Objective and Subjective Measures of Cognitive Decline in Highly Educated Older Adults: A 10-Year Longitudinal Study

Abstract

Background: The timely detection of cognitive decline in highly educated adults is challenging due to their resilient cognitive abilities and the limited sensitivity of neuropsychological tests for this group. Therefore, evaluating subjective facets such as subjective cognitive decline (SCD) becomes imperative, potentially enabling the early identification of cognitive decline. Objective: Our primary objective was to identify effective methods, both objective and subjective, for the early detection of cognitive decline in highly educated older adults. A secondary objective was to translate and validate a Hebrew adaptation of the SCD questionnaire. Methods: Initially (T0), the study included 28 highly educated participants (mean age = 72.6, SD = 4.54; mean education 17.6, SD = 3.41). By the final evaluation (T7), 20 participants remained. Annual assessments involved objective neuropsychological tests and self-report questionnaires evaluating depression, anxiety, and SCD with changes analyzed over time using repeated measures ANOVA. Results: Significant declines were observed in the following objective neuropsychological tests: Rey–Osterrieth Complex Figure Test (ROCFT) copy, F(3,57) = 9.05, p < 0.001, ηp2 = 0.32, and Rey Auditory Verbal Learning Test (RAVLT) trial six, F(1,19) = 7.32, p < 0.05, ηp2 = 0.28, which is consistent with previous findings. The Hebrew SCD questionnaire demonstrated high reliability and validity and was highly correlated with cognitive decline. Conclusions: The ROCFT copy and the Hebrew SCD questionnaire can serve as valuable indicators for the early detection of cognitive decline in highly educated older adults.

1. Introduction

As life expectancy rises, the incidence of cognitive decline among older individuals is on the rise [1]. While cognitive decline is a natural aspect of aging, it may, in some cases, indicate an underlying neuropathological condition such as dementia or Alzheimer’s disease (AD) [2,3]. Extensive research has been conducted to develop effective assessments for detecting the earliest clinical signs of dementia, particularly AD, including neuropsychological tests, biomarkers, and neuroimaging techniques [4]. However, there is a lack of systematic investigation focusing specifically on highly educated older adults.

Highly educated older adults often experience a delay in detecting initial signs of neuropathological conditions using standardized tools, which is likely due to their well-developed cognitive reserve (see ‘Highly Educated Older Adults’ section for further discussion) [5,6]. Nevertheless, many of them report subjective cognitive decline (SCD) and psychological distress (e.g., depression and anxiety) [7], both of which are recognized as risk factors for dementia and specifically AD [8]. Given this recognition—that SCD and psychological distress are established risk factors for dementia and may reflect underlying neuropathological changes even when objective impairments are not yet detected—it is imperative to consider the subjective experiences of highly educated individuals in the diagnostic process. Therefore, the implementation of early and accurate diagnostic methods that incorporate subjective aspects could prove especially advantageous for this demographic, facilitating timely intervention and support. This study aims to bridge this research gap through a comprehensive longitudinal investigation spanning over a decade. Its primary objective is to identify effective methods, encompassing both objective and subjective measures, for the early detection of cognitive decline in highly educated older adults.

1.1. The Subjective Aspects of Cognitive Decline: SCD and Psychological Distress

Subjective cognitive decline (SCD) refers to an individual’s self-perception of experiencing ongoing cognitive decline, often linked to memory issues, even though objective neuropsychological assessments do not show noticeable decline [8].

SCD is often considered the earliest symptomatic manifestation of MCI and dementia due to AD [9], as supported by imaging studies [10]. Furthermore, SCD is recognized as a risk factor for both future cognitive decline and the progression to dementia, particularly AD [8,11]. Therefore, it represents an opportune stage for early intervention [12]. Similar research using a validated SCD questionnaire found a sensitivity of 85% and a specificity of 80% in distinguishing individuals with cognitive impairment from those without, further underscoring the importance of SCD as a predictive tool [13].

Anxiety and depression can also lead to SCD that is transient and not necessarily indicative of neurodegenerative pathology. However, the presence of anxiety, depression, and concerns among older adults with SCD heightens the probability of transitioning to AD [14].

Thus, further research is needed to examine the association between subjective aspects, such as psychological distress, and objective cognitive decline over time, as this aspect remains inadequately addressed in the current literature.

To accurately assess subjective aspects of cognitive decline, valid and reliable diagnostic tools are essential [8,12]. While established questionnaires like the Beck Depression Inventory (BDI) [15] and the State–Trait Anxiety Inventory (STAI) questionnaire [16] are widely used to measure depression and anxiety, no Hebrew questionnaire exists specifically for SCD. Among the validated SCD tools, Gifford’s 50-item questionnaire [17] is notable for its reliability across multiple languages [13,18]; however, it has yet to be translated into Hebrew. Therefore, our secondary objective is to initiate the process of translating and validating the questionnaire in Hebrew as part of this research effort.

1.2. Highly Educated Older Adults

Highly educated older adults are considered to benefit from cognitive reserve (CR), which may enhance their ability to sustain cognitive function and delays age-related cognitive decline [19,20]. However, once diagnosed with MCI or AD, they may experience faster cognitive decline, which is potentially due to delayed diagnosis [17,19,21]. Cognitive reserve may enable individuals to compensate for underlying neuropathological changes, allowing them to maintain cognitive function despite accumulating pathology. This compensatory ability may delay the manifestation of clinical symptoms, extending the preclinical phase and complicating early detection, particularly in highly educated individuals [5,6,19,21]. Importantly, alternative interpretations of cognitive reserve offer different perspectives on its neural basis. The neural efficiency model suggests that individuals with higher reserve require less neural activation to perform cognitive tasks [22], whereas the com-pensatory mechanisms model posits that reserve allows for the recruitment of alterna-tive neural networks to sustain function despite damage [19]. These models are not mutually exclusive and may operate in parallel depending on the context.

While the concept of CR has gained widespread acceptance in explaining individ-ual differences in cognitive aging [6], it is important to acknowledge that it is not without controversy. Critics have noted that CR is often inferred indirectly, lacks standardized measurement, and may conflate with other constructs such as brain re-serve or brain maintenance. Some studies highlighted ongoing challenges in defining and operationalizing CR across studies and emphasized the need for clearer differenti-ation between related constructs [21]. Despite these limitations, the current study adopts the CR framework as it remains a valuable lens through which to understand differential trajectories of cognitive decline, particularly in highly educated individuals who may compensate more effectively for underlying neuropathology.

Unlike large-scale studies that encompass participants with diverse educational backgrounds [22,23], our study specifically examines highly educated individuals, who may experience a delayed onset of decline followed by a steeper trajectory compared to those with lower education levels [24]. In contrast to ‘Superagers’, known for main-taining exceptional cognitive performance [24], our cohort represents the broader, al-beit delayed, pattern of typical cognitive aging. By integrating objective assessments with subjective SCD measures, our study offers valuable insights into early markers of decline within this unique population.

1.3. Neuropsychological Assessment: Longitudinal Study

Standard cognitive screening tools like the MMSE and MoCA may miss subtle cognitive changes, whereas comprehensive neuropsychological testing provides a more accurate evaluation across cognitive domains, especially in longitudinal studies that compare results to an individual’s baseline [2,8,19,20,21,24]. However, practice effects, where repeated testing improves performance, pose challenges in longitudinal studies, particularly in highly educated individuals [25,26]. These effects may diminish with adequate time gaps between tests, and their absence could indicate preclinical dementia [25,26].

Effective tools with minimal practice effects, such as the Rey Auditory Verbal Learning Test (RAVLT), verbal fluency (VF), and the Rey–Osterrieth Complex Figure Test (ROCFT), show promise in detecting early cognitive decline [27,28,29,30,31,32,33,34,35,36,37,38]. Specific RAVLT measures, like immediate memory and proactive interference, and the ROCFT’s copy phase, are particularly sensitive to subtle changes with minimal practice effects [25,29,37,38,39,40]. While these tests are promising, more research is needed to optimize early detection in highly educated older adults through longitudinal studies [2,3,8,9,10,40].

1.4. Current Research

To address existing research gaps, we conducted a decade-long longitudinal study utilizing a comprehensive neuropsychological battery. Our primary goal was to identify effective methods, both objective and subjective, for the early detection of cognitive decline in highly educated older adults. In terms of objective assessments, we aimed to build on our previous findings [40], focusing on declines in RAVLT trial six, Semantic Fluency (SF), and ROCFT copy scores within a 12-month period. Anticipating score reductions in these tests from the initial evaluation (T0) to the final assessment (T7), we recognize that these changes might take time to manifest, as the study population was initially healthy. To gain a deeper understanding of these trends, we also examined changes in the intermediate assessments (T5 and T6).

For subjective assessments, our goal was to examine whether subjective measures (depression, anxiety and SCD) are associated with objective cognitive performance and decline. Our investigation was systematic, considering relationships at specific time points (cognitive functions at T5, T6, and T7), and over time (cognitive decline from the study’s beginning at T0 to subsequent assessments at T5–T7). Given the varied research findings on subjective aspects, we refrained from establishing a unidirectional hypothesis regarding their relationship with objective cognitive decline.

The characterization of SCD as an early marker of cognitive decline is well established in the literature [12,13,14]. Our study does not aim to propose SCD as a novel construct per se but rather seeks to extend existing knowledge by focusing on its predictive value in a specific and often underexamined subpopulation: highly educated older adults. Due to the absence of a validated Hebrew-language SCD questionnaire, our secondary goal was to translate and validate a Hebrew adaptation of Gifford’s SCD questionnaire [17], which was used in our study from T5 to T7. The study was conducted with Hebrew-speaking older adults, which necessitated the use of a culturally and linguistically appropriate Hebrew version. To ensure linguistic and cultural validity, we followed a structured translation and validation process (see Section 2 for details).

2. Methods

2.1. Participants

A sample of highly educated older adults completed eight comprehensive cognitive assessments for over a decade (T0 to T7). The intervals between these assessments ranged from one year to a year and a half, which was subject to participant availability and external factors like COVID-19 lockdowns. The initial sample (T0) consisted of 28 healthy older adults with ages spanning from 66 to 80 years (M = 72.6, SD = 4.54). However, attrition and participant withdrawals occurred during the study. At T1, one participant chose not to continue their participation for personal reasons. Then, in the sixth year of follow-up (T6), three participants withdrew from the study. Two of them did so due to deteriorating general medical conditions, while one participant decided to discontinue participation for personal reasons. Finally, in the seventh year of follow-up (T7), an additional four participants requested to discontinue their involvement. Among them, two were diagnosed with dementia, one was diagnosed with Parkinson’s disease, and one participant chose not to continue the study for personal reasons. As a result, the final sample at T7 consisted of 20 participants, evenly split between men and women, with ages ranging from 74 to 89 years (M = 80.45, SD = 4.78). Their educational backgrounds varied from 12 to 21 years with an average of 17.6 years of education (SD = 3.41). Additional demographic information for all participants is presented in

Table 1. Sociodemographic characteristics of participants at T7.

Characteristic	Values
Age (years) Gender	M = 80.45 (SD = 4.78) 10 men/10 female
Education (years)	M = 17.6 (SD = 3.41)
BMI a	M = 26 (SD = 2.82)
Engagement in physical exercise b (number of times per week)	M = 2.8 (SD = 1.39)
Leisure activity c (times per week) Social gathering d (times per week) Living with a partner or alone?	M = 2.2 (SD = 1.57) M = 1.9 (SD = 1.14) 5 alone/15 with a partner
Country of birth, N	Israel-12 (60%) Poland-2 (10%) Rumania-2 (10%) England; Germany; Russia; Argentina-4 (20%)
Age of immigration, years	M = 10.43 (SD = 10.61)
Native language, n	Hebrew-12 (60%) Yiddish-4 (20%) English; Hungarian; Russian; Spanish; Romanian-5 (20%)

Note. N = 20. M = mean. SD = standard deviation. N = number of participants. ^a BMI = body mass index [weight (kg)/height² (m²)]. ^b Engagement in physical exercise = aerobic activity (e.g., tennis, swimming), or other activities (e.g., yoga, Feldenkrais). ^c Leisure activity = concerts, plays, movies, hobbies (e.g., playing an instrument, artistic activity). ^d Social gathering = meetings with friends or relatives.

.

Participants were initially recruited through advertisements placed in recreational centers frequented by independent older adults engaged in a variety of activities. The participants received modest financial compensation for participation in each session (50 shekels ~about USD 15). We used the following inclusion criteria to enter the study: (1) men and women over the age of 65 and (2) fluent Hebrew. We used the following exclusion criteria: (1) existence of a neurological or psychiatric disorder via a comprehensive neurological interview, evaluation, and review of medical information by the study neurologist, (2) history of head injuries and/or head surgeries, (3) diagnosis of cancer in the last five years, and (4) subjective reports of cognitive decline.

The initial survey, which involved conducting detailed telephone interviews and gathering relevant medical and surgical records, included 83 older adults. After analyzing the survey data, 49 potential participants were excluded due to MRI incompatibility, which was primarily caused by metallic implants. The remaining 34 participants underwent a comprehensive neurological assessment at the Tel Aviv Sourasky Medical Center. Out of these, six participants were further excluded due to a history of neurological disorders (such as attention-deficit hyperactive disorder, stroke, etc.) or psychiatric disorders (including documented depression). Finally, 28 individuals completed the MRI procedure and the neuropsychological assessment, and they were included in the study. None of the participants reported a significant objective cognitive decline, or SCD, at T0.

2.2. Procedure

Approval for the research was originally obtained by the Helsinki Committee of Tel Aviv Sourasky Medical Center in 2012, and an official extension was granted in 2022 (Approval number 0623-11-TLV, approval date is 12 December 2022) and the Ethics Committee of the Tel Aviv-Jaffa Academic College.

The research team maintained regular contact with participants on an annual basis via telephone calls and emails, ensuring ongoing engagement and scheduling follow-up assessments. Each contact session typically lasted 10 to 15 min and focused on (1) confirming participants’ willingness to continue in the study and ensuring their availability for upcoming assessments; (2) providing logistical details regarding the next evaluation session; and (3) addressing any brief questions participants had about the study process. All communications were conducted in Hebrew, ensuring that participants fully understood the purpose of the study and their continued involvement. Neuropsychological assessments were conducted at the Center for Memory and Attention Disorders, the Department of Neurology, and the Functional Brain Center, Wohl Institute for Advanced Imaging, which are all located at Tel Aviv Sourasky Medical Center in Israel. To accommodate participants facing travel difficulties, starting at time point T5 and beyond, data collection was extended to include home visits as needed.

Prior to participant involvement, researchers elucidated the study’s purpose and obtained written informed consent. Participants who agreed to participate were invited annually for evaluation, during which their consent was reaffirmed. Additionally, verbal consent was obtained for recording parts of the assessment to ensure the accurate coding of responses. Participants were provided with details about the session’s procedure, duration, and the availability of breaks.

Subsequently, participants underwent a series of assessments administered in Hebrew, following a predetermined test sequence designed to minimize interference among tests that rely on similar cognitive domains while accommodating time constraints. The sequence included the MoCA [41], RAVLT [27], ROCFT [37,38], Digit Span and General Knowledge Subtests—both from the Wechsler Adult Intelligence Scale—Third Edition (WAIS-III) [42], Logical Memory from Wechsler Memory Scale—Third Edition (WMS-III) [43], Trail-Making Test (TMT) Parts A and B [44], and a verbal fluency test (VF) test encompassing both phonemic fluency (PF) and semantic fluency (SF) [45].

After completing the cognitive tests, participants were given questionnaires in Hebrew, which were presented in a randomized order. These questionnaires included the BDI [15] and the STAI [16], which were aimed at evaluating participants’ psychological distress. Additionally, an annual questionnaire gathering demographic information was administered, and from the fourth year of the follow-up onward (T4–T7), a lifestyle assessment was introduced. From the fifth year of follow-up onward (T5–T7), participants were also given Gifford’s SCD questionnaire [17] in its translated Hebrew version. At T7, participants also filled out the Big Five Inventory (BFI) questionnaire [46]. Following the conclusion of the assessment session, participants received financial compensation.

2.3. Assessments

2.3.1. Neuropsychological Tests

RAVLT [27]: Evaluation of verbal episodic memory. We utilized the Hebrew version of the RAVLT, which was established by Vakil and Blachstein [29]. The scores for trials one to nine were determined based on norms provided by Vakil and Blachstein [27]. For interference measurements, we used Vakil’s updated norms [47]. The RAVLT exhibited satisfactory test–retest reliability (ranging between 0.36 to 0.68), good internal consistency (α = 0.80), and high construct validity [48].

ROCFT [37,38]: The assessment comprises two integral components—copying and delayed recall. Normalized scores were established utilizing well-established norms widely adopted in Israel [49]. The test exhibited a good level of reliability (α = 0.80), internal consistency (α = 0.82) [50], and high construct validity [51].

WAIS-III Knowledge Subtest [42]: This subtest evaluates an individual’s crystallized intelligence, which is a cognitive capacity facet that embodies pre-existing knowledge and skills. The evaluation was carried out in Hebrew with adherence to the well-established standard norms of the WAIS-III translated by Goodman [52]. This subtest demonstrated high internal consistency of 0.85 to 0.99 and good test–retest reliability of 0.75 to 0.99 [53].

TMT, Trails A and B [44]: The assessment evaluates visual working memory, attention, and processing speed. The calculation of normalized scores was based on widely recognized norms that are prevalent in Israel [54]. The test–retest consistency for Part A ranged from 0.76 to 0.89, and for Part B, it ranged from 0.86 to 0.94, affirming commendable stability over time [55]. Both TMT trials exhibit strong construct validity [56].

VF Test [45]: The assessment examines retrieval of words, initiation, and mental flexibility. Normalized scores for the VF test were calculated based on published norms in Hebrew [31]. The internal reliability was notably high with a coefficient of 0.91 for PF and 0.89 for SF [31].

2.3.2. Questionnaires

BDI [15]: The BDI is a well-known and widely validated depression assessment questionnaire that is used in both clinical and research settings. The overall score ranges from 0 to 63 and is calculated by the sum of the items [57]. A higher total score indicates a higher level of depression. Scores between 0 and 10 typically fall within the normal range [15]. The BDI exhibited strong construct validity [58], and its internal reliability ranged from 0.73 to 0.92 [59]. The validated Hebrew version of the BDI [60] was used in this research. The internal consistency of the BDI in our research was 0.98 at T0, 0.68 at T5, 0.70 at T6, and 0.78 at T7.

STAI [16]: A well-known and widely validated anxiety assessment questionnaire comprises 20 items. A threshold score of 40 is often employed to designate potential clinical levels of anxiety [61]. The internal consistency of the STAI spanned from 0.86 to 0.95 [16], and the construct validity was high [62]. The validated Hebrew version of the STAI [63] was used in this research. The internal consistency of the STAI in our research was excellent with values of 0.92 at T0, 0.95 at T5, 0.91 at T6, and 0.96 at T7.

BFI [46]: A brief self-report questionnaire assesses the five major personality dimensions. The questionnaire comprises 44 items, which are each rated on a 5-point Likert scale. Higher scores on a specific dimension reflect a stronger expression of that trait. The Hebrew version of the BFI [64] was used in the research. For this study, only the score on the Agreeableness dimension from the questionnaire was calculated in order to establish discriminant validity for the SCD questionnaire, as prior research had not previously linked this personality trait to SCD or to processes associated with dementia [65,66]. The internal consistency of the Agreeableness dimension in this research was satisfactory (α = 0.79).

Lifestyle Assessment: A comprehensive in-house questionnaire, modeled after a well-established survey structure [67], delved into leisure, social, and engagement in physical exercise.

SCD Questionnaire [17]: An established self-report tool designed to evaluate SCD was developed following the criteria proposed by the SCD Initiative Working Group [8]. The SCD50 questionnaire comprises 50 items that evaluate shifts in memory patterns, encompassing alterations in overall cognition, including attention and executive functions. Scoring involves 37 dichotomous questions (scored as either 0 or 1 on a Likert scale), such as “Do you have difficulty with your memory?” (35 “yes”/“no” responses, one “good”/“bad” response, and one “agree”/”disagree” response). Additionally, 13 questions are rated on a 3-point Likert scale (0–2) to indicate frequency or severity (“always”/“sometimes”/“never”; “significant”/“minor”/“no problems”), e.g., “How often is the following a problem for you: Personal dates (e.g., birthdays)?” Three items (questions 11, 19, and 30) require reversed scoring to maintain consistency. The overall score is obtained by summing the scores for all items, ranging from 0 (indicating no SCD) to 61 (indicating substantial SCD).

The questionnaire demonstrated high internal consistency (α = 0.94), and its construct validity has been well established [17]. Additionally, a condensed version of the questionnaire, comprising the first 21 items and scored from 0 to 27, has been explored (SCD21), showing high construct validity and internal reliability of 0.95 [17]. No prior cross-cultural validations of the SCD questionnaire in Hebrew have been published. For the Hebrew version of the questionnaire used in our study, the translation process involved the principal investigators (Odelia Elkana and Elissa Ash), along with a master’s student (Noy Tal), who were all proficient in both Hebrew and English. They individually translated the questionnaire and collaborated to synthesize a unified Hebrew version, addressing any disparities and linguistic challenges. Subsequently, a professional translator, unacquainted with the questionnaire’s concepts and original wording, executed a back-translation into English. Adjustments were made to the Hebrew version based on the translator’s feedback, ensuring consistent cross-cultural adaptations while preserving the questionnaire’s intended meanings. The final Hebrew version of the SCD questionnaire was confirmed by the authors.

At T0, participants had no complaints of cognitive decline and performed within the normal range on neuropsychological testing, thus meeting the inclusion criteria. By T5, many participants reported cognitive decline, prompting the introduction of a relevant SCD questionnaire.

We assessed the internal consistency of the Hebrew versions of the SCD50 and the SCD21 in this study at three distinct time points: T5, T6, and T7. Notably, the SCD50 demonstrated excellent internal consistency, as evidenced by Cronbach’s alpha values of 0.93 to 0.96. The SCD21 exhibited good internal consistency with Cronbach’s alpha values of 0.81 to 0.90. Remarkably high to excellent test–retest reliabilities were observed for both versions, ranging from 0.78 to 0.97.

2.4. Data Analyses

We utilized the 23rd version of the SPSS software. All neuropsychological tests were scored using normative data and standard scores. To address the seven percent missing data at T0 (two participants were unable to complete two questionnaires), we used regression imputation. We derived neuropsychological scores based on age-matched norms and examined gender differences using t-tests. We used Pearson correlations for normally distributed variables and Spearman correlations for non-normally distributed ones. We compared cognitive function and subjective measures over time using one-way analyses of variance (ANOVA) for repeated measures, which were followed by post hoc pairwise comparisons with Sidak adjustment. Our study employs a longitudinal within-subject design, leveraging multiple assessment points to enhance statistical power despite sample attrition over time. The repeated-measures approach increases sensitivity to cognitive change by reducing inter-individual variability and maximizing the number of observations per participant, ensuring the robust detection of within-subject effects. We used a mixed model to analyze the pattern of the participants’ cognitive deterioration over the study period. We examine the relationship between objective and subjective cognitive tests using correlations and partial correlations. We applied the Bonferroni correction for multiple comparisons to control the increased risk of Type I errors. We evaluated the reliability and validity of the Hebrew-translated SCD questionnaire by employing Cronbach’s alpha, test–retest correlations, and correlations with both related and unrelated constructs. We used repeated measures ANOVA to assess changes in cognitive function over time while controlling for potential practice effects with post hoc analyses designed to detect significant changes in performance.

3. Results

3.1. Participants Characteristics

Participants were highly educated, and they were also quite active physically and socially (see Table 1). Throughout the years, the average Z scores for all neuropsychological tests consistently fell within the normal range associated with age norms (0.23–0.53).

Their performance on the knowledge subtest of the WAIS-III, a reliable indicator of crystallized intelligence, was notably high (mean z = 1.38, SD = 0.80, indicating 1.3 SD above the age-matched average norm). Furthermore, the average BDI (M = 5.5, SD = 2.56) and STAI (M = 24.08, SD = 9.84) scores fell within the normal range. The participants did not exhibit clinically significant levels of symptoms related to depression or anxiety.

3.2. Change in Objective Assessments

We assessed the differences between the participants’ cognitive performance at the beginning of the study (T0) and the last three time points (T5, T6, T7), during which we introduced the SCD questionnaire in response to their subjective complaints regarding cognitive decline. The outcomes of the repeated measures of one-way ANOVA for the neuropsychological tests are presented in

Table 2. Analysis of changes in objective measures over time using repeated measures ANOVA.

Time Point/ Test	T0 M (SD)	T5 M (SD)	T6 M (SD)	T7 M (SD)	F(57)	p	ηp2
ROCFT copy	1.1 (0.52)	0.4 (0.79)	0.1 (0.93)	0.1 (0.84)	9.05	<0.001	0.32
ROCFT delay	0.8 (0.71)	1 (0.8)	0.9 (1.03)	1 (0.95)	0.51	0.67	0.03
PF	0.2 (0.92)	0.4 (1.03)	0.6 (0.97)	0.6 (0.83)	1.26	0.29	0.06
SF	0.5 (1.18)	0.4 (0.92)	0.5 (1.29)	0.4 (1.3)	0.10	0.90	0.01
RAVLT trial six	0.6 (1.34)	0.3 (0.88)	0.1 (1.12)	−0.2 (1.23)	2.16	0.10	0.10
TMT B	0.7 (0.53)	0.8 (0.46)	0.4 (0.47)	0.1 (0.74)	5.81	<0.001	0.26

Note. N = 20. M = mean z score. SD = standard deviation. ROCFT = Rey–Osterrieth Complex Figure Test. RAVLT = Rey Auditory Verbal Learning Test. PF = phonemic fluency. SF = semantic fluency. TMT = Trail-Making Test. The bold results indicate significant contrasts that were hypothesized in advance. p values less than 0.05 are italicized.

. The influence of gender on the objective variables was examined and found to be insignificant. Consequently, gender was not included as a covariate in the ANOVA. In line with our initial hypothesis, a significant decline was observed in the ROCFT copy score over time (see Table 2). Subsequent follow-up analyses demonstrated statistically significant declines between the initial assessment and all other time points (p < 0.05). Furthermore, in line with our primary hypothesis, a marked decline in the score of the RAVLT trial six from the study’s commencement to its conclusion was also identified (the contrast between T0 to T7), F(1,19) = 7.32, p < 0.05, ηp2 = 0.28. Contrary to our primary hypothesis, SF did not demonstrate a significant decline between the initial assessment and all subsequent time points (see Table 2). For a comprehensive overview of the results in additional RAVLT test trials over the years, see Supplementary Table S1. Scores on all assessments conducted between T1 and T4 have been previously published [68].

Table 2 also includes the variance analysis results for the TMT B test. Although we did not initially speculate on the changes in this test a priori, a marked decline in the TMT B score from the study’s commencement to its conclusion was identified. The contrast between T0 and T7 was significant, F(1,19) = 7.28, p < 0.05, ηp² = 0.28.

Figure 1 presents a visual representation through a linear mixed model depicting the trajectory for the ROCFT copy task, encompassing all study participants, including those who discontinued their involvement. The general trend of the performance in this test is shown in Figure 1 by a bold red line. Notably, 17 out of the 20 participants who completed the study (T0 to T7) showed a decrease in their scores over time. It is pertinent to highlight that two participants who discontinued their involvement in the study due to medical reasons exhibited a noticeable decline in their ROCFT copy scores in year T5, (participants 16 and 19). Furthermore, at T6, the three participants who displayed the most substantial decrease in ROCFT copy scores were the same individuals who did not participate in the last year of the study (participants 15, 21, and 24), which made it seem that there was an upward trend between T6 and T7, despite the general trend of decline in this test. Among these three, two received a diagnosis of dementia (participants 15 and 21), while the third declined participation due to personal reasons (participant 24). Therefore, the false rise in T7 is attributed to the absence of those three subjects. Nonetheless, it is evident that in comparison to T5, the overall mean score in the test has declined.

3.3. Change in Subjective Assessments

To explore the relationship between subjective and objective measurements, our initial step was to investigate differences in participants’ subjective aspects during the last three time points of the study, coinciding with the emergence of reported cognitive decline. The outcomes of the repeated measures ANOVA for the subjective measures are summarized in

Table 3. Analysis of changes in subjective measures over time using repeated measures ANOVA.

Time Point/ Questionnaire	T5 M (SD)	T6 M (SD)	T7 M (SD)	F (38)	p	ηp2
BDI (0–62)	5.2 (3.74)	6 (3.59)	5.6 (2.54)	2.17	0.13	0.11
STAI (20–80)	29.1 (7.96)	30.8 (8.94)	28.9 (9.45)	0.73	0.49	0.04
SCD50 (0–61)	24.7 (12.04)	28.1 (12.26)	26.4 (10.38)	4.54	0.02	0.19
SCD21 (0–27)	10.2 (5.18)	12.7 (5.08)	11.72 (4.3)	7.06	<0.01	0.27

Note. N = 20. M = mean score. SD = standard deviation. BDI = Beck Depression Inventory. SCD = subjective cognitive decline. STAI = State–Trait Anxiety Inventory. p values less than 0.05 are italicized.

. We assessed the potential impact of gender on the subjective variables and found that except for the BDI index in year T5, t(18) = −2.24, p < 0.05, where women exhibited higher levels of depression than men, gender did not significantly influence the subjective variables. Consequently, gender was included as a covariate in the relevant ANOVA for BDI. This addition yielded a noteworthy result in the depression contrast, indicating a significant increase between time points T5 and T6, F(1,18) = 7.35, p < 0.05, ηp² = 0.29. Furthermore, we identified a significant increase in the SCD50 score over time (see Table 3). Subsequent follow-up analyses revealed a significant increase in SCD50 scores between T5 and T6 (p < 0.05). Similarly, a significant increase was observed in the SCD21 score over time (see Table 3) with a subsequent follow-up analysis showing a significant increase in SCD21 scores between T5 and T6 (p < 0.001).

3.4. Preliminary Validation of the Hebrew SCD Questionnaire

We evaluated the reliability of the SCD50 and SCD21 at three specific time points (T5, T6, T7). The extended questionnaire exhibited remarkably high Cronbach’s alpha values with coefficients of 0.95 at T5, 0.96 at T6, and 0.93 at T7. Similarly, the abbreviated questionnaire demonstrated good internal consistency with Cronbach’s alpha coefficients of 0.90 at T5, 0.89 at T6, and 0.81 at T7. Moreover, to assess the test–retest reliability of the questionnaire, we calculated the correlation between responses collected over the three years for the extended and abbreviated versions. Remarkably high to excellent test–retest reliability was observed for both versions (see

Table 4. Inter-version reliability and test–retest reliability of the SCD questionnaire.

Measure	SCD21_T5	SCD50_T5	SCD21_T6	SCD50_T6	SCD21_T7	SCD50_T7
SCD21_T5	-	0.98 **	0.87 **	0.89 **	0.78 **	0.85 **
SCD50_T5		-	0.80 **	0.80 **	0.78 **	0.89 **
SCD21_T6			-	0.97 **	0.80 **	0.93 **
SCD50_T6				-	0.78 **	0.93 **
SCD21_T7					-	0.94 **
SCD50_T7						-

Note. N = 27 at T5, N = 24 at T6, N = 20 at T7. SCD = subjective cognitive decline. ** p < 0.01.

). Figure 2 illustrates the test–retest correlation analyses of all the participants for the SCD50 test scores at T5, T6, and T7. Similarly, Figure 3 illustrates the test–retest correlation analyses of all the participants for the SCD21 test scores at T5, T6, and T7. Furthermore, the correlation between the two versions of the questionnaire consistently demonstrated a high level of association across all three years, as outlined in Table 4.

To further establish the validity of the SCD questionnaire, we assessed its discriminant and convergent validity. We observed moderate positive correlations between the questionnaires and related constructs (anxiety and depression), as outlined in Supplementary Tables S2–S4. Additionally, moderate to high positive correlations were found between the SCD questionnaire and cognitive decline, determined as the score difference between T0 and T7, using the ROCFT copy test (Δ ROCFT copy T0–T7), as outlined in

Table 5. Discriminant and convergent validity of the SCD questionnaire.

Measure	SCD21_T5	SCD50_T5	SCD21_T6	SCD50_T6	SCD21_T7	SCD50_T7
Δ ROCFT copy T0–T7	0.34	0.42	0.52 *	0.55 *	0.44	0.64 **
Agreeableness BFI (T7)	−0.08	0.09	0.02	0.06	0.05	0.05

Note. N = 27 at T5, N = 24 at T6, N = 20 at T7. SCD = subjective cognitive decline. ROCFT = Rey–Osterrieth Complex Figure Test. BFI = Big Five Inventory. Δ ROCFT copy T0–T7 = the difference between the ROCFT copy score at T7 and T0. * p < 0.05, ** p < 0.01.

. Conversely, minimal correlations were observed between the SCD questionnaire and the unrelated construct (agreeableness from the BFI questionnaire), as outlined in Table 5.

4. Discussion

The current study spans over a decade, offering a significantly longer follow-up compared to our previous 12-month study [40]. By integrating subjective aspects of cognitive decline and exploring their relationship with objective cognitive decline, this research highlights the potential of SCD as a critical marker. This is particularly relevant for highly educated individuals, providing a nuanced perspective on early detection and the progression of cognitive trajectories. The primary objective of this study was to identify effective methods—encompassing both objective and subjective assessments—for the early detection of cognitive decline in highly educated older adults over a decade. This aim stemmed from the challenges of diagnosing objective cognitive decline in this population, using single-point assessments on the one hand, and addressing the practice effect in longitudinal studies on the other, due to their significant cognitive reserve [5,6].

In line with these goals, the decision to focus on time points T5 to T7 was based on both theoretical and empirical considerations. Subjective memory complaints—central to the study’s aims—began to emerge only at T5. Prior to that point, no participants reported subjective decline, rendering earlier time points less relevant to the questions under investigation. This pattern also aligns with the study’s inclusion criteria, which required that participants exhibit no evidence of either subjective or objective memory decline at baseline. Thus, it is both expected and methodologically appropriate that subjective cognitive concerns would first be observed only after several waves of follow-up. In the objective assessment, the study focused on three neuropsychological tests that have previously shown efficacy for detecting cognitive decline after one year (ROCFT copy, SF, and RAVLT trial six) [40]. The study aims to examine which of these tests will remain effective for detecting cognitive decline over an extended period, spanning from T0 to T5–T7 (from baseline to 60 months, 72 months, and 84 months). In the subjective assessment, we focused on depression, anxiety, and subjective cognitive decline as potential predictors of objective cognitive decline. We examined the relationships between the subjective measures and the objective measures (ROCFT copy, SF, and RAVLT trial six) at specific time points (T5, T6, and T7) and across the study period (from T0 to T5–T7). Acknowledging the absence of an appropriate SCD questionnaire in Hebrew, the secondary goal was to conduct an initial validation of Gifford’s SCD questionnaire [17] in the Hebrew language.

4.1. Objective Assessments

In line with our primary hypothesis, a substantial decline with strong effect sizes was evident in the ROCFT copy phase and RAVLT trial six when comparing the initial assessments (T0) with subsequent follow-up evaluations (T5–T7). The ROCFT evaluates participants’ ability to copy a complex multi-part shape, involving various functions such as visual–spatial perception, grapho-motor skills, attention, visual memory, and executive functions like planning, organization, and attention to detail [37,38,69]. Comparisons with performance in other tests assessing similar cognitive domains, and the second part of the ROCFT test, in which participants reproduce the complex shape from memory, offer insights into potential areas of cognitive decline. During the initial drawing phase of the task, some participants exhibited errors by adding or omitting details. Nevertheless, their overall perception of the shape appeared to remain intact. In contrast, participants maintained a consistent level of performance in the delayed memory phase, which was similar to their initial performance at the beginning of the study. However, drawing conclusions about cognitive function from this phase of the test can be challenging due to a practice effect observed during the delayed section of the ROCFT, arising from participants becoming more familiar with the test over time, as demonstrated in previous studies [25]. Consequently, after their initial exposure to the test during the drawing phase, participants can more easily recall and apply strategies that proved effective in the earlier copying phase. This can lead to decreased effectiveness in detecting cognitive decline in the delayed segment compared to the copying phase.

The conclusions drawn from participants’ performances on the ROCFT and other neuropsychological assessments indicate that their cognitive decline is unlikely to be solely attributed to a decline in visual perception or motor skills [37,38,69]. These findings gain additional support from the results of the MoCA subtest, which involves replicating a simpler two-dimensional cube, in which all participants successfully recreated the shape. Moreover, other evaluations related to attention (e.g., TMT A, digit recall) and memory (e.g., story recall) did not display a similar decline, pointing toward a probable reduction in the executive functions required for the complex ROCFT task. The conclusion gains further support from the consistent decline in the sixth trial of the RAVLT, which is a task that heavily relies on executive functions [47]. This specific trial evaluates the ability to learn a new word list (list 6) after repeated exposure to previously learned lists (lists 1–5). Decreased performance in the sixth trial indicates proactive interference, whereas previously acquired information obstructs the assimilation of new material [70]. Additional support for this conclusion comes from the results of the TMT B test, which also relies heavily on executive functions [44] and also showed a significant decrease over the years. Multiple studies have documented that older adults are notably vulnerable to this phenomenon, reflecting challenges in executive function, particularly in the suppression of irrelevant information such as the previously learned list [71]. In summary, it can be deduced that highly educated adults encounter executive function decline when engaging in complex tasks in their later years. These challenges lead to a decline in their scores compared to their initially high baseline scores. However, their scores still fall within the normal range when compared to the broader population (up to 1.5 standard deviations below the mean) [72]. Remarkably, within this demographic, these challenges may manifest before memory or perceptual difficulties, which are typically observed in less educated older adults [73].

Strengthening the assertion that these tests are finely tuned to detect cognitive decline is the minimal impact of the practice effect on assessments introducing novel information, such as the ROCFT copying phase [25] and the RAVLT Trial Six [74]. The minimal practice effects underscore the increased responsiveness of these tests in detecting cognitive decline in our ongoing study. This enhanced effectiveness is attributed to the cognitive demands inherent in these assessments and their reliance on executive functions [27,28,69], which is in line with similar findings in related research [29,39,40]. Notably, the ROCFT copying test showcases exceptional effectiveness within this population, as evident from the continued detection of significant decline, even when individuals experiencing the most pronounced decline opted not to participate in the study in subsequent years. Hence, it can be concluded that those tests, and especially the ROCFT copy, may serve as a valuable indicator of the early detection of objective cognitive decline [26] within this specific group of highly educated older adults.

To deepen our understanding of the properties underlining the significant decline in performance on ROCFT, we further analyzed the data of the digit span subtests and the TMT trail B since they are considered to rely heavily on executive functions [37,38,44].

The results (Supplementary Table S5) indicate that the mean of digit span subtests remained relatively stable over the years, with no significant changes observed, particularly in digit span backward, which primarily assesses working memory—a key component of the executive function system in some models [75]. In contrast, TMT trail B, evaluating both working memory and shifting abilities (letters and numbers), displayed a notable decline over time, especially from T5 to T7 (Supplementary Table S5). Intriguingly, this subtest shares a reliance on visuo-spatial abilities with the ROCFT copy. The visuo-spatial component introduces additional complexity to the tasks, suggesting that this combination may contribute to the sensitivity of these tests in detecting decline over time, particularly in highly educated older adults. The widespread use of assessments like the ROCFT Copy and TMT-A and B in studies of highly educated adults, as highlighted in a meta-analysis [76], further underscores the importance of these tests in evaluating cognitive decline.

In contrast to our initial hypothesis, the SF test did not demonstrate a significant decline beyond the anticipated age-related changes. This test evaluates the capacity to recall as many words as possible from a specified category within a minute, drawing upon vocabulary, verbal memory retrieval, and executive functions, specifically shifting and clustering strategies [31]. While the raw scores of the SF test did decrease (please see Supplementary Table S6), standardized scores did not manifest this decline despite prior research indicating age-related declines in language function using the SF test [31,36,77]. This absence of significant cognitive decline in the standardized scores can be attributed to several factors. Firstly, it is possible that the cognitive reserves of highly educated older adults may mitigate age-related declines in certain cognitive domains [78]. Notably, the norms of the VF test were based on participants with an average of 12.76 years of education, which is significantly lower than the average of 17.6 years of education in our study. Secondly, while the ROCFT and RAVLT trial six primarily rely on fluid intelligence due to their basis in executive functions [36,47], the SF test predominantly draws on crystallized knowledge (vocabulary) and, to a lesser extent, on fluid intelligence (extracting words from categories or letters). Thus, the SF test predominantly engages language areas in the brain [45], suggesting that the extensive semantic knowledge of highly educated older adults [79] may mask noticeable declines when using existing norms. Thirdly, fluctuations in semantic fluency test scores, declining after one year [40] but not after a decade, may result from adjusted norms at T1 compared to T7. Initially, participants were categorized into different age groups (under 70,70–74,75) based on test norms [45]. However, at T7, all participants were in the same age group of over 75, without separation, as there are no tailored norms for older adults. This underscores the importance of developing norms for neuropsychological tests, considering the sensitivities of older adults and prioritizing years of education in addition to age [40].

4.2. Subjective Assessments

Our investigation into the complex interplay between subjective measures (SCD, depression, anxiety) and objective cognitive functions/cognitive decline at T5–T7 uncovered notable dynamics. There was a notable increase in depression and SCD observed between T5 and T6 (2020–2022), which was succeeded by a modest decrease and stabilization at T7.

The surge in emotional measures during T5–T6 can be attributed to the initial waves of the COVID-19 outbreak, which are marked by heightened uncertainty and perceived life-threatening circumstances. This period coincided with prolonged lockdowns, aligning with recent studies emphasizing the substantial influence of COVID-19 on subjective aspects [80], including heightened levels of depression and anxiety observed in older adults. Remarkably, our highly educated study participants demonstrated a relatively rapid recovery, which was possibly due to their robust cognitive reserves contributing to enhanced emotional regulation and resilience to psychological distress [5,6,81]. This aligns with the psychological adaptation process commonly observed after negative life events, leading to resilience or complete recovery [82].

However, when considering cognitive decline over the years, we observed moderate to high positive correlations between SCD (T5–T7) and cognitive decline, as evidenced by the alteration (Δ) in ROCFT copy scores from T0 to T7. Even after adjusting for anxiety and depression, the partial correlations of SCD with cognitive decline remained moderate. In comparison, correlations between depression or anxiety at T5–T7 and cognitive decline, assessed by Δ ROCFT copy T0–T7, were minimal and of smaller magnitude than the correlations observed between SCD and Δ ROCFT copy T0–T7 during the same timeframe. Hence, anxiety and depression failed to predict cognitive decline over time. These findings align with other studies establishing a meaningful connection between SCD and objective cognitive decline [12,13,23] and support studies that did not find a significant association between anxiety or depression and cognitive decline [83]. Consequently, it can be concluded that SCD may serve as a valuable predictor of cognitive decline in highly educated older adults, potentially surpassing the predictive power of depression and anxiety due to its specificity for cognitive decline.

These findings resonate with a growing body of longitudinal research examining the complex relationship between educational attainment and cognitive aging. For example, large-scale studies of initially healthy older adults have demonstrated that while higher education is consistently associated with superior baseline cognitive performance, it does not appear to moderate the rate of cognitive decline over time [84,85]. Similarly, in clinical populations, education has not been shown to alter the trajectory of decline once Alzheimer’s disease is manifest [20]. These results highlight a crucial distinction between cognitive level and cognitive change: whereas education and by extension cognitive reserve may delay the onset of clinical symptoms, it does not necessarily prevent decline itself. This distinction is central to theoretical models of cognitive reserve, reinforcing the importance of long-term, repeated assessments to differentiate between initial performance advantages and actual protective effects against cognitive deterioration. Our findings support this perspective, emphasizing the value of combining both subjective and objective measures over time to detect meaningful change in highly educated individuals.

4.3. The Hebrew SCD Questionnaire

Additionally, this study serves as an initial validation of the Hebrew-translated SCD questionnaire, demonstrating high internal and test–retest reliability across multiple time points for both test versions. In assessing the questionnaire’s convergent and discriminant validity, we explored its associations with related constructs like depression and anxiety, aligning with prior studies [8,13,14,33,86]. We further examined its correlation with the related construct of cognitive decline [8,13,14,86] assessed by the difference in ROCFT copy scores between T0 and T7. Furthermore, we investigated the questionnaire’s associations with unrelated constructs, such as agreeableness from the BFI [14]. The significant positive correlations observed between SCD and cognitive decline, moderate positive correlations with psychological distress, and zero correlation with agreeableness strongly validate our questionnaire. The findings align with research that found similar associations of SCD with cognitive decline and psychological distress among cognitively healthy older adults [86]. This further confirms the questionnaire’s suitability as a diagnostic tool for assessing cognitive decline among highly educated older adults.

The current findings support our hypothesis that SCD is a meaningful predictor of future objective cognitive decline, especially among highly educated older adults. However, the novelty of this study does not rest solely on the predictive role of SCD, which has been established in previous work. Rather, our findings contribute to the literature by demonstrating that SCD remains a robust predictor in a population characterized by high cognitive reserve, where early detection is particularly challenging. Moreover, by including both subjective and objective measures, our study reveals that SCD predicts future cognitive decline above and beyond the effects of mood symptoms such as depression and anxiety, which are often comorbid and can confound interpretations of self-reported complaints. This highlights the utility of combining self-report with objective testing in longitudinal designs, especially when seeking early markers of cognitive decline in highly educated individuals.

4.4. Clinical and Research Implications

The study sheds light on the effectivity of objective and subjective methods for detecting early cognitive decline in highly educated older adults with potential implications to assist cognitive health diagnosis and intervention. Notably, the results emphasize the heightened effectivity of objective methods over subjective ones. The robust effect sizes observed in the ROCFT copy and the RAVLT trial six and TMT trial B underscore their effectiveness even within a population of well-educated older adults.

Among the subjective methods assessed in this study, the SCD questionnaire emerges as a standout, exhibiting a moderate to high correlation with cognitive decline and proving to be an important tool for predicting objective cognitive decline. Recognizing the predictive value of SCD in the initial stages of MCI [87], and its potential to precede objective assessments [8,23], this study recommends incorporating the SCD questionnaire as an initial indicator alongside with the objective tools. In this population, it should take precedence over instruments designed for assessing anxiety and depression.

The study’s conclusions provide valuable insights for the early identification of individuals at risk of subsequent objective cognitive decline [12]. By endorsing the use of both nuanced neuropsychological tests and SCD questionnaires, healthcare professionals can implement early intervention strategies. These interventions, including cognitive training, social engagement, lifestyle modifications, and appropriate medical management, are widely recognized for their role in reducing the risk of cognitive decline and Alzheimer’s disease across populations [5]. However, highly educated individuals are often already engaged in such activities, raising the question of whether additional engagement provides further benefits or whether cognitive reserve enables them to sustain performance despite underlying decline. Future research should explore how these factors interact in this population and whether intervention strategies should be adapted accordingly. Based on our findings, we propose several practical applications: (1) routine screening for subjective cognitive decline (SCD) during medical visits using validated self-report tools; (2) the enhancement of neuropsychological assessments with executive function-sensitive tasks and longitudinal tracking; (3) public education encouraging older adults, particularly the highly educated, to report subtle cognitive concerns; and (4) the development of personalized cognitive interventions that go beyond generic training, incorporating novel, executive function–focused challenges. These strategies may facilitate earlier detection and a more effective prevention of functional decline.

4.5. Limitations and Future Directions

This study was subject to several limitations. First, the small sample size challenges the identification of statistical significance, and generalization should be made with caution, necessitating the consideration of effect size. The sample size also constrained our ability to use advanced methods like growth curve modeling. Second, the COVID-19 pandemic’s outbreak during the sixth year of follow-up (T6) may have influenced cognitive and emotional responses, warranting careful interpretation. Third, the absence of SCD questionnaire data at the study’s inception (T0) limits causal understanding of the relationships between cognitive decline, SCD, anxiety, and depression. Fourth, the study’s external validity is constrained by the cohort’s homogeneity, comprising predominantly well-educated individuals of white ethnicity. Our participants were initially selected at T0 as cognitively healthy older adults with no reported subjective complaints, which could suggest a potential selection bias. However, our longitudinal design specifically aimed to track the emergence of subjective complaints over time. By focusing our analyses on T5, T6, and T7, when some participants began to report SCD, we were able to investigate how new onset subjective complaints relate to subsequent objective cognitive changes. Therefore, while the initial selection at T0 ensured a homogenous and healthy baseline, our findings reflect meaningful within-subject variability that emerged over the follow-up period. Lastly, the participants’ repeated exposure to the same neuropsychological tests raises concerns regarding familiarity. However, research indicates that these tests maintain their effectiveness in tracking cognitive capabilities across multiple uses even in the face of familiarity [31,48,51]. The observed significant decline in performance over time further suggests genuine cognitive decline.

Future studies should prioritize the integration of longitudinal MRI data to deepen our understanding of the relationship between cognitive decline and brain structure. Such data will allow researchers to better distinguish between successful aging and abnormal cognitive trajectories, shedding light on the neurobiological mechanisms underlying these patterns.

Another crucial direction is the development of localized normative data that account for varying levels of educational attainment. Implementing such norms would enhance the sensitivity of cognitive tests, especially in highly educated older adults, enabling a more accurate detection of early cognitive decline. This approach would also provide more specific insights into particular cognitive domains that may be affected, offering a clearer understanding of the cognitive profiles of this population.

5. Conclusions

In sum, this study highlights the importance of combining subjective and objective assessments to identify early cognitive changes in highly educated older adults. By demonstrating the predictive utility of SCD alongside neuropsychological markers of executive decline, it offers a valuable framework for improving early detection in cognitively resilient populations. These findings underscore the need for tailored diagnostic approaches that consider cognitive reserve and intra-individual decline over time.