Association between longitudinal trajectories of cognitive functioning and all-cause mortality among middle-aged and older adults in South Korea: Exploring the mediating role of oral health-related quality of life

Abstract

The associations between distinct cognitive function trajectories and mortality risk, as well as the potential mediating pathway underlying this relationship, have not been fully elucidated. This cohort study explored the association between cognitive function trajectories and all-cause mortality and the mediating role of oral health-related quality of life (OHRQoL). A nationwide cohort of 5486 individuals aged ≥45 years was included in this study. Longitudinal courses of cognitive function measured using the Mini-Mental State Examination (MMSE) were classified using growth mixture modeling. The OHRQoL was measured using the Geriatric Oral Health Assessment Index. The association between cognitive function trajectories and all-cause mortality was determined using a Cox model. Mediation analyses were conducted to evaluate the mediating role of OHRQoL in this association. Hazard ratios (HRs) and 95% confidence intervals (CIs) were computed. Four trajectories of cognitive function were identified: high-stable (n = 4469; 81.5%), moderate-increasing (n = 418; 7.6%), decreasing (n = 454; 8.3%), and low-stable (n = 145; 2.6%). Compared to those experiencing high-stable trajectory, individuals experiencing decreasing (HR = 1.39, 95% CI = 1.08–1.78) and low-stable (HR = 1.84, 95% CI = 1.31–2.57) trajectories had an elevated risk for all-cause mortality. The mediation analyses showed that OHRQoL accounted for 14.6% and 11.5% of the elevated mortality risk among decreasing (indirect HR = 1.04; 95% CI = 1.01–1.08) and low-stable trajectories (indirect HR = 1.06; 95% CI = 1.01–1.11) compared to the high-stable trajectory, respectively. While modest, a moderating role of poor OHRQoL on the association between decreasing or low-stable cognitive function trajectories and all-cause mortality was observed. However, considering the uncertainty in the estimated mediation proportion, these findings should be interpreted with caution and require confirmation in future studies.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11357-026-02398-2.

Introduction

Maintaining cognitive function is essential for promoting healthy aging [1]. According to the Global Burden of Disease Study, dementia affects approximately 57 million people worldwide and is associated with 2.0 million deaths in 2021 [2]. The health burden related to dementia has been recognized as an important public health concern in South Korea owing to rapid population aging. In South Korea, approximately one in ten older adults, aged ≥65 years, live with dementia [3]. Furthermore, the number of deaths attributed to dementia has increased from 8.7 thousand in 2013 to 14.4 thousand in 2023, underscoring the need for policies and initiatives to address cognitive health among older adults [4].

Poor cognitive function is a well-documented risk factor of all-cause mortality. For instance, previous meta-analyses showed that poor cognitive function is linked to increased mortality risk [5, 6]. The Mini-Mental State Examination (MMSE), one of the most widely used measures of cognitive function, has consistently served as a predictor of all-cause mortality [7, 8]. Complex mechanisms may underpin the pathways linking poor cognitive function and mortality. For instance, individuals with poor cognitive function are susceptible to cachexia, malnutrition, and pneumonia, which are risk factors for mortality [9].

Oral health–related quality of life (OHRQoL) is a multi-faceted concept that evaluates one’s subjective well-being and encompasses the physical, functional, emotional, and social domains related to oral health [10]. Some previous studies have linked cognitive function to OHRQoL [11–14]. For instance, a study of Korean middle-aged or older adults indicated that the MMSE score was positively associated with OHRQoL [14]. Moreover, a longitudinal study of English adults showed that cognitive function was positively associated with subsequent OHRQoL [15]. Various pathways may be linked to poor cognitive function and OHRQoL. Individuals with impaired cognitive function may be more likely to exhibit poor oral hygiene, neglect oral health problems, and encounter barriers to accessing oral healthcare services, which, in turn, contributes to the deterioration of OHRQoL [16, 17].

Previous studies have found that poor OHRQoL is linked to an increased risk of all-cause mortality. A cohort study in Japan showed that the OHRQoL was inversely associated with the risk of all-cause mortality [18]. Similarly, OHRQoL is inversely associated with the risk of all-cause mortality among middle-aged or older adults in South Korea [19]. Poor OHRQoL may lead to increased mortality risk via various pathways, such as poor nutritional status, systemic inflammation, and social exclusion [20].

Despite the established associations between cognitive function, OHRQoL, and mortality, the mediating role of OHRQoL in the relationship between cognitive function and all-cause mortality remains poorly understood. Additionally, recent studies have shown varied trajectories of cognitive function over time and their relationship with mortality risk, rather than relying solely on cognitive function assessed at a single time point [7, 21, 22]. Examining the patterns of cognitive change over time may provide more informative insights into subsequent mortality risk. Therefore, this study aimed to explore the association between cognitive function trajectories and all-cause mortality in middle-aged and older adults and the mediating role of OHRQoL in this relationship.

Methods

Study participants

The study analyzed data from participants of the 2006–2022 Korean Longitudinal Study of Aging (KLoSA). To collect information on the health and social status of middle-aged and older adults, approximately 10,000 individuals, aged ≥45 years, were recruited nationwide across South Korea through systematic sampling in 2006 (1st wave) [23]. Subsequently, participants were followed up every 2 years until 2022 (9th wave). For each survey wave, information was collected by trained interviewers through household surveys. Survey items related to cognitive function have been collected since 2006, while those related to ORHQoL have been collected since 2018. The study design of the present study is illustrated in Fig. 1. The study variables were measured to ensure a longitudinal sequence between exposure (2006–2016), mediator (2018), and mortality (through 2022). The sample selection process is illustrated in Fig. 2. Initially, 10,041 adults provided valid information on cognitive function in 2006 (1st wave). Among these participants, individuals with missing information on baseline sociodemographic characteristics were excluded. Subsequently, those who were lost to follow-up or died before 2018 were also excluded. After further excluding individuals with missing information on OHRQoL, the final sample consisted of 5486 adults. Written informed consent was provided by all individuals. The study was approved by the Institutional Review Board of Severance Hospital (no. 4–2025-1535).

Fig. 1.

Study design (GOHAI, Geriatric Oral Health Assessment Index)

Fig. 2.

Participant selection flowchart

Variables

Cognitive function

Cognitive function was evaluated using the Korean version of the MMSE. The validity and reliability of the MMSE in the Korean population have been established in a previous study [24]. The total MMSE score ranges from 0 to 30, with higher scores indicating better cognitive functioning.

OHRQoL

The OHRQoL was evaluated using the Geriatric Oral Health Assessment Index (GOHAI), a validated measure for assessing OHRQoL in the Korean population [25]. The total GOHAI score ranges from 0 to 60, with higher scores indicating a better OHRQoL.

Mortality

The outcome variable was the all-cause mortality. For those who died, the date and cause of death were determined through detailed exit interviews with the participant’s spouse, family members, or other close relatives. The follow-up time was defined as the period from the baseline examination in 2018 to the date of death for deceased participants or to the date of the most recent follow-up survey.

Control variables

The following baseline characteristics measured in 2006 were adjusted for: sex, age, educational attainment (elementary school or below, middle school, high school or above), income (Q1–Q4), marital status (married, not married), smoking status (non-smoker, current smoker), physical activity (<150 min/week, ≥150 min/week), chronic conditions (none, one, two, or more). Income was divided into four quartiles based on household income for each year. Chronic conditions were categorized on the basis of the presence of hypertension, diabetes mellitus, cancer, pulmonary disease, liver disease, or cardiovascular disease.

Data analysis

GMM

The 10-year trajectory of cognitive function from 2006 to 2016 was identified using the growth mixture model (GMM). For the preliminary analysis, the optimal number of latent trajectories was selected by sequentially fitting the GMMs with different numbers of trajectories, from one to five. For comparison, model fit statistics including the Akaike information criterion (AIC), Bayesian information criterion (BIC), sample size-adjusted BIC (SABIC), Lo-Mendell-Rubin adjusted likelihood ratio test (LMR-LRT), parametric bootstrapped likelihood ratio test (BLRT), average posterior probability (AvePP), and entropy were computed for each GMM. The optimal number of latent trajectories was selected using pre-specified criteria [26], including low AIC, BIC, and SABIC values, minimum AvePP > 0.7, entropy > 0.8, and the smallest proportion of the class > 1.0%. Additionally, the p-value for LMR-LRT or BLRT > 0.05 was interpreted as favoring a GMM with one fewer class. Missing values in the MMSE during the study period were handled using the full-information maximum likelihood method in the GMMs. In addition, a sensitivity analysis was conducted using complete cases with MMSE scores available at all survey waves from 2006 to 2016 (n = 4453).

Cox regression

After determining the optimal GMM, a Cox regression model was used to determine the association between the cognitive function trajectories and mortality. Both unadjusted and fully adjusted models were fitted, and hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated. The proportional hazard assumption was confirmed using the Schoenfeld residual test. The R package “survival” was used for Cox regression analyses.

Mediation analysis

An explanatory mediation analysis was performed to examine the potential role of OHRQoL in the relationship between the trajectories of cognitive function and mortality. A counterfactual-based mediation model was employed to decompose the total effect (TE) of the cognitive function trajectories of all-cause mortality into natural direct effects (NDE) and natural indirect effects (NIE). Linear regression was employed to determine the association between cognitive function trajectories and OHRQoL, whereas Cox models were employed to examine the relationship between OHRQoL and all-cause mortality. In total, 5000 bootstrap resamplings were performed to calculate the NDE, NIE, and mediated proportions. The association among TE, NDE, and NIE was expressed as ${\mathit{HR}}^{\mathit{TE}}={\mathit{HR}}^{\mathit{NDE}}\times {\mathit{HR}}^{\mathit{NIE}}$ the proportion mediated by the excess HR attributable to the indirect pathway was calculated as follows: $\text{proportion}\text{mediated}={(HR}^{\mathit{NDE}})\times {(HR}^{\mathit{NIE}}-1)/({\mathit{HR}}^{\mathit{TE}}-1)$. Further methodological details have been provided in the literature [28]. The R package “CMAverse” was employed for the mediation models. All data analyses were conducted using R (version 4.5.2). For sensitivity analysis, an additional analysis was conducted using complete cases with MMSE scores available at all time points from 2006 to 2016 (n = 4453). To further explore whether specific dimensions of OHRQoL contributed to the observed mediation, we conducted additional mediation analyses using the three GOHAI subdomain scores—physical function (4 items), pain/discomfort (3 items), and psychosocial function (5 items)—as separate mediators in each model. These analyses were performed to determine whether the estimated mediation was primarily attributable to specific aspects of OHRQoL or was relatively consistent across the different domains. Additionally, to fully utilize the baseline study population enrolled in 2006, we employed a time-dependent Cox regression model that incorporated cognitive function measured across repeated survey waves as a time-varying independent variable. Unlike the primary analysis, this approach did not require the exclusion of participants who died before 2018, thereby minimizing the potential for survivor bias. After excluding observations with missing values, 9449 individuals were included in this sensitivity analysis.

Results

Preliminary analysis

The baseline (2006) characteristics of the included and excluded participants are presented in Table S1. Compared with participants included in the final analytic sample, those excluded from the analysis were more likely to be male, older, have lower levels of income and education, have chronic conditions, and have lower mean MMSE scores. Additionally, the mortality rate was higher among the excluded participants than among those included in the final analytic sample. The model fit statistics across the GMMs with different numbers of classes are presented in Table S2. Although AIC, BIC, and SABIC decreased as the number of classes increased, the p-value for the LMR-LRT was 0.136, which indicates a preference for the four-class models. The four-class model also met the criteria for entropy, AvePP, and proportion of the smallest classes; therefore, the four-class GMM was chosen as the final model.

GMM

The final GMM model yielded four cognitive function trajectories from 2006 to 2016 (Fig. 3).

1. High-stable group (n = 4469; 81.5%): This group, representing the majority of study participants, was characterized by consistently high cognitive function, maintaining mean MMSE scores at approximately 27–28 throughout the follow-up period.

2. Moderate-increasing group (n = 418, 7.6%): This group was characterized by relatively low cognitive function in 2006 (MMSE score, 19–20), followed by a gradual increase in MMSE scores over the follow-up period.

3. Decreasing group (n = 454, 8.3%): This group was characterized by a relatively high baseline cognitive function in 2006 (MMSE score: 25), followed by a marked decline over time.

4. Low-stable group (n = 145, 2.6%): This group was characterized by persistently low cognitive function (MMSE score: 12–14) with a slight downward trend during follow-up.

Fig. 3.

The MMSE scores from 2006 to 2016 (MMSE, Mini-Mental State Examination)

Descriptive analysis

The sample characteristics at baseline (2006) are presented in Table 1. Compared with the high-stable group, those with a decreasing and low-stable cognitive function trajectory were more likely to be women, older, with lower education and income levels, not married, and had chronic conditions. The mean OHRQoL score in 2018 was 39.55 for the high-stable group, 33.35 for the moderate-increasing group, 32.62 for the decreasing group, and 30.12 for the low-stable group.

Table 1.

Baseline participant characteristics (2006) and OHRQoL (2018)

	Overall	Cognitive function trajectory
		High-stable	Moderate-increasing	Decreasing	Low-stable
	N = 5,486	N = 4,469	N = 418	N = 454	N = 145
Sex
Male	2,319 (42.3)	2,088 (46.7)	77 (18.4)	139 (30.6)	15 (10.3)
Female	3,167 (57.7)	2,381 (53.3)	341 (81.6)	315 (69.4)	130 (89.7)
Age (years)
Mean (SD)	58.86 (9.11)	56.95 (8.23)	66.07 (8.02)	66.88 (7.58)	71.63 (7.99)
Education
Elementary school or below	2342 (42.7)	1496 (33.5)	365 (87.3)	341 (75.1)	140 (96.6)
Middle school	995 (18.1)	910 (20.4)	25 (6.0)	58 (12.8)	2 (1.4)
High school or above	2149 (39.2)	2063 (46.2)	28 (6.7)	55 (12.1)	3 (2.1)
Income
Q1	1178 (21.5)	779 (17.4)	163 (39.0)	170 (37.4)	66 (45.5)
Q2	1471 (26.8)	1129 (25.3)	154 (36.8)	147 (32.4)	41 (28.3)
Q3	1769 (32.2)	1549 (34.7)	80 (19.1)	112 (24.7)	28 (19.3)
Q4	1068 (19.5)	1012 (22.6)	21 (5.0)	25 (5.5)	10 (6.9)
Marital status
Married	4672 (85.2)	3972 (88.9)	295 (70.6)	335 (73.8)	70 (48.3)
Not married	814 (14.8)	497 (11.1)	123 (29.4)	119 (26.2)	75 (51.7)
Smoking status
Non-smoker	4492 (81.9)	3592 (80.4)	376 (90.0)	392 (86.3)	132 (91.0)
Current smoker	994 (18.1)	877 (19.6)	42 (10.0)	62 (13.7)	13 (9.0)
Physical activity
<150 min/week	3892 (70.9)	3061 (68.5)	350 (83.7)	350 (77.1)	131 (90.3)
≥150 min/week	1594 (29.1)	1408 (31.5)	68 (16.3)	104 (22.9)	14 (9.7)
Chronic conditions
None	3683 (67.1)	3123 (69.9)	231 (55.3)	262 (57.7)	67 (46.2)
One	1371 (25.0)	1043 (23.3)	128 (30.6)	146 (32.2)	54 (37.2)
Two or more	432 (7.9)	303 (6.8)	59 (14.1)	46 (10.1)	24 (16.6)
OHRQoL (2018)
Mean (SD)	38.26 (8.92)	39.55 (8.37)	33.35 (9.64)	32.62 (8.65)	30.12 (8.05)

MMSE mini-mental state examination, SD standard deviation, OHRQoL oral health-related quality of life

Cox models

The Kaplan–Meier curve for all-cause mortality across cognitive function trajectories is shown in Fig. 4. Table 2 shows the results of Cox regression models. In the adjusted model, compared to the high-stable group, the HRs (95% CIs) for all-cause mortality were 0.93 (0.68–1.25) for moderate-increasing group, 1.39 (1.08–1.78) for decreasing group, and 1.84 (1.31–2.57) for low-stable group, respectively.

Fig. 4.

Kaplan–Meier curve

Table 2.

The longitudinal trajectory of cognitive function and all-cause mortality during the follow-up

	N of exposed	N of death	Model 1	Model 2
			HR (95% CI)	HR (95% CI)
Cognitive function trajectory
High-stable group	4469	308	Reference	Reference
Moderate-stable group	418	61	2.23 (1.70–2.94)	0.93 (0.68–1.25)
Decreasing group	454	100	3.57 (2.85–4.47)	1.39 (1.08–1.78)
Low-stable group	145	57	7.24 (5.45–9.60)	1.84 (1.31–2.57)
Sex
Male	2319	266		Reference
Female	3167	260		0.53 (0.42–0.67)
Age (years)
Mean (SD)	5486	526		1.13 (1.12–1.15)
Education
Elementary school or below	2342	337		Reference
Middle school	995	76		1.17 (0.89–1.54)
High school or above	2149	113		0.93 (0.72–1.21)
Income
Q1	1178	161		Reference
Q2	1471	189		1.20 (0.97–1.49)
Q3	1769	127		1.01 (0.80–1.29)
Q4	1068	49		0.96 (0.69–1.34)
Marital status
Married	4672	380		Reference
Not married	814	146		1.27 (1.01–1.59)
Smoking status
Non-smoker	4492	409		Reference
Current smoker	994	117		1.43 (1.14–1.81)
Physical activity
<150 min/week	3892	413		Reference
≥150 min/week	1594	113		0.72 (0.58–0.90)
Chronic conditions
None	3683	256		Reference
One	1371	188		1.27 (1.05–1.54)
Two or more	432	82		1.60 (1.24–2.06)

HR hazard ratio, CI confidence interval

Model 1, unadjusted model; Model 2, Model 1 + covariates

^aper 1,000 person-years

Mediation analysis

Figure 5 presents the results of the mediation model. Compared to the high-stable group, the decreasing cognitive function trajectory was inversely associated with OHRQoL (β: −3.72, 95% CI: −4.55 – −2.88). Additionally, the OHRQoL was inversely associated with all-cause mortality (HR: 0.98, 95% CI: 0.97–0.99). The HRs (95% CI) for indirect effects were 1.04 (1.01–1.08), accounting for 14.6% (95% CI: 1.0%–58.8%) for the total association. Similarly, compared with the high-stable group, the low-stable cognitive function trajectory was inversely associated with OHRQoL (β: −6.17, 95% CI: −4.74 – −3.32). The HRs (95% CI) for indirect effects were 1.06 (1.01–1.11), accounting for 11.5% (95% CI: 1.9%–29.7%) for the total association.

Fig. 5.

The mediating role of oral health-related quality of life (OHRQoL) in the relationship between cognitive function trajectories and all-cause mortality (HR, hazard ratio; CI, confidence interval)

The sensitivity analysis using the complete cases (Table S3) also produced similar mediating role of OHRQoL in the association between the decreasing or low-stable group and mortality risk compared with those from the main analyses (indirect HR: 1.05, 95% CI: 1.00–1.10 for the decreasing group and HR: 1.05, 95% CI: 1.00–1.11 for the low-stable group). When each of the three GOHAI domains was analyzed separately, the adjusted HRs for the NIE were comparable across domains, suggesting that the observed mediation was relatively consistent across the physical function, pain/discomfort, and psychosocial domains rather than being attributable to a specific dimension of OHRQoL (Table S4). Additionally, the results from the time-dependent Cox model showed a significant inverse association between cognitive function (MMSE score) and the mortality risk (adjusted HR: 0.94, 95% CI: 0.93–0.94; Table S5).

Discussion

This study identified four distinct trajectories of cognitive functioning among middle-aged and older adults in South Korea, namely, high-stable, moderate-increasing, decreasing, and low-stable groups. Compared with the high-stable group, those with decreasing and low-stable cognitive function trajectories had a significantly higher mortality risk. Furthermore, poor OHRQoL partially accounted for the increased mortality risk observed in those with decreasing and low-stable cognitive function trajectories. However, the estimated mediation proportions were modest (<15%), with wide CIs, indicating considerable uncertainty regarding the magnitude of the mediating role of OHRQoL. Therefore, these findings should be interpreted with caution.

These findings are consistent with those of prior studies showing associations between cognitive function trajectories and mortality risk. For instance, low and declining cognitive function trajectories were linked to an increased risk of all-cause mortality among middle-aged and older adults in Spain [7]. Cohort studies of older Chinese adults also showed that declining and low-stable cognitive function trajectories were linked to an elevated risk of all-cause mortality [21, 27]. Additionally, OHRQoL (measured via GOHAI) was inversely associated with all-cause mortality risk among adults in Japan [18] and South Korea [19]. Conversely, the association between cognitive function and OHRQoL is inconsistent. While some studies found a positive association between cognitive function and OHRQoL [14, 15], findings from a meta-analysis indicated no significant differences in GOHAI-measured OHRQoL between individuals with or without Alzheimer’s disease (standardized mean difference: 0.09, 95% CI −0.66–0.85) [20]. In contrast to previous studies, our study provides novel insights into how the longitudinal trajectories of cognitive function are associated with OHRQoL.

Complex mechanisms may underpin the relationship between cognitive function trajectories and OHRQoL. Poor cognitive function trajectories may be associated with unfavorable oral hygiene behaviors and inadequate oral disease management. Poor cognitive function was associated with infrequent tooth brushing in older Chinese adults [28]. A previous study of middle-aged or older adults in the United States reported that those with subjective cognitive decline were less likely to visit dental healthcare services [29]. Moreover, a mediation analysis indicated that oral hygiene behaviors can serve as mediators between poor cognitive function and oral health status among older adults [16]. Therefore, unfavorable oral health behaviors among individuals with poor cognitive function may constitute a pathway leading to dental caries or poor OHRQoL in older adults [30].

The relationship between OHRQoL and mortality risk may be mediated by complex biological or psychosocial mechanisms. A previous meta-analysis reported that poor OHRQoL, as measured by the GOHAI, increased the risk of malnutrition [31], which in turn may contribute to physical frailty and heightened mortality risk [32]. Additionally, another study demonstrated that poor OHRQoL may be linked to systemic inflammation, potentially increasing the risk of cardiovascular disease among older adults in South Korea [33]. Psychosocially, poor OHRQoL may be linked to poor mental health, including loneliness and depression by inducing low social participation [34, 35], a potential risk factor for subsequent mortality.

This study has some limitations. First, despite the longitudinal nature of this study, the causal relationships between cognitive function trajectories, OHRQoL, and mortality could not be fully determined. For instance, certain factors, such as genetic predisposition, history of neuropsychiatric disorders and dental diseases, and earlier-life trajectories of cognitive function, could not be analyzed due to a lack of data. Additionally, individuals with declining or persistently low cognitive function may experience deterioration in socioeconomic status and health conditions during the follow-up. Although our analyses adjusted for baseline covariates measured in 2006, they did not account for changes in these factors over time. Therefore, time-varying socioeconomic and health conditions may have acted as unmeasured confounders, potentially influencing the observed associations. The standard causal mediation analysis relies on several assumptions, including the absence of unmeasured confounding for the exposure–mediator, exposure–outcome, and mediator–outcome relationships. Because information on OHRQoL before 2018 was unavailable, the temporal sequence between cognitive decline and OHRQoL could not be fully established. Therefore, these assumptions cannot be fully verified in our study. Our mediation analysis should be interpreted as a decomposition of associations, rather than a representation of causal effects. Second, the trajectory analysis necessitated the inclusion of only individuals who survived between 2006 and 2018, which may have introduced a risk of survival bias. Specifically, compared with the included sample, the excluded participants were older, had poorer baseline health status, had a higher mortality rate during the follow-up. Therefore, the exclusion of these participants may have introduced a healthy survivor bias, which should be considered when interpreting the study findings. The selection bias may have contributed to an underestimation of the relationship between cognitive function and mortality risk. Third, specific causes of mortality were not recorded using standardized codes (e.g., International Classification of Disease). Therefore, the associations between cognitive trajectory, OHRQoL, and cause-specific mortality could not be examined. This limitation precluded a more in-depth investigation of the mechanisms linking poor OHRQoL to mortality, including cardiovascular, nutritional, and infectious pathways. Future studies incorporating cause-specific mortality data may provide greater insight into the biological and clinical mechanisms underlying this association. Fourth, the MMSE and GOHAI were primarily developed as screening tools and may be subject to recall bias. Finally, the sample included only middle-aged or older adults in South Korea. Therefore, the findings of this study may not necessarily be generalizable to individuals from other regions. Fifth, the direct application of our findings to clinical practice may be limited because accurate classification into the cognitive trajectory groups identified in this study requires repeated cognitive assessments over a prolonged period. Nevertheless, the objective of this study was to provide epidemiological evidence regarding the association between long-term patterns of cognitive decline, OHRQoL, and mortality. Our findings suggest that individuals exhibiting persistently low cognitive function or evidence of rapid cognitive decline, even without 10 years of follow-up data, may benefit from closer monitoring and targeted interventions aimed at maintaining oral health and reducing subsequent mortality risk.

Despite its limitations, this study is among the first to identify varying typologies of cognitive function trajectories among community-dwelling middle-aged or older individuals and to explore their relationship with mortality risk. Furthermore, the mediating role of OHRQoL in this relationship, which is understudied in existing literature, may offer novel insights into the pathways through which poor cognitive function contributes to mortality risk.

Conclusion

This study identified four distinct cognitive functional trajectories in middle-aged and older individuals in South Korea. Compared to those with the high-stable trajectory, those with decreasing and low-stable cognitive function trajectories exhibited an increased mortality risk. Furthermore, a moderating role of poor OHRQoL on the association between decreasing or low-stable cognitive function trajectories and all-cause mortality was identified in this study. However, given the uncertainty in the estimated mediation proportion, this finding should be interpreted with caution and warrants further investigation in future studies with larger sample sizes and cause-specific mortality data.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

Seong-Uk Baek, conceptualization, methodology, investigation, software, visualization, writing-original draft preparation; Jin-Ha Yoon, conceptualization, supervision, writing-review and editing.

Funding

None to declare.

Data availability

Data is available at https://survey.keis.or.kr/eng/index.jsp.

Declarations

Ethics approval

Written informed consent was provided by all individuals. The study was approved by the Institutional Review Board of Severance Hospital (no. 4–2025-1535). This study was conducted in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.

Conflict of interest

The authors declare no competing interests.