are not applicable (-) for need and enabling/behavioral factors as these are not hypothesized to operate through further mediators in the model. R2 (ODA scores) = 0.064; R2 (OHB scores) = 0.056.Abbreviations: β, unstandardized coefficient; SE, standard error; Std. β, standardized coefficient; DFT, decayed and filled teeth; ODA, oral disease awareness; OHB, oral health belief; CFI, comparative fit index; TLI, Tucker-Lewis index; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual. Direct, Indirect, and Total Effects of Predisposing, Need, and Enabling/Behavioral Factors on Oral Health Literacy Based on Path Analysis Note: Model fit: χ2/df = 1.457; CFI = 0.960; TLI = 0.879; RMSEA = 0.033 (90% CI: 0.000–0.057); SRMR = 0.031. Bold P values indicate statistical significance at the 0.05 level. Indirect effects represent the total indirect effect of each predisposing factor through all mediating variables (need factors and enabling/behavioral factors combined). Indirect effects are not applicable (-) for need and enabling/behavioral factors as these are not hypothesized to operate through further mediators in the model. R2 (ODA scores) = 0.064; R2 (OHB scores) = 0.056. Abbreviations: β, unstandardized coefficient; SE, standard error; Std. β, standardized coefficient; DFT, decayed and filled teeth; ODA, oral disease awareness; OHB, oral health belief; CFI, comparative fit index; TLI, Tucker-Lewis index; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual. The significant associations of dental visit and dental prosthesis use differed between the Poisson regressions and the path model. This is expected because the two approaches differ in structure: the Poisson models estimated each outcome separately using only its own set of screened covariates, whereas the path model simultaneously estimated all relationships for both outcomes within a single framework using the robust maximum likelihood estimator. Consequently, the same predictor can reach significance for a different outcome under each method. Mediation Analysis Mediation analysis results are presented in Table 4. Gender (female) exerted a significant positive indirect effect on ODA (Std. β = 0.034, P = 0.011) and OHB (Std. β = 0.024, P = 0.044) through mediating variables, despite having a negative direct effect on ODA. High education level exerted a significant positive indirect effect on ODA (Std. β = 0.033, P = 0.023). The decomposition of specific indirect pathways (Table 5) revealed three positive and two negative paths (P < 0.10).Table 4Mediation Analysis Results of Direct, Indirect, and Total Effects of Predisposing Factors on Oral Health LiteracyPredisposing FactorsOutcomeEffect TypeβSEPStd. β% MediatedGender (Female)OHBDirect−0.1310.1030.202−0.066Total indirect0.0480.0240.0440.02457.8Total−0.0830.1020.418−0.041ODADirect−0.4710.2150.028−0.116Total indirect0.1370.0540.0110.03441.0Total−0.3340.2130.118−0.082Education level (High education level)OHBDirect0.0560.1050.5910.028Total indirect0.0430.0300.1540.02143.4Total0.0990.1050.3480.049ODADirect0.3110.2090.1370.075Total indirect0.1380.0610.0230.03330.7Total0.4490.2080.0310.108Residence (Rural area)OHBDirect−0.1260.0930.175−0.063Total indirect−0.0460.0260.075−0.02326.7Total−0.1720.0970.074−0.086ODADirect0.0170.1980.9320.004Total indirect−0.1030.0580.076−0.025119.8Total−0.0860.2010.668−0.021Note: Path analysis using maximum likelihood (ML) estimator with nonparametric bootstrap (5,000 resamples). Statistical significance of indirect effects was determined by whether the bootstrap bias-corrected 95% confidence interval (BC 95% CI) excluded zero. % Mediated = |Indirect effect| / |total effect| × 100%. % Mediated values exceeding 100% indicate suppression effects, where the indirect effect exceeds the total effect due to opposing direct and indirect effect directions. Bold P values indicate statistical significance at the 0.05 level.Abbreviations: β, unstandardized coefficient; SE, standard error; Std. β, standardized coefficient. Table 5Specific Indirect Effects Reaching Marginal Significance (P < 0.10)Predisposing FactorsMediatorOutcomeStd. βPDirectionGenderDental visitODA0.014<0.10PositiveEducation levelDental visitODA0.015<0.10PositiveEducation levelRoot DFTOHB0.015<0.10PositiveResidenceDental prosthesisOHB−0.017<0.10NegativeResidenceDental visitODA−0.015<0.10NegativeNotes: Indirect effects estimated using maximum likelihood (ML) estimator with nonparametric bootstrap (5,000 resamples). A relaxed significance threshold of P <0.10 was adopted for specific indirect effects given the limited statistical power for detecting individual mediation pathways in the current sample. Only specific indirect paths with P <0.10 are shown.Abbreviations: Std. β, standardized coefficient. Table 4Mediation Analysis Results of Direct, Indirect, and Total Effects of Predisposing Factors on Oral Health LiteracyPredisposing FactorsOutcomeEffect TypeβSEPStd. β% MediatedGender (Female)OHBDirect−0.1310.1030.202−0.066Total indirect0.0480.0240.0440.02457.8Total−0.0830.1020.418−0.041ODADirect−0.4710.2150.028−0.116Total indirect0.1370.0540.0110.03441.0Total−0.3340.2130.118−0.082Education level (High education level)OHBDirect0.0560.1050.5910.028Total indirect0.0430.0300.1540.02143.4Total0.0990.1050.3480.049ODADirect0.3110.2090.1370.075Total indirect0.1380.0610.0230.03330.7Total0.4490.2080.0310.108Residence (Rural area)OHBDirect−0.1260.0930.175−0.063Total indirect−0.0460.0260.075−0.02326.7Total−0.1720.0970.074−0.086ODADirect0.0170.1980.9320.004Total indirect−0.1030.0580.076−0.025119.8Total−0.0860.2010.668−0.021Note: Path analysis using maximum likelihood (ML) estimator with nonparametric bootstrap (5,000 resamples). Statistical significance of indirect effects was determined by whether the bootstrap bias-corrected 95% confidence interval (BC 95% CI) excluded zero. % Mediated = |Indirect effect| / |total effect| × 100%. % Mediated values exceeding 100% indicate suppression effects, where the indirect effect exceeds the total effect due to opposing direct and indirect effect directions. Bold P values indicate statistical significance at the 0.05 level.Abbreviations: β, unstandardized coefficient; SE, standard error; Std. β, standardized coefficient. Mediation Analysis Results of Direct, Indirect, and Total Effects of Predisposing Factors on Oral Health Literacy Note: Path analysis using maximum likelihood (ML) estimator with nonparametric bootstrap (5,000 resamples). Statistical significance of indirect effects was determined by whether the bootstrap bias-corrected 95% confidence interval (BC 95% CI) excluded zero. % Mediated = |Indirect effect| / |total effect| × 100%. % Mediated values exceeding 100% indicate suppression effects, where the indirect effect exceeds the total effect due to opposing direct and indirect effect directions. Bold P values indicate statistical significance at the 0.05 level. Abbreviations: β, unstandardized coefficient; SE, standard error; Std. β, standardized coefficient. Table 5Specific Indirect Effects Reaching Marginal Significance (P < 0.10)Predisposing FactorsMediatorOutcomeStd. βPDirectionGenderDental visitODA0.014<0.10PositiveEducation levelDental visitODA0.015<0.10PositiveEducation levelRoot DFTOHB0.015<0.10PositiveResidenceDental prosthesisOHB−0.017<0.10NegativeResidenceDental visitODA−0.015<0.10NegativeNotes: Indirect effects estimated using maximum likelihood (ML) estimator with nonparametric bootstrap (5,000 resamples). A relaxed significance threshold of P <0.10 was adopted for specific indirect effects given the limited statistical power for detecting individual mediation pathways in the current sample. Only specific indirect paths with P <0.10 are shown.Abbreviations: Std. β, standardized coefficient. Specific Indirect Effects Reaching Marginal Significance (P < 0.10) Notes: Indirect effects estimated using maximum likelihood (ML) estimator with nonparametric bootstrap (5,000 resamples). A relaxed significance threshold of P <0.10 was adopted for specific indirect effects given the limited statistical power for detecting individual mediation pathways in the current sample. Only specific indirect paths with P <0.10 are shown. Abbreviations: Std. β, standardized coefficient. Moderation and Multi-Group Analyses In exploratory configural multi-group analyses stratified by education level (Figure 3A and Supplementary Table 4), dental prosthesis use was significantly and positively associated with OHB in the high-education group. Root DFT significantly predicted OHB in the high-education group and ODA in the low-education group. The number of teeth with periodontal pockets was negatively associated with OHB in the low-education group, while dental visit was positively associated with ODA in the high-education group. When stratified by residence (Figure 3B and Supplementary Table 4), both dental prosthesis use significantly and Root DFT both enhanced OHB among urban residents. However, given that no interaction terms reached statistical significance (Supplementary Table 3), these patterns should be interpreted as exploratory observations rather than evidence of statistically confirmed between-group heterogeneity.Figure 3Standardized path coefficients (β, 95% CI) from configural multi-group path analyses, stratified by education level (A) and residence (B). Paths represent effects of clinical and behavioral predictors on oral health literacy.Two forest plots of standardized beta for oral health literacy predictors by education and residence groups.Image A′s forest plot shows standardized beta for eight paths (-0.6 to 0.6) with y-axis labels: Root DFT to OHB/ODA, Dental prosthesis to OHB/ODA, Dental visit to ODA, Sweetened beverage consumption to OHB and periodontal pockets to OHB/ODA. Paths have High/Low education markers, 95% confidence intervals, red for P < 0.05. Point estimates: Root DFT to OHB (0.2-0.3), ODA (0.1-0.2); Dental prosthesis to OHB (0.0-0.3), ODA (0.1); Dental visit to ODA (0.1); Sweetened beverage to OHB (0.0-0.1); periodontal pockets to OHB (-0.2), ODA (0.0-0.1). Image B′s plot, similar paths (-0.50 to 0.75), uses Rural/Urban markers, 95% confidence intervals, red for P < 0.05. Point estimates: Root DFT to OHB (0.1-0.2), ODA (0.1); Dental prosthesis to OHB (0.0-0.2), ODA (0.1); Dental visit to ODA (0.1); Sweetened beverage to OHB (0.0-0.1); periodontal pockets to OHB (-0.1), ODA (0.0-0.1).Abbreviations: DFT, decayed and filled teeth; OHB, oral health belief; ODA, oral disease awareness. Standardized path coefficients (β, 95% CI) from configural multi-group path analyses, stratified by education level (A) and residence (B). Paths represent effects of clinical and behavioral predictors on oral health literacy. Discussion To the authors’ knowledge, this is among the first studies to employ path analysis grounded in Andersen’s behavioral model to systematically investigate the direct and indirect pathways associated with OHL among older adults. The findings provide novel insights beyond conventional regression analyses by delineating how predisposing demographic characteristics shape OHL through complex mediating mechanisms involving clinical oral conditions and health behaviors. Although Andersen’s behavioral model has been widely applied to dental service utilization,23,24 few prior studies have used it within a path-analytic framework to examine OHL itself, and none, to our knowledge, in older Chinese adults. This study extends this literature by jointly modelling predisposing, need, and enabling/behavioral factors. The positive association between dental service utilization and OHL is consistent with reports from other population,25 whereas the indirect role of education and gender observed here adds nuance to studies that examined only direct associations. The overall OHL levels in this population were suboptimal. While awareness of basic oral health concepts such as the cariogenicity of sugar (81.8%) and the importance of oral health (92.7%) was relatively high, knowledge of evidence-based preventive measures—fluoride protection (27.4%) and pit and fissure sealants (23.9%)—was alarmingly low. These findings are consistent with previous reports showing that older adults in China have limited awareness of specific preventive strategies despite acknowledging the general importance of oral health.26,27 The low correct rate for fluoride-related knowledge may reflect generational exposure to oral health education that predated widespread fluoride advocacy in China.28 This knowledge gap represents an actionable target for community health education programs. A key finding was the consistent positive association between Root DFT and both OHB and ODA scores across regression and path analyses. This may appear counterintuitive, as greater disease burden might be expected to correlate with lower health knowledge. The Root DFT index reflects cumulative root caries together with its treatment history (filled teeth) rather than treatment experience alone. One possible interpretation, consistent with these cross-sectional associations, is that the process of seeking and receiving dental treatment for root caries may serve as an opportunity through which individuals acquire oral health knowledge and develop positive beliefs; however, this mechanism cannot be confirmed in the present design. This suggests a possible “experience-driven knowledge acquisition” pattern aligns with the health belief model, which posits that personal susceptibility and disease experience can motivate health-related cognition.29 Previous studies have similarly reported positive associations between treatment experience and health literacy in chronic disease populations.30,31 Clinically, this finding supports the concept of “teachable moments”, suggesting that dental treatment encounters may represent valuable opportunities for clinicians to deliver targeted oral health education.32–34 In addition, dental prosthesis use and dental visit emerged as important direct positive predictors of OHL, though through different domains. Prosthetic treatment experience directly enhanced OHB scores, likely because the prolonged treatment process and follow-up visits involved in prosthodontic rehabilitation necessitate patient-provider communication that reinforces beliefs about oral health importance. Dental visit, in contrast, directly improved ODA scores, suggesting that dental encounters provide exposure to disease-specific information that broadens knowledge of oral pathology and prevention.35,36 These differential effects imply that OHB and ODA, while both components of OHL, are sensitive to different types of clinical experiences and may require distinct educational approaches. Sweetened beverage consumption ≥ 1/day was positively associated with OHB in the Poisson regression but negatively associated in the path model. Because only 10 participants reported this consumption level, this estimate is highly unstable and direction-sensitive to model specification. Larger studies are needed to clarify any association. The mediation analysis revealed that predisposing factors, such as gender and education level, influence OHL not merely through direct channels but through indirect pathways mediated by need and enabling/behavioral factors. Women may face lower direct exposure to oral health information through social or occupational channels,14 yet their generally higher rates of dental service utilization and prosthetic treatment may partially offset this deficit.37,38 This may explain why female gender showed a negative direct effect on ODA but a positive indirect effect through mediators. This suppression effect is particularly informative, as it underscores the importance of path analysis for revealing relationships that would be obscured in conventional regression models. The consistent role of dental visit behavior as a mediator across multiple pathways reinforces the centrality of dental service access in the OHL formation process. This aligns with the “information exposure” hypothesis: dental visits may represent opportunities for receiving health education, such that greater access to dental services could be associated with greater oral health knowledge. As the data are cross-sectional, this temporal, accumulative interpretation remains hypothetical.39 The exploratory multi-group analyses revealed descriptive patterns of variation in path coefficients across subgroups, though formal interaction tests did not confirm statistically significant moderation. The finding that dental prosthesis use was associated with OHB only among highly-educated and urban individuals may be consistent with the possibility that translating clinical experience into health beliefs is facilitated by a foundational level of cognitive resources and health system engagement, although this was not formally confirmed. Highly-educated individuals may be better equipped to process and internalize the oral health information conveyed during prosthodontic treatment,40,41 while urban settings may provide a more conducive environment for sustained patient education through better-resourced dental clinics.38,42 The observation that periodontal pockets were negatively associated with OHB in the low-education group, though not statistically confirmed as a between-group difference, warrants further investigation. Unlike root caries, which typically involves a discrete treatment event that may trigger knowledge acquisition, periodontal disease progresses insidiously and is often asymptomatic in early stages.43 This may provide fewer “teachable moments” for patients without sufficient baseline knowledge to recognize its significance. This finding has clear clinical implications: periodontal disease management for low-education patients should be accompanied by proactive, adapted health education that addresses misconceptions and promotes self-efficacy. However, these exploratory findings should be interpreted with caution and require confirmation in future studies with adequate sample sizes for formal moderation testing. This study has several strengths. First, the use of path analysis within a theoretically grounded framework of Andersen’s behavioral model enabled a more nuanced understanding of the complex interrelationship among predisposing, need, and enabling/behavioral factors. Second, the nationally representative sampling design and the inclusion of both clinical and questionnaire data enhance the generalizability and comprehensiveness of the findings. Third, the multi-group analyses provided exploratory insights into potential heterogeneity across population subgroups, providing a foundation for personalized intervention design. However, several limitations should be acknowledged. First, the cross-sectional design precludes causal inference. The identified pathways should be interpreted as associational structures rather than definitive causal mechanisms. Longitudinal studies are still needed to confirm the temporal sequence of the relationships. Second, the explained variance for both OHL outcomes was modest, indicating that additional unmeasured factors—such as media exposure and social networks—likely contribute to OHL variation. Third, OHL was assessed using a self-reported instrument with dichotomous scoring, which may have limited sensitivity in capturing the full spectrum of oral health knowledge. Fourth, the sample was derived from 13 surveillance sites, which, although representative at the national level, may not capture the full diversity of China’s regional socioeconomic and cultural landscapes. Fifth, the exploratory multi-group analyses were conducted on subgroups with limited sample sizes, and no interaction terms reached statistical significance. These multi-group analyses were underpowered for formal moderation testing. Thus, the observed subgroup-specific patterns should therefore be regarded as hypothesis-generating rather than confirmatory. Future studies should expand the geographic scope and include a broader age range. In addition, the statistical power for detecting individual specific indirect effects may have been limited in the current sample, so these pathway-specific mediation findings should be interpreted as exploratory and require confirmation in future studies with larger samples. The findings yield several targeted clinical recommendations. First, dental professionals could consider treating routine treatment encounters as potential educational opportunities and, where feasible, incorporating oral health messaging into clinical workflows.44 Second, these findings suggest that gender-sensitive educational approaches could be explored; for example, materials addressing disease-awareness gaps might be offered to female patients during clinical encounters. Third, education-stratified strategies are essential. For low-education populations, clinicians should prioritize simplified, visually oriented health education materials and proactively address the risk of untreated periodontal disease. For high-education populations, the focus could be on using treatment experiences to reinforce preventive behaviors. Fourth, rural oral health programs may consider expanding dental service access. Because the relevant subgroup interaction tests were not statistically significant, the suggestion that limited access is a key barrier for rural older adults should be regarded as a hypothesis requiring confirmation in adequately powered studies. Conclusion This study suggests that, among older Chinese adults, Root DFT and dental service utilization were associated with higher oral health literacy, while gender and education level were indirectly associated with oral health literacy through the pathways examined. Dental treatment encounters may provide opportunities to promote oral health literacy, particularly by addressing gaps in knowledge of preventive measures such as fluoride use and pit and fissure sealants. Exploratory subgroup findings also suggest that tailored oral health education strategies for different education and residence groups warrant further study.