Risk Factors and Annual Rates of Cognitive Decline in Elderly Mild Cognitive Impairment: A Retrospective Cohort Study

Introduction

Mild cognitive impairment (MCI) is widely recognized as a transitional state between normal aging and dementia, with a high risk of progression to Alzheimer’s disease (AD) and other dementias.^1,2 It is characterized by a decline in cognitive function beyond that expected for an individual’s age and education level, but without significant impairment in activities of daily living.³ The prevalence of MCI increases with age, affecting approximately 10–20% of adults aged 65 and older.⁴ Given the growing aging population, understanding the trajectories of cognitive decline in elderly patients with MCI and identifying related risk factors has become a major public health priority.

Research concerning cognitive decline has long been a focal point in the field of neuroscience. MCI, positioned between normal aging and the more severe stages of dementia, has garnered significant attention. As indicated by prior investigations, MCI patients are at a heightened risk of progressing to dementia, with annual conversion rates approximated between 5% and 15%.⁵ Nevertheless, the progression of cognitive decline in MCI patients is not uniform. It exhibits heterogeneity, potentially shaped by an intricate interplay of factors such as age, sex, education level, genetic predisposition, vascular risk factors, and lifestyle choices.⁶

The transition from MCI to dementia is a critical issue that demands thorough exploration. A more rapid cognitive decline may serve as a harbinger of a higher likelihood of progressing to dementia, thereby underscoring the urgency for timely intervention. Existing evidence suggests that advanced age, male gender, lower educational attainment, vascular risk factors, and poorer baseline cognitive function are linked to an accelerated rate of cognitive decline in MCI patients.^7,8 However, the specific relationships between these factors and cognitive decline trajectories warrant more in-depth exploration to enhance our comprehension and guide the development of precision medicine approaches in dementia prevention.⁹ The complexity of these relationships and the potential for interaction among these factors make this an area ripe for further investigation. However, existing studies show inconsistent decline rates due to heterogeneous designs, while longitudinal trajectory analyses with ≥3 assessments in Chinese elderly populations remain scarce.¹⁰ Most prior research relied on simple pre-post comparisons rather than trajectory-based approaches that capture non-linear patterns.¹¹ These limitations hamper personalized intervention strategies. Chinese elderly cohorts are underrepresented in international literature,¹⁰ yet their application in elderly MCI cohorts with comprehensive neuropsychological batteries and extended follow-up remains insufficient. Moreover, few studies have simultaneously examined both dementia conversion risk and annual decline rates within the same analytical framework.¹²

Given the limitations of previous studies and the need for a more comprehensive understanding of cognitive decline trajectories and related risk factors in elderly patients with MCI, this retrospective cohort study aims to analyze the trajectories of cognitive decline in elderly patients with MCI and identify the risk factors associated with progression from MCI to dementia and the annual rate of cognitive decline. The findings of this study may provide valuable insights for early identification and intervention in high-risk individuals to delay or prevent progression to dementia.

Methods

Study Design and Data Source

This retrospective cohort study included elderly patients diagnosed with MCI at our hospital between January 2021 and December 2024. The inclusion criteria were as follows: 1) age ≥60 years old; 2) met the Petersen criteria for MCI,¹³ including subjective cognitive decline reported by the patient or informant, objective cognitive decline from a previously normal baseline, preserved activities of daily living, and absence of dementia; 3) available medical records with sufficient data for cognitive function assessment and risk factors analysis; 4) Eligibility required ≥3 serial cognitive evaluations at 12-month intervals. Exclusion criteria included: 1) presence of other neurological disorders that could affect cognition (eg, stroke, Parkinson’s disease, multiple sclerosis); 2) severe psychiatric disorders (eg, dementia, schizophrenia, bipolar disorder); 3) reversible cognitive impairment (eg, untreated hypothyroidism, vitamin B12 deficiency); 4) history of alcohol or substance abuse within the past year; 5) incomplete or unreliable medical records.

Based on pilot data from our institution (n=100) showing a 12% annual conversion rate with SD=2.1 for SLUMS decline, we projected that 400 participants would provide 80% power to detect a 0.5-point difference in annual SLUMS decline rate at α=0.05, allowing for 15% attrition.

This study was approved by the Institutional Review Board of Heilongjiang University of Chinese Medicine (Approval No.: 202411544) and conducted in accordance with the ethical standards of the Declaration of Helsinki. Individual informed consent was waived due to the retrospective and anonymized nature of the data analysis.

Cohort Follow-Up and Outcome Definitions

The follow-up comprised annual cognitive assessments at 12-month intervals. Actual completion rates: 12 months (n=487, 97.4%), 24 months (n=456, 91.2%), 36 months (n=398, 79.6%), 48 months (n=312, 62.4%), and 60 months (n=156, 31.2%). Median follow-up was 36 months (IQR 24–48). Participants contributed median 4 assessments (range 3–6). Those lost to follow-up (n=23, 4.6%) were censored at last assessment. Progression to dementia was established according to DSM-5 criteria by a panel of neurologists and neuropsychologists. Diagnosis required significant cognitive decline in one or more domains interfering with independence in everyday activities. The process involved: (1) structured clinical interview; (2) neuropsychological assessment; (3) brain MRI/CT to exclude other causes; (4) consensus conference review. Date of diagnosis was the first visit when DSM-5 criteria were met.

Data Collection

Predictors were selected based on established dementia risk factors (demographics, vascular, lifestyle) implicated in cognitive decline.^6,8 Data were collected from electronic medical records, which included demographic information (age, sex, education level, etc), medical history (hypertension, diabetes, hyperlipidemia, cardiovascular disease, etc), lifestyle factors (smoking status, alcohol consumption, physical activity level, etc), and cognitive function assessment results. Physical activity was categorized as low (<30 min/week), moderate (30–150 min/week), or high (>150 min/week) based on self-reported frequency and duration of moderate-intensity activities. Smoking status was classified as never smoked, former smoker (quit >1 year), or current smoker. Alcohol consumption was categorized as non-drinker, occasional drinker (<1 drink/week), or regular drinker (≥1 drink/week) based on self-reported intake over the past year. Cognitive function was assessed using a comprehensive neuropsychological test battery at baseline and during follow - up visits. The Saint Louis University Mental Status (SLUMS) Examination was used to evaluate general cognitive function, with a range of 0–30 points.¹⁴ The Montreal Cognitive Assessment (MoCA) was also administered to assess various cognitive domains, including memory, attention and concentration, executive function, language, visuospatial abilities, and abstraction, with a maximum score of 30. Additionally, specific cognitive domain tests were conducted, such as the Rey Auditory Verbal Learning Test (RAVLT) for memory function and the Clock Drawing Test (CDT) for visuospatial abilities.

The Montreal Cognitive Assessment (MoCA) was used to assess multiple domains (range 0–30). Because MoCA performance is strongly influenced by age and education, raw scores were interpreted with caution. Age and education were included as covariates in all multivariable models. We acknowledge that age- and education-adjusted normative approaches may improve clinical interpretability;^15,16 however, such norms specific to this Chinese elderly cohort were not available, and statistical adjustment via covariate inclusion provides appropriate control.

Outcome Measures

The primary outcome was the trajectory of cognitive decline. Cognitive decline was defined as a decrease in cognitive test scores from baseline to follow - up. We calculated the annual rate of change in SLUMS, MoCA, RAVLT, and CDT scores to characterize the cognitive decline trajectory. The secondary outcomes were incidence of progression from MCI to dementia and time to progression, with dementia diagnosis defined in Cohort Follow-up and Outcome Definitions.

Statistical Analysis

All statistical analyses were performed using SPSS 26.0. Continuous variables were presented as mean ± standard deviation (SD) or median (interquartile range, IQR), and categorical variables were presented as frequencies and percentages.

Descriptive statistics were used to summarize the characteristics of the study population. Differences in baseline variables between groups were compared using Student’s t - test or Mann–Whitney U-test for continuous variables and chi - squared test or Fisher’s exact test for categorical variables. For Cox regression, all 500 participants were included (23 lost to follow-up censored at last assessment). For linear regression, 477 participants (95.4%) with ≥2 assessments were included. Baseline characteristics were similar between groups with complete versus limited data (all p>0.05). Sensitivity analyses using multiple imputation (MICE) yielded consistent results. Multivariable Cox proportional hazards regression models included 8 pre-specified predictors (age, sex, education level, hypertension, diabetes mellitus, hyperlipidemia, cardiovascular disease, and baseline SLUMS score), yielding approximately 10 events per variable. For interpretability, baseline SLUMS was coded such that hazard ratios correspond to 1-point decreases in score (ie, HR>1 indicates higher risk with lower baseline cognitive function). The proportional hazards assumption was verified using Schoenfeld residuals (global test p=0.42; all individual p>0.05). Linearity of continuous variables was assessed using restricted cubic splines; no significant non-linearity was detected (all p>0.10). Multicollinearity was excluded using variance inflation factors (all VIF<2.0). To analyze the cognitive decline trajectories, we employed linear mixed-effects models with restricted maximum likelihood (REML) estimation, which accounted for the correlation between repeated measurements and varying assessment intervals inherent to retrospective data collection. These models allowed us to estimate the fixed effects of time and other covariates on cognitive test scores. Restricted cubic splines were used to assess non-linearity of continuous variables; no significant departure from linearity was detected (all p > 0.10). We also assessed the presence of nonlinear trajectories by including quadratic terms of time in the models. Multivariable Cox proportional hazards regression models were used to identify risk factors associated with the progression from MCI to dementia. Variables with a p - value < 0.20 in the univariate analysis were included in the multivariable model. Cox models included 8 predictors (age, sex, education, hypertension, diabetes, hyperlipidemia, cardiovascular disease, baseline SLUMS), with 10 events per variable. Proportional hazards assumption was verified using Schoenfeld residuals (p>0.05 for all variables). Hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated to quantify the strength of associations. For the analysis of cognitive decline trajectories, multivariable linear regression models were used to identify risk factors associated with the annual rate of cognitive decline. Variables were selected using a similar approach as in the Cox regression analysis.

Sensitivity analyses were conducted to assess the robustness of our findings. We performed subgroup analyses based on age, sex, and education level to examine whether the associations between risk factors and cognitive decline trajectories varied across different subgroups.

Results

Baseline Characteristics of Study Participants

A total of 500 elderly patients with MCI were included in this retrospective cohort study. The baseline characteristics of the study participants are presented in Table 1. The mean age of the participants was 75.2 ± 6.8 years, and 265 (53%) were male. The mean education level was 12.3 ± 3.4 years. The majority of participants were non - smokers (68%, 340/500) and had a history of hypertension (52%, 260/500). The mean SLUMS score at baseline was 25.6 ± 2.1, and the mean MoCA score was 22.4 ± 1.9.

Table 1 Baseline Characteristics of Study Participants

Cognitive Decline Trajectories

During median follow-up of 36 months (IQR 24–48), 78 participants (15.6%) progressed to dementia, with an annual conversion rate of 5.2%. The cognitive decline trajectories of the study participants are shown in Table 2. The mean annual rate of decline in SLUMS score was 0.8 points per year (95% CI: 0.6, 1.0, p < 0.001), indicating a significant decrease in general cognitive function over time. Subgroup analyses revealed steeper decline in participants with hypertension compared to those without (mean difference −0.6 points/year, 95% CI: −0.9 to −0.3, p = 0.001). Similarly, participants aged ≥75 years showed accelerated RAVLT decline compared to those <75 years (−1.5 vs −0.9 points/year, p < 0.001). Similarly, the mean annual rate of decline in MoCA score was 0.7 points per year (95% CI: 0.5, 0.9, p < 0.001). For specific cognitive domains, the mean annual rate of decline in RAVLT score was 1.2 (95% CI: 0.9, 1.5, p < 0.001), and the mean annual rate of decline in CDT score was 0.3 (95% CI: 0.2, 0.4, p = 0.002).

Table 2 Cognitive Decline Trajectories of Study Participants

Risk Factors for Cognitive Decline

The results of the multivariable Cox proportional hazards regression analysis for risk factors associated with progression from MCI to dementia are presented in Table 3. Older age (HR = 1.05, 95% CI: 1.02, 1.08, p = 0.001), male sex (HR = 1.42, 95% CI: 1.13, 1.78, p = 0.023), lower education level (HR = 0.94, 95% CI: 0.91, 0.97, p < 0.001), hypertension (HR = 1.35, 95% CI: 1.12, 1.62, p = 0.012), and lower baseline SLUMS score (HR = 1.07 per 1-point decrease, 95% CI: 1.03, 1.11, p < 0.001) were significantly associated with an increased risk of progression from MCI to dementia.

Table 3 Risk Factors for Progression from MCI to Dementia

The results of the multivariable linear regression analysis for risk factors associated with the annual rate of cognitive decline are shown in Table 4. Older age (β = −0.06, 95% CI: −0.08, −0.04, p < 0.001), male sex (β = −0.52, 95% CI: −0.83, −0.21, p = 0.003), lower education level (β = −0.08, 95% CI: −0.11, −0.05, p < 0.001), hypertension (β = −0.45, 95% CI: −0.71, −0.19, p = 0.001), diabetes mellitus (β = −0.38, 95% CI: −0.64, −0.12, p = 0.015), and lower baseline SLUMS score (β = −0.12, 95% CI: −0.15, −0.09, p < 0.001) were significantly associated with a faster annual rate of cognitive decline.

Table 4 Risk Factors for Annual Rate of Cognitive Decline

Sensitivity Analysis

Subgroup analyses stratified by age (<70 vs. ≥70 years), sex, and education level (<12 vs. ≥12 years) showed generally consistent associations (Table 5). Notably, although diabetes mellitus showed marginal non-significance for dementia conversion in the overall cohort (HR = 1.23, p = 0.087), significant associations were observed in both male (HR = 1.24, 95% CI: 1.01, 1.52, p = 0.011) and female (HR = 1.22, 95% CI: 1.01, 1.47, p = 0.037) subgroups. Education level showed universally protective effects across all subgroups (all p < 0.001). These findings support the generalizability of our main results across demographic strata.

Table 5 Sensitivity Analysis in Subgroups

Discussion

The study provides a comprehensive analysis of cognitive decline trajectories and related risk factors in elderly patients with MCI. The findings indicate that cognitive decline in MCI patients is progressive and heterogeneous, with significant declines observed in general cognitive function (SLUMS and MoCA scores) and specific cognitive domains (RAVLT and CDT scores). These results are consistent with previous studies that have documented the gradual progression of cognitive impairment in MCI patients over time.¹⁷ The mean annual rate of decline in SLUMS score (0.8 points per year) and MoCA score (0.7 points per year) aligns with prior research findings, which have reported similar rates of cognitive decline in MCI populations.^18,19

The identification of risk factors for progression from MCI to dementia and for faster cognitive decline is a crucial aspect of our study. Our results highlight several demographic and clinical factors that are significantly associated with adverse cognitive outcomes. Older age emerged as a strong predictor of both progression to dementia and a faster rate of cognitive decline. This finding is well - supported by existing literature, as advanced age is a well - established risk factor for dementia.^20,21 The association between male sex and increased risk of cognitive decline and dementia progression is also in line with previous studies.²² Sex-related disparities in cognitive vulnerability likely stem from a dual pathway: biologically, intrinsic differences in brain structure and function, including hormonal and vascular profiles, may contribute; while socially and behaviourally, distinct life-course exposures amplify or mitigate risk.²³ The observed male predominance in cognitive decline may reflect these multifactorial influences, though the specific mechanisms require further investigation with dedicated hormonal and biomarker assessments.

Lower educational attainment was associated with both higher dementia conversion risk and faster cognitive decline, consistent with the cognitive reserve hypothesis.^24,25 This framework posits that higher education provides greater neural connectivity and synaptic plasticity, enabling individuals to better tolerate accumulating neuropathology before clinical symptoms become apparent.^26,27 In our cohort, the association between lower education and poorer outcomes likely reflects reduced resilience to pathological changes, whereby individuals with fewer reserve-related resources manifest cognitive impairment earlier and progress more rapidly.²⁸ Recent evidence suggests that reserve-related factors are associated with delayed progression to objective cognitive impairment,²⁷ which may explain why education remained independently linked to both outcomes despite adjustment for vascular risk factors. This mechanistic interpretation provides a theoretical basis for the observed educational gradient and informs potential intervention strategies.

Hypertension was another significant risk factor identified in our study. This finding underscores the importance of vascular health in maintaining cognitive function and preventing dementia.²⁹ Effective management of hypertension through lifestyle modifications and pharmacological interventions may have a protective effect on cognitive function.^30,31 The relationship between vascular health and cognitive decline is complex and multifactorial. Hypertension can lead to small vessel disease, white matter hyperintensities, and reduced cerebral blood flow, all of which can contribute to cognitive impairment.³² Therefore, early detection and treatment of hypertension should be a priority in the management of patients with MCI.

Interestingly, diabetes mellitus was associated with a faster annual rate of cognitive decline (β = −0.38, p = 0.015). Although the association with dementia conversion was marginally non-significant in the overall cohort (HR = 1.23, p = 0.087), subgroup analyses revealed significant effects in both male (HR = 1.24, p = 0.011) and female (HR = 1.22, p = 0.037) subgroups (Table 5). This pattern suggests that diabetes effects on dementia risk may be obscured in aggregate analyses due to sex-specific baseline risk differences, but are consistently present across sexes. These findings align with prior studies reporting significant associations between diabetes and dementia risk,^33,34 and highlight the importance of considering sex-stratified analyses in MCI cohorts. The similar effect sizes in both subgroups indicate sex-independent vascular/metabolic mechanisms. The consistency of risk factor associations across age, sex, and education subgroups strengthens the generalizability of our findings. The observation that diabetes effects were significant in both sexes, while education effects were universally protective, suggests these factors operate through stable biological and cognitive reserve mechanisms, respectively. Further research with larger sample sizes and longer follow - up periods is needed to clarify the relationship between diabetes and dementia risk in MCI patients. The potential mechanisms underlying the association between diabetes and cognitive decline may include hyperglycemia-induced oxidative stress, insulin resistance, and the development of microvascular complications.^35,36 Therefore, the role of diabetes in cognitive decline warrants further investigation.

The significant association between lower baseline SLUMS score and worse cognitive outcomes suggests the importance of early cognitive assessment in identifying individuals at high risk of progression to dementia. Regular monitoring of cognitive function in MCI patients can help detect early signs of decline and facilitate timely intervention. Early detection and intervention are crucial in managing MCI, as they can potentially delay the progression to dementia and improve quality of life for patients and their caregivers.^37,38

Our study has several strengths. First, it is a large - scale retrospective cohort study with a relatively long follow - up period, providing robust data on cognitive decline trajectories and risk factors. Second, we used a comprehensive neuropsychological test battery to assess cognitive function, allowing for a detailed analysis of both general and specific cognitive domains. Third, we employed rigorous statistical methods to account for the correlation between repeated measurements and to identify independent risk factors.

However, there are also some limitations to consider. First, the retrospective design may introduce selection bias, as participants with regular healthcare access and complete longitudinal assessments may differ systematically from those lost to follow-up or with incomplete records. Second, reliance on existing medical records may result in information bias, particularly for lifestyle factors and medical history documentation. Third, despite multivariable adjustment, residual confounding from unmeasured factors (eg, APOE ε4 genotype, social engagement, inflammatory biomarkers) cannot be excluded, and observed associations may reflect reverse causation rather than true causal effects. Fourth, our single-center setting may limit generalizability to other populations. Future research should employ prospective designs, multicenter collaboration, comprehensive biomarker assessment, and standardized protocols to address these limitations.

Conclusion

In conclusion, this retrospective study suggests that demographic and clinical factors—including older age, male sex, lower education level, hypertension, and reduced baseline cognitive function—are associated with cognitive decline in elderly MCI patients. These observational findings may inform early identification of high-risk individuals, though causal inferences are limited by the retrospective design. Future prospective studies are warranted to validate these associations and guide intervention strategies.