Understanding Cognitive Trajectories in Middle-Aged and Older Cancer Survivors: An Analysis of the Korean Longitudinal Study of Aging

Article information

J Korean Acad Fundam Nurs. 2025;32(4):507-519

Publication date (electronic) : 2025 November 30

doi : https://doi.org/10.7739/jkafn.2025.32.4.507

Mi Sook Jung¹⁾, Munkyung Park²⁾, Kyeongin Cha³⁾, Xirong Cui³⁾, Nondumiso Satiso Dlamini⁴⁾, Ah Rim Lee^5),

¹⁾Professor, College of Nursing, Chungnam National University, Daejeon, Korea

²⁾Associate Professor, College of Nursing, Chungnam National University, Daejeon, Korea

³⁾Doctoral Student, College of Nursing, Chungnam National University, Daejeon, Korea

⁴⁾Doctoral Student, University of the Witwatersrand, Gauteng Province, South Africa

⁵⁾Doctoral Student, College of Nursing, Chungnam National University, Daejeon, Korea

Corresponding author: Lee, Ah Rim College of Nursing, Chungnam National University 266, Munhwa-ro, Jung-gu, Daejeon 35015, Korea Tel: +82-42-580-8410, Fax: +82-42-580-8309, E-mail: rimee0901@gmail.com

*This study was supported by Chungnam National University and the National Research Foundation of Korea (No. RS-2021-NR059632).

Received 2025 July 29; Revised 2025 September 18; Accepted 2025 November 16.

Abstract

INTRODUCTION

Cancer continues to pose a global health challenge, with older adults shouldering a disproportionate share of its burden. Globally, an estimated 18.1 million people were newly diagnosed with cancer and 9.9 million individuals died in 2020, with older adults representing 62.3% of all cases and 71.2% of cancer-related deaths [1]. In South Korea, approximately 5.0% of the total population were cancer survivors as of 2022, and nearly one in every seven individuals aged 65 or older fell into this group [2]. Ongoing advances in screening and treatment have contributed to a substantial improvement in survival, with the five-year survival rate for those diagnosed between 2018 and 2022 reaching 72.9%. This improvement marked an 18.7%- point increase over the previous 20 years [2]. At the same time, South Korea has rapidly transitioned to a super-aged society, with adults aged 65 and older reaching 20.3% of the population in 2025. This figure is forecasted to climb 30.9% by 2036 [3]. The demographic shift, in combination with older adults comprising an increasingly large proportion of cancer survivors, has created a pressing need to broaden survivorship strategies, particularly in addressing the long-term effects of various treatments.

As cancer survivorship is increasingly understood as a continuum in which each phase presents unique challenges, the need to preserve cognitive health has been one of the critical issues in long-term care and public health [4]. Reduced cognitive function has been identified as a potential long-term neurotoxic effect of cancer treatment and an unintended consequence of extended survival [5]. Although the reported prevalence of post-treatment cognitive changes varies by treatment trajectory, modality, and method of assessment, up to 75% of survivors experience difficulties performing simple and complex tasks compared to their pre-cancer level of functioning [6]. In a study of 228 older survivors, 32% exhibited cognitive impairment within one year of diagnosis, and more than half of them continued to show deficits at the two-year follow-up [7]. Notably, impairments in attention, memory, and exec-utive process have been observed in a subset of survivors, with some deficits persisting for up to 20 years following treatment [8]. While cognitive decline is most pronounced in older adults, evidence also supports that younger and middle-aged survivors are vulnerable to treatment-related cognitive difficulties [6,8]. These difficulties may be com-pounded by psychological distress and changes in social roles and could increase survivors’ long-term risk of cognitive impairment as they transition into older age [6].

Reduced cognitive function among middle-aged and older cancer survivors can extend beyond subjective com-plaints, compromising their ability to maintain functional independence and promote healthy aging throughout the long-term survivorship phase [9]. Such impairments can lead to loss of autonomy, declines in self-efficacy, and dis-ruption of social roles, collectively diminishing survivors’ overall quality of life [9]. Specifically, cognitive impairment disrupts instrumental activities of daily living such as medication management, financial planning, and maintaining social relationships [9]. Although these difficulties were initially attributed to normal aging or generalized fatigue, a growing body of research supports that the cognitive impairment experienced by cancer survivors is partly driven by treatment-related biological mechanisms including chronic inflammation, oxidative stress, and alter-ations in neural circuitry [5,10].

Changes in cognitive function after cancer treatment have been conceptualized as the outcome of multiple interaction influences beyond therapy itself. Cognitive impairment among cancer survivors may not be limited to subtle post-treatment symptoms but could instead signal a clinically meaningful trajectory toward age-related neuro-degeneration including mild cognitive impairment or dementia [10,11]. Recent research indicates that post-treat-ment cognitive impairment arises from a complex inter-play of treatment-induced neurotoxicity, comorbid con-ditions— such as neurodegenerative diseases, cardiovascular dysfunction, and chronic psychological stress— and individual vulnerabilities including age, genetic predis-position, socioeconomic status, and behavioral determinants [10-13]. Given this convergence of cancer-and age-related risk factors, cognitive changes observed in cancer survivors should be viewed through the lens of shared mechanisms with age-related cognitive impairment. However, much of the existing literature has often relied on narrowly focused, cross-sectional studies and linear analytic approaches, which may not fully capture the dynam-ic, multifactorial nature of cognitive change, possibly driven by the interactions among these interrelated factors [10,11].

Determinants of cognitive change can be conceptualized within three interrelated domains based on prior research [10-14]. Health status, including comorbid conditions such as hypertension, diabetes, and cerebrovascular disease, contributes to cognitive aging through vascular and metabolic mechanisms [11,14]. Lifestyle behaviors such as physical activity, smoking cessation, and healthy nutrition enhance brain plasticity and resilience [13,14]. Psychosocial factors, including depression and social support, shape stress regulation and engagement in healthy practices and represent critical pathways through which cognitive changes may develop over time [10,12,14]. Collectively, these domains reflect the multifactorial nature of cognitive change and underscore the importance of in-tegrating biological, behavioral, and psychosocial per-spectives in survival research.

To better elucidate the mechanisms underlying long-term cognitive change among cancer survivors, the appli-cation of machine learning approaches has been suggested as a valuable tool for modeling such complexity [14]. Recent studies have employed techniques such as random forests, neural networks, and gradient-boosting models to develop predictive models of cognitive function [7,15,16]. These approaches captured complex patterns across clinical, genetic, and psychosocial variables, demonstrating improved predictive accuracy over traditional methods. However, the machine learning models have also shown limitations regarding stability, generalizability, and clinical applicability across heterogeneous cancer populations with diverse socio-clinical profiles [14]. Therefore, further validation using large-scale, longitudinal datasets that reflect real-world complexity is essential for refining machine learning models predicting cognitive impairment among cancer survivors.

The Korean Longitudinal Study of Aging (KLoSA) is a nationally representative panel survey conducted bien-nially with individuals aged 45 years and older [17]. It col-lects extensive information on cognitive status, chronic conditions, health behaviors, and psychosocial indicators across multiple waves, enabling long-term assessment of cognitive changes and their multifactorial determinants [17]. KLoSA provides several advantages, including national representativeness, repeated longitudinal assessments, a focus on middle-aged and older adults, comprehensive coverage of health and social domains, and har-monization with international aging studies that facilitate cross-national comparisons of aging patterns. Leveraging these strengths, the present study utilized data from the KLoSA to characterize longitudinal trajectories of cognitive function and to explore key determinants of long-term decline among cancer survivors.

METHODS

2. Study Sample

KLoSA is an extensive, ongoing panel survey designed to collect comprehensive information on the aging process among adults in South Korea. The survey targets com-munity-dwelling individuals aged 45 years and older, excluding residents of Jeju Island, and uses a multistage stratified sampling framework based on geographic re-gion and housing type, with sampling frames derived from the 2005 Population and Housing Census. Data from waves 1 (2006) through 7 (2018) were used for the present analysis. The KLoSA dataset includes a wide range of variables across demographics, family structure, health, em-ployment, income, assets, and quality of life indicators. At baseline, 10,254 participants were enrolled, and 6,940 individuals remained by wave 7 [17].

For the current study, a subsample of older adults who self-reported a cancer diagnosis was identified from the KLoSA cohort. Participants were eligible for inclusion if they reported a physician-confirmed history of cancer or malignant tumor and completed at least three waves of the KLoSA survey. Participants were excluded if they had fewer than three cognitive function assessments following their cancer diagnosis or had a prior diagnosis of cerebrovascular disease or dementia.

At baseline (wave 1, 2006), 245 participants reported a history of cancer. An additional 509 participants subse-quently reported a new cancer diagnosis between wave 2 (2008) and wave 7 (2018), yielding a total of 754 identified cancer survivors. Among these, 324 individuals were excluded due to participation in fewer than three survey waves, resulting in 430 eligible participants. After further excluding 19 participants with fewer than three cognitive assessments after cancer diagnosis and 12 participants with pre-existing cerebrovascular disease or dementia, the final sample consisted of 399 participants (Figure 1).

Figure 1.

Flowchart for participant data selection.

RESULTS

1. Sample Characteristics

The final analytic sample consisted of 399 middle-aged and older cancer survivors (Table 1). The mean age of participants was 71.97 years (SD=9.36), and more than half of them were women (61.9%). Regarding educational attainment, 45.3% had completed elementary school or less, 16.7% had completed middle school, 26.8% had graduated from high school, and 11.1% had attained a college or higher. Overall, only 38.0% of the sample had completed high school education or higher. Hypertension (44.1%) and diabetes mellitus (22.1%) were the most commonly reported chronic conditions. BMI categories were distributed as follows: 7.8% were underweight, 47.6% were normal weight, 23.4% were overweight, and 21.2% were classified as class I or II obesity. In terms of health-related be-haviors, 5.8% of participants were current smokers, and 13.5% reported current alcohol consumption. Additional-ly, 36.6% of participants engaged in regular exercise at least once a week, while 92.7% maintained regular meal patterns. The mean of social support score was 7.19 (SD=2.74), and the mean depressive symptom was 1.91 (SD=2.34). The mean time since cancer diagnosis was 11.39 years (SD=6.67), ranging from 4 to 40 years. The most commonly reported primary cancer sites were stomach (22.8%), breast (17.5%), and colorectal (15.0%). At the time of the last survey, 25.3% of participants were receiving ongoing cancer-related treatment or symptom management. Regarding symptom progression, only 22.1% of them reported that their cancer had been cured, 32.2% reported improvement, 37.4% reported no change in symptoms, and 8.3% reported symptoms worsening since the last survey. About 16.3% of participants reported experiencing functional limitations in daily activities due to cancer.

Table 1.

Demographic Characteristics, Health Status, and Health-Related Habits of the Participants

2. Type of Trajectory in Cognitive Function

Latent class growth analysis identified two distinct cognitive trajectory groups among older cancer survivors (Table 2). Model fit indices, including AIC, BIC, and SABIC, decreased progressively with additional classes, suggesting improved statistical fit. However, when con-sidering classification quality and model parsimony, the two-class solution demonstrated the most favorable overall performance. The two-class model yielded a high en-tropy value (0.895), indicating strong classification precision, and showed significant improvements compared to the one-class model based on both LMRT (p<.001) and BLRT (p=.001). Although the three- and four-class models achieved further reductions in model fit indices, their LMRT and BLRT were either nonsignificant or less robust compared to the two-class model. Considering both statistical indicators and interpretability, the two-class sol-ution was selected as the optimal model. The identified latent classes exhibited distinct patterns of cognitive change over time (Figure 2). The cognitive maintenance trajectory group (85.2%) was characterized by a relatively high baseline cognitive score (mean=27.06) and only a modest decline of approximately 0.09 points per survey wave. In contrast, the cognitive decline trajectory group (14.8%) demonstrated a lower baseline score (mean=20.63) and a steeper decline of approximately 0.88 points per wave.

Table 2.

Model Fit Identification of Trajectory Groups by Cognitive Function

Figure 2.

Longitudinal trajectories of cognitive function scores among older cancer survivors. Two distinct trajectories were identified through latent class growth analysis: a cognitive maintenance trajectory characterized by stable cognitive function over time and a cognitive decline trajectory characterized by progressive decline across survey waves.

Following the classification of cognitive change trajectories, comparisons of sociodemographic, health-re-lated, psychosocial, and cancer-related characteristics between trajectory groups are presented in Table 1. Signifi-cant differences were found across multiple domains. Survivors in the decline group were older than the maintenance group (p<.001) and more likely to be female (p=.03). Educational attainment differed, with 91.30% of the decline group having completed middle school or less compared to 57.0% in the maintenance group (p<.001). In terms of health status, the decline group exhibited higher prevalence of physician-diagnosed hypertension (p=.023), diabetes mellitus (p=.002), and cerebrovascular accident (p=.018). BMI distributions differed, with the decline group more likely to be less obese (p<.001). In addition, the decline group demonstrated lower rates of regular exercise (p<.001) and alcohol consumption (p=.040), while no significant differences were observed in current smoking status and regular meal intake. Survivors in the decline group reported higher depressive symptom scores than those in the maintenance group (p=.001). Social support scores were comparable between groups. The decline group reported more difficulties with daily activities due to cancer than the maintenance group (p<.001).

3. Predictive Modeling of Cognitive Change Trajectories

Multiple machine learning algorithms were employed further to delineate predictors of cognitive change trajectories among cancer survivors. Model performance metrics revealed high overall accuracy across algorithms, with Random Forest demonstrating the highest accuracy (0.92) and AUC (0.93), followed by Logistic Regression (accuracy 0.86, AUC 0.86), Support Vector Machine (accuracy 0.86, AUC 0.86), and Neural Network (accuracy 0.82, AUC 0.78). Variable importance analyses identified both consistent and model-specific predictors of cognitive change trajectory membership (Table 3). Across all modeling approaches, older age consistently emerged as the most influential predictor, reflecting its robust association with vulnerability to cognitive deterioration. In Logistic Regression, lower educational attainment, lack of regular physical activity, functional limitations related to cancer, and longer time since cancer diagnosis were identified as key statistically significant factors. Random Forest priori-tized age, time since diagnosis, social engagement, depressive symptoms, and BMI as important variables contributing to prediction. In the Neural Network model, functional limitations related to cancer, diabetes mellitus, and symptom progression status were notable contributors, in addition to age and physical activity. The Support Vector Machine model emphasized age, BMI stage (class I obesity), symptom progression status, and educational attainment as variables of predictive value.

Table 3.

Comparison of Factor Ranking

DISCUSSION

This study investigated longitudinal trajectories of cognitive function among middle-aged and older cancer survivors and identified key predictors of decline using multiple supervised machine learning models. Cognitive trajectories were derived from individuals with K-MMSE da-ta collected at three or more time points. Based on these longitudinal data, individuals were grouped according to distinct patterns of cognitive changes: those who maintain cognitive function and those who exhibited continuous decline. Across all models, age consistently emerged as the most influential predictor, followed by educational attainment, physical activity, body mass index, time since cancer diagnosis, symptom progression, and cancer-re-lated limitations in daily functioning. Psychological factors, including depressive symptoms and social engagement, also contributed meaningfully to model performance. These findings highlight the multifactorial nature of cognitive vulnerability in long-term cancer survivorship and underscore the potential of machine learning approaches to enhance the early identification of at-risk indi-viduals.

The identification of two distinct cognitive trajectories aligns with prior research indicating that cognitive outcomes in cancer survivorship are heterogeneous and can persist well beyond the treatment phase [26]. In our cohort, approximately 15% of participants exhibited a sus-tained cognitive decline even more than a decade post-di-agnosis, highlighting a subset at heightened risk of ongoing neurocognitive vulnerability. This pattern supports the emerging view that cancer-related cognitive impairment may represent premature or accelerated aging rather than a transient, treatment-limited effect [10,12]. These findings underscore the importance of implementing targeted, long-term cognitive monitoring and individualized interventions within survivorship care to address the multifactorial and persistent nature of cognitive decline in middle-aged and older cancer survivors [10].

Survivorship cognitive deficits share biological vulnerabilities with age-related neurogenerative processes, raising the possibility that early cognitive changes during the post-treatment period could signal heightened long-term risk [10,27]. Recognizing and tracking these changes is critical for informing interventions aimed at maintaining cognitive health, delaying functional decline, and preventing dementia in an aging survivor population.

Collectively, these findings highlight the persistence of cognitive decline across diverse treatment modalities and underscore the need for longitudinal monitoring and supportive interventions for middle-aged and older survivors. The national survivorship programs in South Korea have played an important role in supporting patients’ recovery and reintegration, however, current frameworks have not yet fully evolved to reflect the distinct needs of a middle-aged and older survivor population [28]. As middle-aged and older adults now comprise an increasingly large proportion of cancer survivors, there is a pressing need to expand survivorship strategies that incorporate geriatric expertise and address the continuum of care across later life stages. Integrating early cognitive monitoring into survivorship care may offer a critical oppor-tunity to preserve cognitive resilience, sustain daily functioning, and improve long-term outcomes in cancer survivor populations.

The random forest algorithm appeared as the top-per-forming model among those evaluated in this study, pro-viding higher accuracy, specificity, and area under the curve compared to other methods. This finding highlights the potential of ensemble approaches to identify complex, nonlinear relationships across diverse health datasets [29]. The support vector machine model also showed higher sensitivity than logistic regression, indicating its utility in detecting individuals at increased risk for cognitive decline. Conversely, the neural network model exhibited lower predictive performance, likely due to limitations related to the modest sample size and class imbalance between cognitive trajectory groups [30]. Although logistic regression has been widely used as a common and effective method in cognitive aging research, its reliance on linear assumptions may have restricted its ability to capture the multidimensional and interactive nature of the variables examined in this study [14,15]. This finding reflects a broader limitation of traditional statistical models since, although they can identify isolated predictors, they are poorly suited to capturing nonlinear interactions, time de-pendent changes, and high-dimensional interdependence. In contrast, machine learning approaches, particularly ensemble methods like random forests, offer robust tools for modeling such complexity, enabling the detection of hid-den risk patterns and improving predictive accuracy across heterogeneous survivor populations. These findings underscore the benefits of employing adaptive, data-driven analytic approaches for risk stratification among cancer survivors when examining complex outcomes such as cognitive decline.

Cancer-related factors were salient predictors of cognitive decline in this study, indicating the persistent influence of disease burden and functional status on neurocognitive health in survivors. Longer time since diagnosis was linked to greater cognitive decline, reflecting that risk may accumulate with age rather than diminish over time. This challenges the assumption that prolonged survivorship leads to cognitive recovery and highlights the importance of continuous monitoring beyond the treatment phase [10,12]. Functional limitations related to cancer consistently predicted decline across multiple machine learning models. These findings support the "disuse hypothesis" in cognitive aging, which suggests that reduced physical and cognitive activity may accelerate neural deterioration through diminished synaptic activity, lower levels of brain-derived neurotrophic factor, and heightened physiological stress response [31]. Symptom progression was also identified as a prominent predictor, particularly within the support vector machine and neural network models. This reflects the cumulative impacts of unre-solved symptoms and potential physiological dysregulation, including chronic inflammation and hypothalamic-pituitary-adrenal (HPA) axis activation [32]. Persistent symptoms may also contribute to distress, fatigue, and reduced engagement in social or cognitive activities, further increasing vulnerability to cognitive decline among survivors [9].

Not surprisingly, sociodemographic, health-related, and psychosocial factors were also important predictors of cognitive decline, underscoring how personal vulnerabilities contribute to neurocognitive aging among cancer survivors. Age was consistently found as the strongest predictor across models, reflecting its established association with cognitive decline and reduced brain resilience in older adulthood [10-13]. Lower educational attainment was associated with a higher risk of decline, supporting the cognitive reserve hypothesis that individuals with fewer years of formal education may have less neural capacity to adapt to aging or treatment-related challenges [10]. Participants reported exercising an average of 4.87 days per week for approximately 60.8 minutes per session, meeting physical activity guidelines recommended by the World Health Organization [33]. Physical activity and body mass index were also significant, with regular exercise benefiting neuroplasticity and cerebral perfusion, and reduced neuroinflammation [31]. BMI showed a nonlinear relationship with cognitive risk, as both underweight and obesity have been associated with poorer cognitive outcomes. Among older adults, mild to moderate adiposity may con-fer metabolic reserve and enhance resilience during period of physiological stress, while underweight status may reflect frailty, malnutrition, or sarcopenic risk that ex-acerbates neural vulnerability [34]. Evidence regarding the association between BMI and mortality among middle-aged and older cancer survivors has been inconsistent, likely because BMI does not adequately reflect body com-position. A recent meta-analysis reported that cancer patients with overweight or obesity combined with low muscle mass exhibited higher cancer-specific mortality than those within the normal BMI range [35]. These findings suggest that BMI remains a valuable but incomplete in-dicator of health risk, and its sole use fails to capture the in-dependent roles of adiposity and muscle mass in shaping outcomes [35]. Furthermore, the reliance on self-reported BMI and lifestyle variables in this study introduces the possibility of recall and reporting biases, which should be considered when interpreting the findings. Future research should incorporate objective, multidimensional assessments to more accurately evaluate the influence of lifestyle and health factors on long-term cognitive outcomes among cancer survivors.

Psychosocial factors influenced cognitive trajectories in this study. Depressive symptoms were associated with greater cognitive decline, consistent with neurobiological evidence implicating HPA axis dysregulation, elevated inflammation, and hippocampal atrophy contribute to cognitive deterioration in late-life depression [32]. Social support was inversely associated with a lower risk of decline. This suggests that social connectedness may help maintain cognitive reserve, buffer stress-related neural dam-age, and serve as resources for maintaining cognitive vital-ity in later life [12].

Several limitations should be considered when interpreting these findings. First, this study applied latent class growth analysis (LCGA) to examine longitudinal changes in cognitive function. LCGA simultaneously estimates the intercept and the rate of change, thereby partially adjusting for aging effects [24]. However, long-term temporal influences unrelated to disease may still have accumulated. Such influences may be particularly relevant for participants with extended follow-up, such as those enrolled in 2006 and observed through 2018. Future research should consider anchoring trajectories to cancer diagnosis or disease duration rather than survey year, in order to dis-tinguish disease-related changes more clearly from general temporal influences. In addition, combining pop-ulation-based surveys with clinical cohorts that include detailed treatment information may allow a more precise characterization of cognitive trajectories in cancer survivors. Second, although the data were obtained from a nationally representative longitudinal cohort, the analytic sample of cancer survivors may not fully reflect the broader survivor population. Attrition was more likely among older or less healthy individuals, which raises the potential for selection bias. Moreover, the analytic sample was relatively limited because cancer survivors represented only a small subset of the KLoSA cohort. This constraint may reduce generalizability and may compromise the stability and interpretability of certain machine learning models, particularly the neural network [30]. Third, cognitive function was assessed using the K-MMSE, a widely used instrument that offers convenient global screening but lacks sensitivity for detecting subtle or domain-specif-ic impairments [19]. This limitation may have led to an underestimation of cognitive changes among survivors. Future research should consider incorporating objective neuropsychological assessments alongside the screening tool to improve the detection of cognitive decline in cancer survivors. Fourth, as this was a secondary data analysis, several important variables that strongly influence cognitive function were unavailable. Details on cancer-related characteristics such as treatment types, pathological features, and cancer stage could not be included in the analysis. This makes it impossible to determine whether cognitive decline persists independently of illness and treatment modalities. Similarly, the KLoSA data did not include information on electronic cigarette use. Smoking status was restricted to conventional cigarette use, which may have resulted in incomplete measurement of smoking-related behaviors. Finally, although KLoSA was designed with stratification, and sampling weights to ensure national representativeness, these features could not be applied because the machine learning frameworks used in this study do not accommodate complex survey weighting. While this omission is unlikely to have sub-stantially influenced model performance, it may affect the representativeness of descriptive estimates and should be considered when interpreting the findings.

Despite these limitations, the findings of this study offer meaningful implications for clinical practice and future research. Above all, by leveraging the unique advantages of the Korean Longitudinal Study of Aging (KLoSA), which provides nationally representative sampling, repeated longitudinal assessments, and comprehensive coverage of health and social indicators, we were able to conduct trajectory analyses of cognitive change that would not have been feasible with clinical or cross-sectional datasets. The identification of symptom progression status and cancer-related functional limitation as key predictors of cognitive decline emphasizes the need to integrate these factors into survivorship care planning. Moreover, the robust performance of the random forest model demonstrates the potential utility of machine learning for predicting long-term cognitive outcomes by enabling early identification of high-risk survivors and informing strategies for personalized care. Future research should extend this work by advancing predictive modeling and exploring potential interaction effects among sociodemographic, clinical, and health-related factors to provide deeper insights into the mechanisms of cognitive decline. Integrating machine learning-based insights with nursing expertise may in-form the development of targeted, nurse-led interventions that preserve cognitive resilience, sustain daily functioning, and ultimately enhance the quality of life for aging cancer survivors.

CONCLUSION

This study leveraged longitudinal data and machine learning methods to explore predictors of cognitive decline among middle-aged and older cancer survivors. Factors such as age, time since cancer diagnosis, symptom progression status, and functional limitation related to cancer were identified as key risk indicators. The random forest model yielded the strongest predictive performance, demonstrating the potential of ensemble methods in capturing complex patterns related to cognitive outcomes. Despite limitations due to the use of secondary data, these findings provide direction for enhancing survivorship care. Future research should aim to integrate detailed clinical and treatment information and test predictive models in diverse populations to advance personalized strategies for maintaining cognitive health in middle-aged and older cancer survivors.

Article information Continued