Abstract
Introduction:
COVID-19 has affected millions worldwide and has been associated with persistent symptoms across various physiological systems. Among these, executive dysfunction (ED) has been frequently reported in long COVID, potentially compromising academic performance and autonomy in daily activities. This study aimed to evaluate Executive Functions (EF) in medical students during the long-term post-infection phase, specifically controlling for fluid intelligence (g-factor) and psychiatric symptoms to distinguish potential viral sequelae from baseline characteristics.
Methods:
A cross-sectional design was employed to evaluate 49 medical students (32 women). Data collection occurred between February and September 2023. Infected participants (n = 26) were assessed on average 20.81 ± 7.35 months post-infection. Psychiatric symptoms were screened using the Depression, Anxiety, and Stress Scale (DASS-21). Additionally, the Paper Folding and Cutting (PF&C) task served as a non-verbal proxy for fluid intelligence (PF&C) task was administered as a non-verbal proxy measure for fluid intelligence (g-factor) to establish a cognitive baseline. EFs were assessed using the 2-back test (working memory), Stroop Test (inhibitory control), and Wisconsin Card Sorting Test – WCST (cognitive flexibility).
Results:
Participants had a mean age of 22.2 years. All were vaccinated prior to testing, with the majority having received three or more doses. Among the infected group, 17 were unvaccinated at the time of infection. Regarding cognitive performance, the infected group exhibited superior accuracy (lower error rates) in working memory and inhibitory control tasks. Similarly, in the WCST, infected participants required significantly fewer trials to complete the first category (p = 0.012), indicating greater initial efficiency, while total categories completed remained similar between groups. Crucially, ANCOVA analyses revealed that this high-level performance was significantly accounted for by fluid intelligence (g-factor) rather than limited by infection history. Scores on the DASS-21 did not differ significantly, suggesting that current psychiatric symptoms did not confound the cognitive outcomes.
Conclusion:
Our findings demonstrate a robust preservation of executive domains, suggesting that high cognitive reserve buffers against neurocognitive sequelae approximately two years post-infection. Consequently, specialized cognitive monitoring may not be necessary for this population. Instead, educational resources should prioritize mental health support to address academic stressors.
1 Introduction
The COVID-19 pandemic caused by SARS-CoV-2 has resulted in millions of infections worldwide (Rutter et al., 2025). Beyond the acute respiratory manifestations, a significant proportion of survivors report persistent symptoms, a condition commonly known as post-COVID-19 syndrome or long COVID (Saucier et al., 2023). Mounting evidence indicates that SARS-CoV-2 possesses neurotropic capabilities, potentially leading to central nervous system alterations (Hugon et al., 2022). Notably, longitudinal imaging studies comparing pre- and post-infection data have identified structural changes in the brain, including a reduction in gray matter thickness in the orbitofrontal cortex and parahippocampal gyrus (Deuter et al., 2024; Douaud et al., 2022). These regions are functionally critical for higher-order cognition, particularly executive functions (EF) and may be affected by long COVID (Ardila and Lahiri, 2020).
Executive functions encompass a set of cognitive processes responsible for the top-down control of behavior, emotion regulation, and complex reasoning (Diamond, 2013; Dias et al., 2015). While previous research has reported cognitive deficits proportional to infection severity (Serafim et al., 2024), interpreting these findings often proves challenging due to methodological limitations such as the absence of control groups, the failure to account for mental health variables, or the lack of a precise description of the recovery period. It is well-established that conditions such as depression, anxiety, and stress - prevalence of which increased significantly during the pandemic (Damiano et al., 2022; Rebello et al., 2022) - can independently impair cognitive performance, acting as confounding variables in the assessment of viral sequelae. Furthermore, some studies have indicated that recovery periods of at least twelve months may be sufficient for the remission of long COVID symptoms.
Understanding the long-term persistence of these deficits is particularly critical for populations under high cognitive demand, such as university students, with consequences for well-being (Hamaideh and Hamdan-Mansour, 2014). For medical students in particular, intact executive functioning is vital not only for academic success but for the development of complex clinical reasoning, decision-making under pressure, and the regulation of cognitive resources during long shifts - core competencies of medical training (Ramos-Galarza et al., 2019). However, there is a scarcity of data regarding the persistence of executive dysfunction in young, vaccinated adults with high educational levels post-infection.
Crucially, individual differences in cognitive outcomes following brain insults are often modulated by cognitive reserve - the brain's capacity to optimize performance through differential recruitment of brain networks (Stern, 2009). Fluid intelligence (or the g-factor) serves as a robust proxy for this reserve (Salas et al., 2021), potentially buffering against the neurocognitive sequelae of infection. However, few studies have controlled for this baseline when assessing long-term COVID-19 outcomes, despite recent evidence suggesting that high cognitive reserve significantly mitigates post-infection deficits (Costas-Carrera et al., 2022).
Therefore, this study aimed to evaluate executive function performance in university students previously infected by COVID-19 compared to a non-infected control group approximately twenty months after infection. By controlling for psychiatric symptoms and fluid intelligence, we sought to determine whether specific viral-induced cognitive sequelae persist over this extended recovery period in this specific demographic.
2 Methods
2.1 Participants and ethics
This was a quantitative, experimental, and cross-sectional study. All procedures were conducted in accordance with the Declaration of Helsinki and approved by the Institutional Ethics Committee (CAAE: 62315722.9.0000.5515), following the guidelines of the Brazilian National Commission for Research Ethics (CONEP). The study was pre-registered on Brazilian Registry of Clinical Trials (ReBEC).
Male and female students from a medical program at a private university in Western São Paulo were invited to participate. While the study design did not include a control group with lower educational attainment, the recruitment strategy prioritized a homogeneous high-functioning population to specifically investigate neurocognitive resilience mechanisms in high-stakes educational environments. To address the variability in cognitive reserve within this selected sample, fluid intelligence (g) was measured and treated as a continuous covariate in all analyses (see Sec. 2.3). Recruitment was conducted through social media using messages that briefly described the study and provided contact information for the researchers. No participant data was accessed or collected during this initial recruitment phase. Following an a priori estimation of the minimum sample size (see Sec. 2.5), a total of 51 participants responded to the invitations. Of these, one subject was excluded for not completing the tests, and another due to a prior psychiatric diagnosis not currently under treatment, presenting with acute symptoms. All other participants with a history of psychiatric disorders were currently receiving treatment. Psychiatric symptoms were screened using the Depression, Anxiety, and Stress Scale – Short Form (DASS-21) (Apóstolo et al., 2006). Additionally, the Ishihara Test was used to screen for dyschromatopsia (Zarazaga et al., 2019). The final sample consisted of 49 participants, divided into a Control Group (n = 23; no history of COVID-19 diagnosis) and a COVID Group (n = 26; history of confirmed COVID-19 diagnosis). Infection history and vaccination status were determined via self-report during the initial anamnesis, consistent with standard observational protocols at the time of data collection. Data collection took place between February and September 2023. This period corresponds to an average interval of approximately 20 months post-infection for the COVID group (see Table 2), reflecting the specific cross-sectional window in which participants were available for assessment.
2.2 Apparatus and setting
Test protocols were programmed using PsychoPy software (Peirce et al., 2019) and executed on a desktop computer running Windows® 10 (64-bit), equipped with an Nvidia GTX 1660 Super graphics card, 16GB of RAM, and a 24-inch monitor with 1,080 p resolution and a 120 Hz refresh rate. Tests were administered in a dedicated university laboratory with controlled lighting and temperature. Participants were seated comfortably approximately 60 cm from the screen, with the keyboard positioned according to their preference.
2.3 Instruments and procedures
A sociodemographic and clinical questionnaire was used to collect data regarding socioeconomic status and clinical history. To ensure data confidentiality, participants were de-identified in the neuropsychological test result files. Specifically regarding COVID-19, the inclusion criteria required a laboratory-confirmed diagnosis via documented positive RT-PCR and/or rapid antigen test results. Other variables considered included clinical signs during the disease course, temporal relationship, disease severity (hospitalization, intubation, respiratory complications), timing of infection relative to vaccination, number of immunization doses, and the presence of long COVID symptoms. It is important to note that while we monitored for lingering symptoms, participants were not recruited based on a formal clinical diagnosis of “Long COVID” or “COVID brain fog” by a specialist. Rather, the presence of any persistent post-acute symptoms was documented as part of the clinical history obtained during the interviews. All COVID-19-related data were obtained through self-report during the clinical interview, where participants reported any perceived cognitive sequelae. Although medical records were not systematically cross-verified, participants were encouraged to consult their official digital or physical vaccination certificates during the session to maximize accuracy and minimize recall bias.
The neuropsychological assessment was theoretically grounded in the general factor of EF, which posits the unity of three separable but related components: inhibitory control, working memory (updating), and cognitive flexibility (shifting) (
Diamond, 2013;
Dias et al., 2015;
Friedman and Miyake, 2017). The selected battery comprised standardized tasks widely recognized for their sensitivity to these specific domains:
Wisconsin Card Sorting Test (WCST): Assessed cognitive flexibility (shifting), the mental ability to adjust to changing environmental demands. The WCST remains a gold-standard measure for this EF dimension (Miles et al., 2021). The computerized version followed the color-form-number sequence proposed by Heaton et al. (cited in Miles et al., 2021). No explicit instructions regarding the sorting rule were given; participants were required to deduce the rule through trial and error. The test concluded once six categories were completed or the deck of response cards was exhausted. Additionally, the test was terminated if 40 consecutive incorrect responses occurred - a condition suggesting an inability to comprehend instructions, potentially compromising rapport and subsequent test performance (Lezak et al., 2012).
2-Back Test: Assessed working memory (updating) using an n-back task (2-back condition) involving a random sequence of letters presented at the center of the screen (Owen et al., 2005). Yaple et al. (2019) argue that the n-back task is the most common measure of working memory performance. A single test session consisted of 240 trials divided into ten blocks of 24 trials each, with fixed 15-second intervals between blocks. A practice block was administered prior to the test, allowing participants to clarify any doubts. Each letter remained visible until the participant responded, with a maximum duration of 2000ms.
Stroop Test: Assessed attention and inhibitory control (Stroop, 1935; Washburn, 2016). This test was administered after the 2-back task, following a break during which the experimenter explained the procedure. The test consisted of ten blocks of 32 trials each. Participants were instructed to respond to the font color of the stimulus word. Stimuli were presented in congruent conditions (word meaning matches font color) and incongruent conditions (word meaning differs from font color). A practice block was provided before the actual test and fixed 15-second intervals separated the test blocks.
Paper Folding and Cutting Task (PF&C): Assessed visual/abstract reasoning and was adapted from the Stanford-Binet Intelligence Scale, Fourth Edition (SB4) (Nantais and Schellenberg, 1999). Crucially for this study, the PF&C was employed to operationalize Cognitive Reserve (CR). Given the educational homogeneity of medical students, traditional proxies for CR such as “years of education” are insufficient to capture variability. The PF&C measures fluid intelligence (Gf), representing raw processing capacity independent of acculturation or verbal knowledge (Haier, 2017). According to Youngstrom et al. (2003), this task holds the highest reliability index (0.94) among SB4 subtests for measuring the general ability factor (g). Thus, individual differences in g served as a proxy for cognitive reserve to control for baseline cognitive variability. The test comprised two sets of 17 figures, presented in increasing order of difficulty. In each trial, a rectangular piece of paper was shown at the top of the screen undergoing a sequence of folding and cutting manipulations. The bottom of the screen displayed five possible outcomes for the unfolded paper; participants were required to select the correct one. Participants completed only the first set of PF&C images to provide an estimate of g. Consequently, the obtained scores were employed as covariates in the statistical analyses of EF tests to strictly control for the influence of baseline CR on executive performance.
2.4 Procedure workflow
All tests were administered directly via PsychoPy in a single session. The experimental workflow followed a strict sequence to ensure consistency across participants. First, a screening and anamnesis were conducted to collect sociodemographic and clinical data via interview. Second, participants completed initial screening tasks that included the DASS-21 (psychiatric screening) and the Ishihara Test for color vision abnormalities, with the experimenter present to assist and ensure eligibility. Finally, each participant performed the main neuropsychological battery, comprising the cognitive tasks in the following fixed order: 2-Back Test (Working Memory), Stroop Test (Inhibitory Control), WCST (Cognitive Flexibility), and PF&C Task (Fluid Intelligence/Cognitive Reserve). All tests were preceded by standardized on-screen instructions. Participants were allowed brief rest periods between tests to minimize fatigue. The total duration of the session was approximately 30 to 40 min.
2.5 Statistical analysis
Sample size calculation was performed using G*Power 3.1.9.7 (Kang, 2021) to determine adequacy for mixed-model Analysis of Variance (ANOVA). Assuming a moderate effect size (f = 0.25), a 5% probability of Type I error (α = 0.05), and a 5% probability of Type II error (1-β = 0.95), a minimum sample of 48 participants was required.
Statistical analyses were performed using JASP software (JASP Team, 2025), based on a frequentist framework with weighted interpretation of p-values (Wasserstein et al., 2019). All data were tested for normality with Shapiro–Wilk tests. Results on the PF&C were compared through a one-way Analysis of Variance (ANOVA). Then, mixed-design repeated measures Analysis of Covariance (rmANCOVAs) were used to compare reaction times (RT) and error rates. Crucially, to isolate the effect of infection status from baseline cognitive ability, raw scores from the PF&C task (fluid intelligence) were entered as the covariate in all ANCOVA models. For the 2-back test, with a 2 (Group) × 10 (Block) design. For the Stroop test, an additional factor of Stroop interference (Congruent vs. Incongruent) was included, resulting in a 2 × 2 × 10 design. For all rmANCOVAs, the first factor was between subjects, while the remaining were within subjects. Finally, Kruskal–Wallis tests – non-parametric equivalents of ANOVA – were used to compare the number of perseverative responses, completed categories, and trials to the first category on the WCST between groups. When appropriate, omega squared (ω2) was used to estimate effect size, with values interpreted as negligible (ω2 < 0.01), small (0.01 ≤ ω2 < 0.06), medium (0.06 ≤ ω2 < 0.14), and large (ω2 ≥ 0.14) (Goss-Sampson, 2025). The same interpretive thresholds were applied to rank epsilon squared (ε2) for the Kruskal–Wallis tests.
3 Results
A one-way ANOVA showed that the Control and COVID-19 groups were comparable in age (F1,47 = 2.429, p = 0.126, ω2 = 0.028). All participants were vaccinated at the time of data collection, and the majority had been tested for COVID-19 via rapid tests or RT-PCR. Table 1 displays these and other sociodemographic and clinical data.
Table 1
| Variablea | Control (n = 23) | COVID-19 (n = 26) | χ² testb |
|---|---|---|---|
| Age (mean ± SD) | 23.00 ± 3.79 | 21.54 ± 2.75 | |
| Sex | |||
| Female | 16 (32.65) | 16 (32.65) | 0.773 |
| Male | 10 (20.41) | 7 (14.29) | |
| Prior psychiatric diagnoses | |||
| None | 14 (28.57) | 14 (28.57) | 0.654 |
| Anxity disorder | 5 (10.02) | 6 (12.24) | |
| Depressive disorder | 0 (0.00) | 2 (4.08) | |
| Mixed disorder (anxiety and depression) | 3 (6.12) | 2 (4.08) | |
| Otherc | 1 (2.04) | 2 (4.08) | |
| Timing of of psychiatric diagnosis | |||
| Pre-infection | 9 (18.37) | 9 (18.37) | 0.322 |
| Post-infection | 0 (0.00) | 3 (6.12) | |
| Number of vaccine doses (current) | |||
| Two doses | 5 (10.20) | 8 (16.33) | 0.733 |
| Three doses | 8 (16.33) | 9 (18.37) | |
| Four doses | 10 (20.41) | 9 (18.37) | |
| Testing method (rapid test or RT-PCR) | |||
| Yes, negative | 17 (34.69) | 0 (0.00) | <0.001 |
| Yes, positive | 0 (0.00) | 26 (53.06) | |
| Did not test (no symptoms) | 6 (12.24) | 0 (0.00) | |
Sociodemographic and clinical data of a sample of medical students from a private higher education institution in the interior of São Paulo state (n = 49).
Data expressed as n (% from total) unless otherwise indicated.
The Fisher–Irwin test was applied to contingency tables equal to 2 × 2.
Other diagnoses included Neurodevelopmental or Personality Disorders.
Participants in the COVID-19 group had been recovered for an average of 20.81 ± 7.35 months, and the majority of the COVID-19 group were infected prior to vaccination. None required hospitalization during the infection. A significant proportion (approximately one-third) reported symptoms following the acute phase or reported difficulties with memory or attention (Table 2).
Table 2
| Variablea | Female (n = 16) | Male (n = 10) | Χ2 testb |
|---|---|---|---|
| Recovery Time from infection (months) | 20.00 ± 8.23 | 22.10 ± 5.86 | |
| Range (Min/Max) | 3/30 | 8/30 | |
| Infection timing relative to vaccination | |||
| Unvaccinated | 10 (38.46) | 7 (26.92) | 0.871 |
| One dose | 1 (3.85) | 1 (3.85) | |
| Two doses | 4 (15.38) | 2 (7.69) | |
| Three doses | 1 (3.85) | 0 (0.00) | |
| Post-COVID-19 status | |||
| Symptoms after acute phase | 8 (30.77) | 2 (7.69) | 0.530 |
| Reported memory difficulties | 8 (30.77) | 2 (7.69) | |
| Reported attention difficulties | 6 (23.08) | 2 (7.69) | |
Clinical data related to COVID-19 infection in a sample of medical students (n = 26).
Data expressed as n (%), relative to COVID group only, unless otherwise indicated.
The Fisher–Irwin test was applied to contingency tables equal to 2 × 2.
Inferential analyses using logistic regression indicated that the presence of symptoms after the acute phase of COVID-19 infection did not predict self-reported difficulties in attention (χ224 = 0.642, p = 0.423) or memory (χ224 = 0.909, p = 0.340).
An rmANOVA for the DASS-21 scale revealed no significant differences between groups (F1,47 = 0.009, p = 0.924, ω2 = 0.000), indicating that symptom levels were equivalent. Crucially, there was no significant Group × Subscale interaction groups (F2,94 = 0.526, p = 0.593, ω2 = 0.000), demonstrating that the specific distributions of depression, anxiety, and stress were statistically identical between groups. The overall prevalence of depression, anxiety, or stress symptoms was low (Table 3).
Table 3
| Scale/Severity | Control (n, %a) | COVID-19 (n, %a) |
|---|---|---|
| Depression | ||
| No symptoms | 18 (78.26) | 17 (65.38) |
| Mild | 3 (13.04) | 3 (11.54) |
| Moderate | 1 (4.35) | 4 (15.38) |
| Severe | 0 (0.00) | 2 (7.69) |
| Extremely severe | 1 (4.35) | 0 (0.00) |
| Anxiety | ||
| No symptoms | 16 (69.57) | 18 (69.23) |
| Mild | 0 (0.00) | 1 (3.85) |
| Moderate | 3 (13.04) | 6 (23.08) |
| Severe | 3 (13.04) | 0 (0.00) |
| Extremely severe | 1 (4.35) | 1 (3.85) |
| Stress | ||
| No symptoms | 18 (78.26) | 19 (73.08) |
| Mild | 1 (4.35) | 4 (15.38) |
| Moderate | 1 (4.35) | 1 (3.85) |
| Severe | 2 (8.7) | 2 (7.69) |
| Extremely severe | 1 (4.35) | 0 (0.00) |
Prevalence of depression, anxiety, and stress symptoms (DASS-21) in a sample of medical students (n = 49).
Percentages calculated based on group totals per subscale.
A one-way ANOVA for the PF&C test scores showed a small effect between groups (F1,47 = 3.584, p = 0.065, ω2 = 0.050) for g-factor performance (Figure 1). Consequently, PF&C scores were used as covariates in the analysis of the EF tests.
Figure 1
The rmANCOVA for reaction times (RT) revealed a small effect for block (F9,414 = 3.610, p < 0.001, ω2 = 0.020), suggesting adequate task learning by both groups (Figure 2). The analysis also showed a non-significant effect between groups (F1,46 = 1.249, p = 0.269, ω2 = 0.003), and the covariate (g-factor) was not significant (F1,46 = 0.116, p = 0.735, ω2 = 0.000). There was no interaction between group and block (F9,414 = 1.216, p = 0.283, ω2 = 0.002). Visual inspection of Figure 2 indicates that both groups performed with similar RTs on the task.
Figure 2
The rmANCOVA for error rate showed a non-significant effect for block (F9,414 = 1.577, p = 0.092, ω2 = 0.002). However, visual inspection indicates an initial reduction in errors from block 1 to block 2, followed by a performance plateau across the remaining blocks (Figure 3).
Figure 3
Although Figure 3 visually suggests a group difference, the rmANCOVA revealed a non-significant main effect for Group (F1,45 = 0.916, p = 0.344, ω2 = 0.000). Instead, results indicated that the covariate (g-factor) significantly accounted for performance variance, showing a small effect (F1,46 = 5.813, p = 0.020, ω2 = 0.049). Furthermore, no significant effects were observed for the interaction between Block and Group (F9,414 = 0.800, p = 0.616, ω2 = 0.000) or the interaction between Block and g-factor (F9,414 = 0.264, p = 0.984, ω2 = 0.000).
The rmANCOVA for RT on the Stroop test revealed statistically significant effects for Stroop interference (F1,46 = 12.373, p < 0.001, ω2 = 0.008) and for block (F9,414 = 2.735, p < 0.001, ω2 = 0.008), although both presented negligible effect sizes. Conversely, the main effect for Group was non-significant (F1,46 = 0.435, p = 0.513, ω2 = 0.000), nor was there a significant effect for the covariate (F1,46 < 0.001, p = 0.991, ω2 = 0.000). No other significant interaction effects were observed (Figure 4).
Figure 4
The rmANCOVA for error rate showed a non-significant main effect for Group (F1,46 = 0.036, p = 0.851, ω2 = 0.000). Consistent with previous analyses, results indicated that the covariate (g-factor) significantly accounted for performance variance, showing a small effect (F1,46 = 6.181, p = 0.017, ω2 = 0.052). No other significant interaction effects were observed, specifically the interaction between Block and Group (F9,414 = 1.594, p = 0.115, ω2 = 0.009) and the interaction between Block and g-factor (F9,414 = 0.746, p = 0.666, ω2 = 0.000) (Figure 5).
Figure 5
Results for the WCST revealed a statistically significant main effect for Group regarding the trials to complete the first category (H1 = 6.312, p = 0.012, ε2 = 0.131). Conversely, results showed non-significant effects for the number of categories completed (H1 = 0.256, p = 0.613, ε2 = 0.005) and for perseverative responses (H1 = 0.234, p = 0.629, ε2 = 0.005) (Figure 6).
Figure 6
4 Discussion
The present study aimed to measure Executive Functions (EF) in individuals with a history of COVID-19 infection, regardless of the timing or total vaccine doses received, and to compare their performance with subjects who had never been diagnosed with COVID-19. The sample predominantly consisted of individuals who were unvaccinated at the time of infection and had recovered almost two years prior to testing. For this specific population of undergraduate medical students, results suggest a robust recovery of executive domains. The COVID-19 group exhibited superior accuracy in working memory and inhibitory control tasks, alongside enhanced efficiency in the initial phase of the WCST, specifically regarding the trials to complete the first category. Crucially, regarding the tasks where covariate analysis was applicable, results revealed that performance was significantly accounted for by the participants' cognitive baseline (g-factor). This finding implies that the overall preservation of executive domains—likely including the efficiency observed in the WCST—is effectively sustained by high cognitive reserve rather than limited by the history of infection.
Throughout the pandemic, research efforts intensified to document the sequelae caused by the virus (Marshall, 2023; Thaweethai et al., 2023) and recent systematic reviews continue to highlight executive function, attention, and memory as the primary domains affected by post-acute sequelae of SARS-CoV-2 infection (Aderinto et al., 2025; Nasir et al., 2025; Panagea et al., 2025). Specifically, studies assessing patients within the first-year post-infection often report persistent deficits. For instance, Becker et al. (2023) identified executive dysfunction with the Trail Making Test B in a diverse cohort approximately 11 months after infection, while Trender et al. (2024), detected subtle deficits persisting up to one year. The discrepancy between these high rates of impairment and our null findings likely reflects the longer recovery period in our study and the specific high-functioning profile of our sample. Indeed, the trajectory of these deficits appears to be time-dependent. Buer et al. (2024), in a large population-based study (n > 8,000), observed that while the risk of reporting executive deficits peaks between 6 and 12 months, it shows signs of decline thereafter.
Our findings extend this timeline, suggesting that by twenty months, any initial executive deficits have likely resolved. Complementing this temporal perspective, severity also plays a crucial role. Serafim et al. (2024) approached executive deficits by categorizing 302 patients into three severity groups (mild, moderate, and severe) approximately 18 months post-diagnosis. While they utilized detailed neuropsychological assessments covering IQ, working memory, and processing speed, they found that persistent deficits at this stage were largely confined to severe patients who had required Intensive Care. Thus, our study, which analyzed the sample as a single group of predominantly mild cases with no hospitalization needed, using valid neuropsychological instruments, found that participants infected approximately twenty months prior did not perform worse than those who were never infected.
In addition, the absence of cognitive impairment in our sample is strongly corroborated by recent findings in similar demographics. Batra et al. (2025) assessed young medical students in India with a history of mild COVID-19 at least two years post-infection and, consistent with our results, found no significant differences in executive function, attention, or memory compared to controls. Similarly, Daher (2025) reported no decline in objective cognitive performance in a cohort evaluated one year after infection. These studies, alongside our own, suggest that young adults with high cognitive reserve - such as medical students - possess a resilience that buffers against long-term neurocognitive sequelae. In contrast, Zamarian et al. (2024) noted persistent attention deficits in 23% of patients at one year in an older cohort (median age 56), reinforcing the protective role of younger age and higher cognitive baseline. From an educational perspective, this suggests that the rigorous admission processes for medical schools may indirectly select individuals with higher neurocognitive resilience. Consequently, medical students and peers in other demanding professions appear better equipped to recruit compensatory neural networks to counteract potential viral-induced deficits. Crucially, this protective mechanism is not exclusive to this demographic but reflects a broader biological principle. A recent meta-analysis by Foreman et al. (2025) confirmed that high cognitive reserve significantly moderates post-COVID-19 outcomes, reducing the magnitude of cognitive deficits by approximately 33% across diverse populations. Thus, our findings in medical students likely represent the “best-case scenario” of this universal neuroprotective effect.
On the other hand, the preserved performance observed in our sample necessitates a precise distinction regarding the mechanism of adaptation. While the terms are sometimes conflated in the literature (Russo et al., 2012), it is important to distinguish between resistance - a passive imperviousness to stress where the system remains unchallenged (Fleshner et al., 2011) - and resilience, defined as an active, adaptive maintenance of function despite significant strain (Savulich et al., 2023). Our data strongly support a resilience framework. Unlike a resistance model, where performance is maintained because the stressor caused no perturbation, our subjects appear to be engaging in active compensation. Crucially, while we found no significant differences in perceived stress, anxiety, or depression (DASS-21), we argue that this absence of psychiatric morbidity does not imply a “cost-free” adaptation. Instead, it aligns with the neural efficiency hypothesis and the concept of high-effort coping (Stern, 2009). It is plausible that these high-functioning students recruited compensatory neural networks to maintain performance levels comparable to controls—incurring a biological cost that is metabolically demanding but psychologically silent in the short term. Therefore, while the stable DASS-21 scores confirm a preservation of affective well-being, the “hidden cost” of this resilience likely resides in reduced neural efficiency rather than overt clinical symptoms. Nevertheless, given the 20-month post-infection interval and the absence of emotional distress, it is reasonable to conclude that our population has achieved a robust functional recovery.
From a pedagogical perspective, recognizing this compensatory cost has direct implications for medical education. Since objective executive deficits do not appear to be the primary long-term threat for this high-reserve population, educational institutions should avoid automatically attributing subjective cognitive complaints to irreversible viral sequelae. Our data suggest that in these undergraduates, such difficulties likely reflect the transient impact of mental fatigue (Ishii et al., 2014), a byproduct of the prolonged compensatory effort described above, or academic burnout (Dunn et al., 2025), rather than structural neurocognitive deficits. Therefore, framing these complaints as permanent brain injury may be counterproductive. Instead, resources should be redirected from generalized cognitive remediation toward psychoeducational support systems. Curricular structures could operationalize this by incorporating regular screenings for mental fatigue, allowing institutions to intervene early with stress management programs to prevent the delayed exhaustion of these compensatory mechanisms. In addition, intervention programs that promote self-regulation have also demonstrated efficacy in mitigating these costs, particularly within the Brazilian context (Polydoro et al., 2015).
Finally, regarding the potential role of vaccination, a recent prospective study by Mukherjee et al. (2024) demonstrated that vaccination prior to SARS-CoV-2 infection did not significantly alter the subsequent neurologic manifestations of long COVID. This suggests that the preserved cognitive performance observed in our study is unlikely to be solely an artifact of vaccination status. Instead, it likely reflects a genuine recovery process facilitated by the passage of time (20 months) and the neuroprotective effects of high fluid intelligence (g-factor) in this specific population.
4.1 Strengths and limitations
The present study has notable strengths that contribute to the literature on post-COVID-19 cognition in educational settings. First, unlike investigations relying solely on self-reported cognitive complaints—which correlate strongly with psychological distress—we employed a battery of standardized, performance-based neuropsychological instruments administered in a controlled laboratory environment. Second, we rigorously controlled for current symptoms of depression, anxiety, and stress using the DASS-21, and for baseline cognitive differences using a non-verbal proxy measure for fluid intelligence (g-factor). Given the high prevalence of burnout and mental health issues among medical students, controlling for emotional distress was a methodological imperative to ensure that the null results were not confounded by psychiatric symptoms. Furthermore, by explicitly accounting for fluid intelligence, we were able to distinguish between genuine viral-induced deficits and variations in the participants' intellectual baseline.
However, limitations must be acknowledged. The relatively small sample size is a concern, as small samples generally increase the probability of Type II statistical errors (Shreffler and Huecker, 2023), although the negligible effect sizes observed suggest a genuine similarity in performance rather than insufficient power. A critical consideration involves the high cognitive reserve of medical students. Standard instruments may be subject to a ceiling effect in this population (Lezak et al., 2012), potentially lacking the sensitivity to detect subtle post-COVID cognitive nuances. The absence of a non-academic control group exacerbates this issue, making it difficult to fully disentangle the protective effects of cognitive reserve from psychometric limitations. However, to mitigate this - and to concurrently address the absence of formal socioeconomic status (SES) stratification - we utilized fluid intelligence (g) as a covariate Given the sample's inherent socioeconomic homogeneity, this unified analytical strategy inherently adjusted for baseline variations typically associated with SES, while also demonstrating that individual variations in processing capacity significantly predicted neurocognitive outcomes, suggesting that the protective mechanism is identifiable even at the upper end of the distribution. Furthermore, survivor bias cannot be ruled out: students with severe sequelae might have interrupted their training and were unavailable for recruitment. Additionally, without comprehensive serological tracking, we cannot definitively rule out that some control participants were asymptomatic carriers of SARS-CoV-2, though our strict criteria (negative tests or complete absence of symptoms) aimed to minimize this risk as effectively as possible within a naturalistic setting. Finally, the lack of longitudinal tracking limits causal inferences regarding the trajectory of recovery. Nevertheless, given widespread reports of initial executive deficits, it is reasonable to assume that many participants may have experienced transient executive deficits that had remitted by the time of our assessment.
5 Conclusion
In summary, while the literature documents executive dysfunction in the aftermath of COVID-19, the present study demonstrates a robust preservation of executive domains in medical students approximately twenty months post-infection. Contrary to the hypothesis of long-term deficits, the infected group exhibited superior accuracy in working memory and inhibitory control, as well as greater efficiency in initial concept formation. These results, however, were largely explained by the participants' cognitive baseline (fluid intelligence/g-factor) rather than by infection history or vaccination status.
These findings are particularly relevant for the educational context, suggesting that young adults with high cognitive reserve possess a resilience mechanism that effectively buffers against potential neurocognitive sequelae of mild COVID-19. Consequently, for this specific demographic, specialized long-term cognitive monitoring may not be necessary. Instead, educational institutions should prioritize resources towards mental health support services to address the high prevalence of burnout and anxiety, which appear to be more significant disruptors of academic performance in this post-pandemic era than direct viral cognitive sequelae. Future research should focus on whether this “neuroprotective shield” of high cognitive reserve persists in older populations or in the face of reinfections.
Statements
Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
Ethics statement
The studies involving humans were approved by the Research Ethics Committee of the University of Western São Paulo (CAAE: 62315722.9.0000.5515). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.
Author contributions
BD: Validation, Writing – original draft, Data curation, Methodology. CM: Writing – review & editing. AR: Writing – review & editing. CB: Data curation, Writing – original draft, Methodology. FR: Formal analysis, Writing – review & editing, Supervision, Conceptualization, Project administration.
Funding
The author(s) declared that financial support was not received for this work and/or its publication.
Acknowledgments
We are appreciated to all the students and their schools who participated in the study.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI was used solely for translation and linguistic refinement. The authors have reviewed and edited the final content and take full responsibility for the manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1
AderintoN.OlatunjiG.KokoriE.OgieuhiI. J.YusufI. A.EgbunuE.et al (2025). COVID-19 and cognitive impairment: a review of the emerging evidence. Discov. Ment. Health5, 56. 10.1007/s44192-025-00189-3
2
ApóstoloJ. L. A.MendesA. C.AzeredoZ. A. (2006). Adaptation to Portuguese of the depression, anxiety and stress scales (DASS). Rev. Lat. Am. Enfermagem14, 863–871. 10.1590/S0104-11692006000600006
3
ArdilaA.LahiriD. (2020). Executive dysfunction in COVID-19 patients. Diabetes Metab. Syndr. Clin. Res. Rev14, 1377–1378. 10.1016/j.dsx.2020.07.032
4
BatraV. A.UnarkatK. B.KaurM.AhluwaliaH.MannaS. (2025). Post-COVID condition does not alter cognitive functions in young adults: a cross-sectional study in north India. Cureus.17 (4), e82208. 10.7759/cureus.82208
5
BeckerJ. H.LinJ. J.TwumasiA.GoswamiR.CarnavaliF.StoneK.et al (2023). Greater executive dysfunction in patients post-COVID-19 compared to those not infected. Brain. Behav. Immun114, 111–117. 10.1016/j.bbi.2023.08.014
6
BuerS.HagenB. I.SøraasA.WhiteR. A.BøR.Ellingjord-DaleM.et al (2024). Executive deficits after SARS-CoV-2 infection: a cross-sectional population study. Brain Behav. Immun. Health41, 100857. 10.1016/j.bbih.2024.100857
7
Costas-CarreraA.Sánchez-RodríguezM. M.CañizaresS.OjedaA.Martín-VillalbaI.Primé-TousM.et al (2022). Neuropsychological functioning in post-ICU patients after severe COVID-19 infection: the role of cognitive reserve. Brain Behav. Immun. Health21, 100425. 10.1016/j.bbih.2022.100425
8
DaherJ.KoberssyZ.DurieuxJ. C.AtiehO.BaissaryJ.AbboudM.et al (2025). Cognitive function 1 year after COVID infection. Open Forum Infect. Dis. 12 (10), ofaf583. 10.1093/ofid/ofaf583
9
DamianoR. F.CarusoM. J. G.CincotoA. V.De Almeida RoccaC. C.De Pádua SerafimA.BacchiP.et al (2022). Post-COVID-19 psychiatric and cognitive morbidity: preliminary findings from a Brazilian cohort study. Gen. Hosp. Psychiatry75, 38–45. 10.1016/j.genhosppsych.2022.01.002
10
DeuterD.HenseK.KunkelK.VollmayrJ.SchachingerS.WendlC.et al (2024). SARS-CoV2 evokes structural brain changes resulting in declined executive function. PLoS One19, e0298837. 10.1371/journal.pone.0298837
11
DiamondA. (2013). Executive functions. Annu. Rev. Psychol64, 135–168. 10.1146/annurev-psych-113011-143750
12
DiasN. M.GomesC. M. A.ReppoldC. T.BastosA. C. M. F.PiresE. U.CarreiroL. R. R.et al (2015). Investigação da estrutura e composição das funções executivas: análise de modelos teóricos. Psicol. Teor. E Prática17, 140–152. Available online at:https://pepsic.bvsalud.org/pdf/ptp/v17n2/11.pdf
13
DouaudG.LeeS.Alfaro-AlmagroF.ArthoferC.WangC.McCarthyP.et al (2022). SARS-CoV-2 is associated with changes in brain structure in UK biobank. Nature604, 697–707. 10.1038/s41586-022-04569-5
14
DunnL.LudemanE.KayingoG.FelterC. (2025). Burnout and mental health in medical education. BMC Med. Educ25, 1604. 10.1186/s12909-025-08009-6
15
FleshnerM.MaierS. F.LyonsD. M.RaskindM. A. (2011). The neurobiology of the stress-resistant brain. Stress14, 498–502. 10.3109/10253890.2011.596865
16
ForemanL.ChildB.SaywellI.Collins-PrainoL.BaetuI. (2025). Cognitive reserve moderates the effect of COVID-19 on cognition: a systematic review and meta-analysis of individual participant data. Neurosci. Biobehav. Rev171, 106067. 10.1016/j.neubiorev.2025.106067
17
FriedmanN. P.MiyakeA. (2017). Unity and diversity of executive functions: individual differences as a window on cognitive structure. Cortex86, 186–204. 10.1016/j.cortex.2016.04.023
18
Goss-SampsonP. M. A. (2025). Statistical analysis in JASP a guide for students.
19
HaierR. (2017). The Neuroscience of Intelligence, Cambridge Fundamentals of Neuroscience in Psychology. Cambridge: Cambridge University Press. 10.1017/9781316105771
20
HamaidehS. H.Hamdan-MansourA. M. (2014). Psychological, cognitive, and personal variables that predict college academic achievement among health sciences students. Nurse Educ. Today34, 703–708. 10.1016/j.nedt.2013.09.010
21
HugonJ.MsikaE.-F.QueneauM.FaridK.PaquetC. (2022). Long COVID: cognitive complaints (brain fog) and dysfunction of the cingulate cortex. J. Neurol269, 44–46. 10.1007/s00415-021-10655-x
22
IshiiA.TanakaM.WatanabeY. (2014). Neural mechanisms of mental fatigue. Rev. Neurosci. 25 (4), 469–479. 10.1515/revneuro-2014-0028
23
JASP Team (2025). JASP (Version 0.95.4)1.
24
KangH. (2021). Sample size determination and power analysis using the G*power software. J. Educ. Eval. Health Prof18, 17. 10.3352/jeehp.2021.18.17
25
LezakM. D.HowiesonD. B.BiglerE. D.TranelD. (2012). Neuropsychological Assessment. 5th ed. Oxford, New York, Auckland, and Cape Town: Oxford University Press.
26
MarshallM. (2023). Long COVID: answers emerge on how many people get better. Nature619, 20. 10.1038/d41586-023-02121-7
27
MilesS.HowlettC. A.BerrymanC.NedeljkovicM.MoseleyG. L.PhillipouA. (2021). Considerations for using the Wisconsin card sorting test to assess cognitive flexibility. Behav. Res. Methods53, 2083–2091. 10.3758/s13428-021-01551-3
28
MukherjeeS.SingerT.VenkateshA.ChoudhuryN. A.Perez GiraldoG. S.JimenezM.et al (2024). Vaccination prior to SARS-CoV-2 infection does not affect the neurologic manifestations of long COVID. Brain Commun.7, fcae448. 10.1093/braincomms/fcae448
29
NantaisK. M.SchellenbergE. G. (1999). The mozart effect: an artifact of preference. Psychol. Sci10, 370–373. 10.1111/1467-9280.00170
30
NasirS. M.YahyaN.YapK. H.MananH. A. (2025). Executive function deficit in patients with long COVID syndrome: a systematic review. Heliyon11, e41987. 10.1016/j.heliyon.2025.e41987
31
OwenA. M.McMillanK. M.LairdA. R.BullmoreE. (2005). N-back working memory paradigm: a meta-analysis of normative functional neuroimaging studies. Hum. Brain Mapp25, 46–59. 10.1002/hbm.20131
32
PanageaE.MessinisL.PetriM. C.LiampasI.AnyfantisE.NasiosG.et al (2025). Neurocognitive impairment in long COVID: a systematic review. Arch. Clin. Neuropsychol40, 125–149. 10.1093/arclin/acae042
33
PeirceJ.GrayJ. R.SimpsonS.MacAskillM.HöchenbergerR.SogoH.et al (2019). Psychopy2: experiments in behavior made easy. Behav. Res. Methods51, 195–203. 10.3758/s13428-018-01193-y
34
PolydoroS. A. J.PelissoniA. M. S.Do CarmoM. C.EmilioE. R. V.DantasM. A.RosárioP. (2015). Promoção da autorregulação da aprendizagem na universidade: percepção do impacto de uma disciplina eletiva. Rev. Educ. PUC-Camp20, 201. 10.24220/2318-0870v20n3a2877
35
Ramos-GalarzaC.Acosta-RodasP.Bolaños-PasquelM.Lepe-MartínezN. (2019). The role of executive functions in academic performance and behaviour of university students. J. Appl. Res. High. Educ12, 444–455. 10.1108/JARHE-10-2018-0221
36
RebelloC. J.AxelrodC. L.ReynoldsC. F.GreenwayF. L.KirwanJ. P. (2022). Exercise as a moderator of persistent neuroendocrine symptoms of COVID-19. Exerc. Sport Sci. Rev50, 65–72. 10.1249/JES.0000000000000284
37
RussoS. J.MurroughJ. W.HanM.-H.CharneyD. S.NestlerE. J. (2012). Neurobiology of resilience. Nat. Neurosci15, 1475–1484. 10.1038/nn.3234
38
RutterH.WabnitzK.NambiarD.GardeA.BentonT. G.HeymannD. L.et al (2025). The lancet commission on improving population health post-COVID-19. Lancet407 (10525), 267–308. 10.1016/S0140-6736(25)02061-6
39
SalasN.EscobarJ.HuepeD. (2021). Two sides of the same coin: fluid intelligence and crystallized intelligence as cognitive reserve predictors of social cognition and executive functions among vulnerable elderly people. Front. Neurol12, 599378. 10.3389/fneur.2021.599378
40
SaucierJ.JoseC.BeroualZ.Al-QadiM.ChartrandS.LibertE.et al (2023). Cognitive inhibition deficit in long COVID-19: an exploratory study. Front. Neurol14, 1125574. 10.3389/fneur.2023.1125574
41
SavulichG.Ferry-BolderE.LimT. V.MakE.ErscheK. D. (2023). The ‘resilient brain’: challenging key characteristics associated with the concept of resilience. Psychol. Med53, 6933–6936. 10.1017/S0033291722003907
42
SerafimA. P.SaffiF.SoaresA. R. A.MoritaA. M.AssedM. M.De ToledoS.et al (2024). Cognitive performance of post-COVID patients in mild, moderate, and severe clinical situations. BMC. Psychol.12, 236. 10.1186/s40359-024-01740-7
43
ShrefflerJ.HueckerM. R. (2023). Type I and Type II Errors and Statistical Power.
44
SternY. (2009). Cognitive reserve⋆. Neuropsychologia47, 2015–2028. 10.1016/j.neuropsychologia.2009.03.004
45
StroopJ. R. (1935). Studies of interference in serial verbal reactions. J. Exp. Psychol18, 643–662. 10.1037/h0054651
46
ThaweethaiT.JolleyS. E.KarlsonE. W.LevitanE. B.LevyB.McComseyG. A.et al (2023). Development of a definition of postacute sequelae of SARS-CoV-2 infection. JAMA329, 1934. 10.1001/jama.2023.8823
47
TrenderW.HellyerP. J.KillingleyB.KalinovaM.MannA. J.CatchpoleA. P.et al (2024). Changes in memory and cognition during the SARS-CoV-2 human challenge study. eClinicalMedicine76, 102842. 10.1016/j.eclinm.2024.102842
48
WashburnD. A. (2016). The stroop effect at 80: the competition between stimulus control and cognitive control: competition for control. J. Exp. Anal. Behav105, 3–13. 10.1002/jeab.194
49
WassersteinR. L.SchirmA. L.LazarN. A. (2019). Moving to a world beyond “ p < 0.05.”. Am. Stat73, 1–19. 10.1080/00031305.2019.1583913
50
YapleZ. A.StevensW. D.ArsalidouM. (2019). Meta-analyses of the n-back working memory task: fMRI evidence of age-related changes in prefrontal cortex involvement across the adult lifespan. NeuroImage196, 16–31. 10.1016/j.neuroimage.2019.03.074
51
YoungstromE. A.GlutiingJ. J.WatkinsM. W. (2003). “Stanford-Binet Intelligence scale: fourth edition (SB4): evaluating the empirical bases for interpretations,” in Handbook of Psychological and Educational Assessment of Children: Intelligence, Aptitude, and Achievement, eds. ReynoldsC. R.KamphausW. (New York: The Guilford Press), 217–242.
52
ZamarianL.RassV.GoettfriedE.MayrV.CarboneF.KindlP.et al (2024). Objective and subjective cognitive outcomes one year after COVID-19. Ann. Clin. Transl. Neurol11, 2360–2371. 10.1002/acn3.52149
53
ZarazagaA. F.Gutiérrez VásquezJ.Pueyo RoyoV. (2019). Revisión de los principales test clínicos para evaluar la visión del color. Arch. Soc. Esp. Oftalmol94, 25–32. 10.1016/j.oftal.2018.08.006
Summary
Keywords
executive function, long-term effects, neurocognition, neuropsychological tests, post-acute COVID-19 syndrome, psychophysics
Citation
Duarte BFS, Madia CS, Rufino AS, Braghin CP and Rodrigues FV (2026) High cognitive reserve sustains neurocognitive resilience in medical education twenty months after COVID-19. Front. Educ. 11:1771044. doi: 10.3389/feduc.2026.1771044
Received
18 December 2025
Revised
27 February 2026
Accepted
09 March 2026
Published
08 April 2026
Volume
11 - 2026
Edited by
Jonathan Martínez-Líbano, Universidad Andrés Bello, Chile
Reviewed by
Joana Senger, Federal University of Rio Grande, Brazil
Almitra Vázquez-Moreno, Autonomous University of the State of Hidalgo, Mexico
Updates
Copyright
© 2026 Duarte, Madia, Rufino, Braghin and Rodrigues.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Felipe Viegas Rodrigues rodrigues.fv@gmail.com
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.