Introduction

In early 2020, the world was struck by the emergence of a novel beta-coronavirus, SARS-CoV-2, which caused the COVID-19 pandemic. The primary mode of transmission for this virus is through respiratory particles released during breathing, coughing, speaking, and sneezing. With the onset of the pandemic, the lives of individuals worldwide were dramatically altered in ways both direct and indirect. Not only did COVID-19 affect physical health, but it has also been associated with changes in cognitive functioning. Various studies have shown that COVID-19 patients experienced cognitive impairments ranging from mild to severe. A systematic analysis by1 revealed that 27% of participants reported subjective cognitive impairments, particularly in executive functions, including working memory. Moreover, Nouraeinejad2 highlighted that COVID-19 triggers neuroinflammatory processes in the hippocampus, reducing neurogenesis and contributing to neuronal degeneration. The hippocampus, a crucial structure for memory, spatial processing, and executive functions, is further impaired by cytokine storms and microglia activation. In addition, COVID-19 infection has been associated with atrophy in the frontal and parietal lobes, as well as the thalamus—key brain structures involved in working memory, attention, and executive functions3. Beaud et al.4 examined cognitive functions related to working memory, executive function, and attention, identifying deficits across these domains in COVID-19 patients. Additionally, these patients exhibited slower processing speeds and delayed verbal and visuospatial retrieval5. Persistent symptoms of COVID-19, known as “long COVID,” have also been associated with neuropsychological deficits, particularly in working memory, processing speed, and attention. Poletti et al.6 reported that 30–80% of individuals with long-term symptoms suffered from these cognitive impairments. Research by7 and2 focused on the neurobiological underpinnings of these impairments, including neuroinflammation and changes in brain structure. When investigating the understanding of working memory impairment, the focus was on aspects such as neuroinflammation and possible changes in brain structures7. The latter relate primarily to changes in parenchymal brain structures, which can lead to functional impairments of the brain. The systematic review by7 shows that working memory performance is influenced by various factors such as age, time and the severity of the infection (β = − 0.132, p < 0.001; β = 0.098, p < 0.001; β = 0.075, p = 0.003). Research by2 emphasises that COVID-19 triggers neuroinflammatory processes in the hippocampus that reduce neurogenesis and lead to neuronal degeneration. The hippocampus, which is central to memory, spatial processing and executive functions, is impaired in its function by microglial activation and the associated cytokine storms.

Although several studies report cognitive changes following COVID‑19, the evidence remains heterogeneous. It is therefore important to consider that psychological stress may also contribute to cognitive performance differences. In this study, we focus on working memory. Depressive symptoms have likewise been associated with impairments in working memory; for example, the meta‑analysis by8 found that individuals with major depressive disorder (MDD) showed longer reaction times and lower accuracy on n‑back tasks compared to healthy controls.

As shown in the meta-analysis by9, self-reported anxiety symptoms are associated with poorer working memory performance. These results were found for both complex and simple working memory tasks. Accordingly, it can be assumed that anxiety has an influence on the working memory performance of the participants in the COVIMMUNE-Clin study. As shown in10, COVID-related anxiety was associated with poorer working memory performance, even after controlling for state and trait anxiety, age, gender, and education.

The studies by Chai et al.20 showed that people with social anxiety disorder have lower activity in the dorsolateral prefrontal cortex. This led to lower n-back task performance in terms of accuracy and responsiveness compared to control participants11. It can be concluded from these research results that the sometimes permanently increased anxiety symptoms triggered by the COVID-19 pandemic could also change these brain regions and thus also reduce working memory performance.

Stress during the pandemic has been exacerbated by various factors, including isolation due to lockdown measures, social distancing, and disruptions to daily life. Studies show that these changes have led to significant increases in psychological stress. In 2021, 31% of respondents reported feeling more stressed compared to just 17% in 202012. Skoda et al.13 highlighted that despite a decrease in COVID-specific fears, psychological stress levels remained high due to long-term exposure to social and economic stressors. Psychological stress was further shown to correlate with cognitive performance changes, including issues such as insomnia and increased cortisol levels during the lockdown period14.

Despite the recognition of cognitive impairments resulting from COVID-19, relatively few studies have specifically examined the impact of psychological distress on visuospatial and auditory working memory. In particular, the comparative analysis of different working memory components has received little attention, highlighting the need for a more differentiated approach. This study aims to fill this gap by assessing a range of stress-related factors—such as psychological distress, somatic symptoms, and anxiety—and their impact on the working memory performance of COVID-19 patients.

Methods

All methods were carried out in accordance with relevant guidelines and regulations. The study protocol was reviewed and approved by the Ethics Committee. The study protocol was reviewed and approved by the local Institutional Review Board (Medical Ethics Committee of the University Hospital Bonn, ID 511/20) on March 10, 2021. Written informed consent was obtained from all participants prior to study participation. All procedures were conducted in accordance with the Declaration of Helsinki and relevant institutional and national guidelines.

Participants

A total of 150 individuals aged between 25 and 75 years participated in the study. Widmann et al.5 outlines the inclusion and exclusion criteria of the study. The first cohort (Cohort I) consists of patients who, after a SARS-CoV-2 infection, either had no symptoms (n = 52) or only symptoms of smell or taste disorders (anosmia, ageusia). The second cohort (Cohort II) includes patients who, after a severe course of a SARS-CoV-2 infection (n = 52), were admitted to a hospital (any type of ward) for at least 24 h. The third cohort (Cohort III) is a healthy control group (n = 23), with age and gender distributions that, based on frequency matching, resemble those of the other cohorts. A SARS-CoV-2 antibody rapid test is performed during the screening for the healthy control arm to exclude any recent or active infection. Of the 150 individuals initially recruited, only 127 could be included in the analyses, and by Visit 3 the sample was further reduced to 92 due to participant drop-outs. The Hopkins Verbal Learning Test is used as a screening method to exclude individuals with verbal episodic memory anomalies before assigning them to the healthy control group. A delayed verbal recall score that is more than − 1.0 SD below the age-specific reference norm value is the excusion criterion. The absence of memory problems, as well as the lack of a known history or current diagnosis of a psychiatric or neurological disorder, were additional criteria for assigning participants to the health control group.

Participants were recruited through various means. COVID-19 patients who had previously been treated at the University Hospital Bonn in February 2020 received a postal invitation to participate in the study. Another recruitment method involved the publication of the study on various social media channels, newspapers, and websites, with a request for interested individuals to contact a provided email address. A subsequent telephone screening was used to exclude or identify potential participants. Participants were screened according to predefined inclusion and exclusion criteria. General inclusion criteria required written informed consent, age between 25 and 75 years, fluent German language skills, and the ability and willingness to participate across all study visits. Cohort-specific criteria ensured that individuals were assigned to the asymptomatic/mild COVID group (ASY), the severe COVID group (SEV), or the healthy control group (CTL) based on clinical course and diagnostic status. Control participants additionally had to meet cognitive, psychiatric, and virological criteria (HVLT > − 1.0 SD, no history of psychiatric disorders, no substance misuse, and a negative SARS-CoV-2 antibody test at baseline). General exclusion criteria comprised any condition that clearly interfered with study participation or with the clinical or neuropsychological procedures, including inability to provide informed consent, sensory impairments that prevented or substantially compromised neuropsychological testing, contraindications for MRI, severe or unstable medical conditions, current severe depressive episodes, psychotic or bipolar disorders, current or past substance misuse, known neurodegenerative diseases (e.g., Alzheimer’s disease, Parkinson’s disease, frontotemporal dementia, Huntington’s disease, amyotrophic lateral sclerosis), vascular dementia, history of stroke, or a history of malignant disease. A detailed flow diagram summarizing participant inclusion, attrition, and exclusions across visits is provided in Fig. 1.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Participant flow diagram illustrating inclusion, exclusion, and attrition across Visit 1 and Visit 3. Note SEV = severe COVID burden; ASY = asymptomatic/mild COVID; CTL = control group. Attrition includes participants who did not return for Visit 3 or provided incomplete/invalid data.

Measurement of psychological distress

The ISR test (ICD-10 Symptom Rating) was developed to psychometrically assess the severity of psychological symptoms. The symptoms of mental illnesses listed in Chapter F of the ICD-10 form the basis of the tool. The ISR scale generally assesses psychological distress, as well as specific symptoms of anxiety, depression, obsessive–compulsive disorder, eating disorders, and somatic complaints. Additionally, there is an auxiliary scale that includes individual items as a screening function for certain syndromes. The items refer to the experiences of the respondents within the past two weeks and are assessed using a five-point Likert scale (0–4). The participant provides a self-assessment regarding 22 different items. The participants should spontaneously rate their agreement with the items by marking either the lowest level of the symptom (does not apply; number 0) or the highest level of the symptom (applies extremely; number 4). Depending on the application area, different average ranges are used. Clinical psychology, neuropsychology, and traffic psychology are fields where a broad average range of 16 to 84 is applied. Completing the test takes approximately 10 min. Depression, anxiety, obsessive–compulsive, somatization, and eating disorder symptoms, as well as auxiliary items, are calculated from the raw scores. The sum of these values results in a total scale. Cut-offs for the presence or severity of symptoms are available for both the scales and the auxiliary items. Norm values are provided for the syndrome scales and the total scale. Since this study focuses on anxiety and somatization symptoms, the evaluation will be explained in more detail below.

To identify the presence or to determine the severity of anxiety symptoms, the categorization of raw scores in Table 3 (see supplement) provides the appropriate normalization. In this scale, an above-average percentage indicates that the participant’s responses suggest the presence of anxiety symptoms, such as physical complaints in anxiety-provoking situations (e.g., palpitations, shortness of breath) and tendencies to avoid anxiety-triggering situations.

In this scale, an above-average percentile indicates that a test subject exhibits some somatization symptoms. For example, the participant might have indicated that they frequently worry about being physically ill, distrust medical judgments, or feel compelled to visit a doctor due to unexplained physical complaints. Based on the raw scores in Table 4 (see supplement), it can be determined whether somatization symptoms are present and how severe they are. In the analysis of the COVIMMUN-Clin study data, the values associated with severe symptom burden will be further examined using the existing norm scale and included in the analysis.

COVID burden was operationalized as the clinical severity of the acute infection, quantified by the number of days spent in hospital. Disease severity therefore reflects an objective clinical indicator of illness intensity. ‘COVID burden’ is used solely to denote the clinical severity classification (SEV, ASY, CTL). This variable reflects an objective clinical grouping and is not be interpreted as a subjective or continuous indicator of perceived disease burden. In contrast, psychological distress was assessed using the ISR (ICD-10 Symptom Rating) scale, which captures subjective symptom burden across multiple psychological domains. Throughout the manuscript, we use the term ‘COVID burden’ exclusively to refer to clinical disease severity (hospitalization duration), whereas ‘psychological distress’ refers to ISR-based symptom severity. These constructs represent distinct dimensions—clinical vs. psychological—and were treated as separate predictors in all analyses.

Measurement of working memory

The visuospatial memory was assessed using the CORSI Block Tapping Test (backward) as part of the Wiener Test System (tablet-based procedure, reference), and the verbal memory was assessed using the Digit Span (WMS-R) backward (reference). For the latter, the test administrator first reads a sequence of numbers, which the study participant is then required to repeat first forwards and then backwards. At the beginning, the Digit Spans are limited to only two to three digits, but they become progressively longer throughout the test. If the participant switches or omits two or more numbers in a sequence, the test is discontinued. The total number of correctly repeated Digit Spans serves as the result of working memory performance. The Reliable Digit Span (RDS), derived from the Wechsler Memory Scale, assesses working memory and effort validity by summing the longest digit spans recalled forwards and backwards. Scores ≤ 7 indicate insufficient effort (Schroeder et al., 2012). In this study, only individuals who achieved a minimum score of 8 at Visit 1 and 9 at Visit 3 were included, ensuring that all analyzed data reflected valid performance. The Block Tapping Test is a method used to determine visuospatial abilities. The test takes place on a board containing nine blocks, arranged irregularly. The participant can see the nine blocks, but not the numbers that are placed on the opposite side (visible to the test administrator). A sequence of blocks with a certain length is tapped by the experimenter three times. The participant is then required to tap the blocks in the same order (immediate block span), followed by tapping them in reverse order. Both Digit Span backward and CORSI backward primarily require executive manipulation in addition to storage components and are therefore interpreted as measures of executive working memory with verbal and visuospatial demands.

Study Procedure

The COVIMMUN-Clin Study is a monocentric, prospective longitudinal study. Data collection at the three assessment time points included: neurocognitive, neurological, and physical examinations; blood samples; SARS-CoV-2 rapid antibody tests; biomarker panels; lung function tests; and imaging procedures, such as the use of a magnetic resonance imaging (MRI) scanner. The data analysis for this study focuses on the results of the neurocognitive assessments, which included various neuropsychological tests and questionnaires. These were used to measure changes in lifestyle due to the pandemic, mental health, sleep behaviour, psychiatric symptoms, and daily activities. The data consists of both objective test results obtained through a computer-based assessment (Wiener Test System) and standardized and normed tests assessing the following domains: complex attention, verbal learning and memory, visual and spatial learning and memory, semantic language skills, and psychological measures for anxiety and depression disorders. Additionally, self-assessment questionnaires were used to measure the subjective change in cognitive abilities (comparing before and during the COVID-19 illness) (Widmann et al.5).

More specified hypothesis: This study examines how two distinct constructs—(1) COVID burden, operationalized as clinical severity group (SEV, ASY, CTL), and (2) psychological distress, measured via anxiety symptoms, psychosomatic complaints, and ISR scores—are associated with working memory performance across two time points. Prior research suggests that both the clinical severity of COVID‑19 and pandemic‑related psychological stress may be linked to cognitive functioning, although these constructs represent different dimensions of post‑COVID experience.

H1:

Working memory performance is expected to be associated with psychological distress (anxiety symptoms, psychosomatic complaints) and with COVID burden (clinical severity group), acknowledging these as conceptually distinct predictors.

H2:

COVID burden (clinical severity group) is expected to be directly associated with working memory performance.

H3:

Psychological distress is expected to correlate with COVID burden, reflecting the possibility that individuals with more severe illness experience higher levels of distress.

H4:

Anxiety symptoms are expected to be associated with working memory performance.

H5:

Psychosomatic complaints are expected to be associated with working memory performance.

H6:

Total psychological distress (ISR total score) is expected to be associated with working memory performance independently of clinical severity group. In summary, H1–H3 address general associations between clinical severity, psychological distress, and working memory, while H4–H6 focus on specific symptom dimensions. The hypotheses explicitly distinguish between COVID burden as a group‑based clinical indicator and psychological distress as a measured construct.

Data Analysis

The study examines the impact of independent variables (sample categorization, psychological stress, somatic and anxiety symptoms) on the dependent variable (working memory: executive working memory with verbal demands and executive working memory with visuospatial demands). For the statistical analysis, data from Visit 1 (study start) and Visit 3 (after twelve months) are used, as psychological stress was assessed at these time points using the Item-Symptom-Rating Scale (ISR). As this study is based on an existing longitudinal clinical cohort, an a priori power calculation was not feasible. To increase transparency, a post-hoc sensitivity analysis was conducted, indicating that the available sample size (N = 127) allows detection of small to-moderate effect sizes. Results should be interpreted within these constraints.

The Reliable Digit Span (RDS), derived from the Wechsler Memory Scale, assesses working memory and effort validity. It sums the longest digit spans recalled forwards and backwards. Scores ≤ 7 suggest insufficient effort (Schroeder et al., 2012). In this study, minimum scores of 8 (Visit 1) and 9 (Visit 3) indicate valid performance. Cut-off values were chosen conservatively to ensure performance validity, in line with established recommendations for Reliable Digit Span interpretation.Descriptive statistics, including measures of central tendency and dispersion (mean, median, variance, and standard deviation), were first calculated to characterize the sample and examine variable distributions. Subsequently, ANOVAs were conducted to compare group differences in memory performance and assess changes over time within a multilevel framework. To exclude potential multicollinearity, correlations between independent variables were examined. Multiple regression analyses were then used to identify predictors of working memory, though these do not capture longitudinal dynamics. Therefore, linear mixed models (LMMs) were employed to analyse changes in working memory over time and their associations with predictor variables. LMMs account for intra-individual dependencies by modelling random intercepts, making them well-suited for the longitudinal structure of this study. The Akaike Information Criterion (AIC) is used to assess model fit in linear mixed models by balancing goodness of fit and model complexity. It supports the selection of models that best explain the data without overfitting, favouring neither causal accuracy nor theoretical validity. Lower AIC values indicate better model fit relative to alternative models . Model selection was exploratory and guided by AIC comparisons rather than confirmatory hypothesis testing. Age, gender, and years of education were included as covariates in all mixed-effects models. These variables were selected a priori based on established theoretical and empirical evidence demonstrating their relevance for cognitive performance across working memory and verbal tasks. Prior research consistently shows that demographic factors—particularly educational attainment—exert substantial and independent effects on tasks such as backward Digit Span and verbal fluency. Including these covariates therefore serves to control for well-documented sources of variance and to reduce potential confounding when estimating the associations between psychological distress and cognitive outcomes.

Results

The sample includes 127 participants, of whom 52 had a severe COVID course or high COVID burden (SEV), 52 had a mild/asymptomatic course (ASY) or low COVID burden, and 23 had no COVID infection (CTL). Individuals who became infected with the virus during the study were excluded from the control group. The group consisted of 59 men and 68 women (see Table 1). The chi-square analysis revealed no significant differences between the groups in terms of gender distribution (χ2(2) = 0.761, p = 0.684). The longitudinal dataset comprised two measurement occasions (Visit 1 and Visit 3). At Visit 1, data were available for 127 participants (SEV: 52; ASY: 52; CTL: 23). At Visit 3, the sample was reduced to 92 participants due to attrition (SEV: 44; ASY: 38; CTL: 10). Consequently, not all individuals contributed data at both time points, resulting in an unbalanced longitudinal structure. The mixed-effects models were estimated using restricted maximum likelihood (REML), which incorporates all available observations under the assumption that data are missing at random (MAR).

Table 1 Frequencies and percentages of gender in the study groups (N = 127).
Full size table

The analyses revealed that participants in the SEV group reported significantly higher levels of psychological distress, anxiety symptoms, and somatic complaints at both measurement points compared to the ASY and CTL groups. These group differences were statistically robust, with the SEV group consistently showing the highest burden across all psychological domains. Anxiety symptoms were the most pronounced across all groups, followed by psychological distress and then somatic complaints.

In terms of cognitive performance, no significant group differences emerged at Visit 1 (see Table 2). Although the SEV group showed an improvement in auditory working memory across visits, they remained significantly impaired compared to both the ASY and CTL groups at Visit 3, with the ASY group performing best. Visual-spatial working memory (CORSI) scores did not differ significantly between groups at either time point, though the ASY group showed a numerical advantage at Visit 1.

Table 2 Means, standard errors, and ANOVA results for the SEV, ASY, and CTL groups at Visit 1 and Visit 3 (N = 127).
Full size table

Overall, psychological burden remained elevated in the SEV group over time, while cognitive differences became more apparent later in the study, particularly in tasks requiring auditory working memory.

Table 3 presents the results of a Kendall’s Tau correlation analysis at Visit 1. The ISR total score (psychological distress) was significantly positively correlated with both the anxiety and somatization scales. A moderate positive correlation was also observed between CORSI and Digit Span backward performance, reflecting consistency across working memory domains. Notably, higher psychological distress was associated with poorer visuospatial working memory (CORSI), whereas no significant relationship emerged between psychological distress and auditory working memory (Digit Span backward), suggesting domain-specific effects.

Table 3 Kendall’s Tau τ (Tau) correlation analysis of the variables at visit 1 (N = 127).
Full size table

Auditory working memory was positively associated with years of education, whereas CORSI performance showed no such link. Further correlations revealed that verbal fluency was also significantly associated with education (τ = 0.320, p < 0.001; see Supplement Table 20). However, no significant associations were found among psychological distress, auditory working memory, and verbal fluency at Visit 1 (see Supplement Table 19). At Visit 3, these variables remained unrelated (see Supplement Table 21). A positive correlation between education and Digit Span backward performance was confirmed (see Supplement Table 22).

Table 4 (Visit 3) again showed significant positive intercorrelations between psychological distress and its symptom subscales. CORSI and Digit Span backward performance remained positively related. Cognitive outcomes were correlated with each other, while psychological measures were internally consistent. Higher educational attainment continued to be associated with better performance on the Digit Span backward test (p = 0.009), whereas CORSI performance remained unaffected by education.

Table 4 Kendall’s Tau τ (Tau) correlation analysis of the variables at visit 3 (N = 127).
Full size table

The influence of psychological burden on working memory performance

Hypothesis 1:

Association between psychological distress and working memory performance

Visual spatial working memory

The models testing Hypothesis 1 showed that higher psychological distress was significantly associated with lower visual-spatial working memory performance in the Corsi task (see Tables 3 and 4). This association remained stable after adjusting for covariates. Model assumptions were met, supporting the robustness of the findings. Overall, individuals reporting higher psychological distress tended to show lower visual-spatial working memory performance across both time points. (Table 5).

Table 5 Summary table of fixed effects and AICs of all sub-models for Model 1: CORSI, ISR-Total, Group, Time (N = 127).
Full size table

In a mixed linear model, the influence of several variables on performance was analysed in the CORSI test. The ISR total raw score had a significant negative effect on visual-spatial working memory performance. Age showed a significant negative effect on performance in the CORSI test (see Table 6). Older participants tended to show poorer working memory performance. The time of the study, gender and grouping had no influence on CORSI performance in this case. The negative estimated value at the time of the study suggests a poorer cognitive performance of the participants at visit 1 compared to visit 3.

Hypothesis 1:

Association between psychological distress and auditory working memory.

Table 6 Parameter estimation of fixed effects: Model 4.
Full size table

Auditory working memory

To examine Hypothesis 1, we tested whether psychological distress was associated with auditory working memory performance. Several linear mixed models were calculated for the Digit Span task. Model 4 was selected as the most appropriate model based on the lowest AIC value (AIC = 877.87; see Table 7). This model was then used to evaluate the associations between psychological distress, covariates, and auditory working memory performance.

Table 7 Summary table of fixed effects and AICs of all submodels for Model 2: Digit Span backwards, ISR total, group, time point (N = 127).
Full size table

As neither the QQ plots nor the Shapiro–Wilk test for the residuals (p = 0.796) indicated significant deviations from normality, the assumption of normally distributed residuals for this model appears to be met. Additionally, the random effects showed no noticeable departures from normality, further reinforcing this assumption. These findings support the validity of the model estimates and confirm the appropriateness of the Restricted Maximum Likelihood (REML) model. Consequently, the model is considered stable, and its estimates can be deemed reliable.

In the linear model in Table 8, positive effects of years of education on auditory working memory performance were found. The more years of education a person had, the better they performed in the test of reciting Digit Spans backwards. Compared to CORSI performance, mental stress had no significant influence on performance in the cognitive test. None of the other parameters included in the analysis showed a statistically significant result. participants.

Table 8 Parameter estimation of fixed effects: Model 4.
Full size table

The influence of COVID burden (group membership) on working memory performance

Hypothesis 2:

Association between COVID burden (group membership) and working memory performance

Visual-spatial working memory

To examine Hypothesis 2, we tested whether COVID burden (clinical severity group) was associated with visual‑spatial working memory performance. Several linear mixed models were compared to identify the most suitable model. Model 4 was selected based on the lowest AIC value (AIC = 608.32; see Table 9). Model diagnostics indicated that assumptions were met: Q–Q plots and the Shapiro–Wilk test for residuals (p = 0.058) showed no significant deviations from normality, and the random effects displayed no irregularities. These results support the stability of the model estimates and the appropriateness of the REML approach.

Table 9 Summary table of fixed effects and AICs of all submodels for Model 3: CORSI, group, time (N = 127).
Full size table

The statistical calculations showed that age had a significant negative influence on the performance of visual-spatial working memory. Participants of older age performed worse in the CORSI block tapping task. All other parameters included in the analysis had no significant influence on the criterion (see Table 10). Differentiating the participants according to group and gender did not appear to have a significant effect on visual-spatial working memory performance. These values are consistent with the results of the confidence intervals. When looking at the difference between two mean values, zero is the value that indicates that there is no difference. This means that the difference is not significant if the confidence interval contains a zero. This was the case for all parameters except the age variable (Cumming & Finch, 2005).

Hypothesis 2:

Association between COVID burden (group membership) and working memory performance

Table 10 Parameter estimation of fixed effects: Model 4.
Full size table

Auditory working memory

To examine Hypothesis 2, we assessed whether COVID burden (clinical severity group) was associated with auditory working memory performance. Several linear mixed models were compared to identify the most suitable model. As shown in Table 15, Submodel 4 was selected for further analysis because it yielded the lowest AIC value (AIC = 878.54).

The QQ plots and the Shapiro–Wilk test for the residuals (p = 0.774) provided no evidence of significant deviations from normality, confirming that the model satisfies this assumption. Likewise, the random effects showed no notable deviations, reinforcing the normality assumption. These results validate the REML model and indicate that the estimates are both stable and reliable. (Table 11).

Table 11 Summary table of the fixed effects and AICs of all sub-models for Model 3: Digit Span, group, time (N = 127).
Full size table

The evaluation of the parameters on auditory working memory performance, which is shown in Table 12, shows that the time of the examination and the years of education have an influence on the cognitive test performance. At visit 1, the participants showed poorer test results than at visit 3. Working memory performance improved over the course of the study. Furthermore, the more years of education the participants had, the better their auditory memory performance. Similar findings can be seen in the correlation analyses and ANOVAs (see Tables 2, 3 and 4). All other variables showed no significant influence.

Hypothesis 3:

Association between psychological distress and COVID burden.

Table 12 Parameter estimation of fixed effects: Model 4.
Full size table

Relationship between psychological complaints and the COVID burden

To examine Hypothesis 3, we assessed the association between psychological distress and COVID burden (clinical severity group). Several linear mixed models were calculated, adding predictors stepwise to identify the most suitable model. As shown in Table 13, Model 1 was selected for further evaluation because it yielded the lowest AIC value (AIC = 262.96).

Table 13 Summary table of fixed effects and AICs of all sub-models for Model 4: ISR total raw score, group, time point (N = 127).
Full size table

. Although deviations from normality were observed, linear mixed-effects models are generally considered robust to moderate violations of this assumption. Nevertheless, these findings should be interpreted with caution and are considered exploratory.This violation of the normality assumption could affect the validity of the model estimates, particularly those of the random effects and residual variance.

Table 14 illustrates the estimation of the fixed parameters for assessing the influence of the study group and the time point on psychological stress. The time of the examination had a significantly positive effect on psychological stress. At visit 1, the participants had increased mental stress due to the positive estimator compared to the reference time of visit 3. In addition, there was a significant positive effect of the examination group on mental stress. People who had a higher COVID burden reported more psychological complaints.

Table 14 Parameter estimation of fixed effects: Model 1.
Full size table

Relationship between anxiety symptoms and COVID burden

Hypothesis 4:

Association between anxiety symptoms and COVID burden.

To examine Hypothesis 4, we assessed the association between anxiety symptoms and COVID burden (clinical severity group). Several linear mixed models were calculated, and predictors were added stepwise to identify the most suitable model. As shown in Table 15, Model 4 was selected for further analysis because it yielded the lowest AIC value (AIC = 513.90).

Table 15 Summary table of the fixed effects and AICs of all sub-models for Model 5: Anxiety scale, time point, group (N = 127).
Full size table

The results of the QQ plots demonstrated that neither the residuals nor the random effects were normally distributed. Despite observed departures from normality, linear mixed-effects models typically tolerate moderate assumption violations. Even so, the findings warrant cautious interpretation and should be viewed as exploratory. This deviation from normality could affect the reliability of the model estimates, especially concerning the random effects and residual variance.

Table 16 shows the estimation of the fixed parameters of the model for identifying influencing variables on anxiety symptoms. The time of the study as well as the categorisation of the test participants into the respective group had a significant influence on the anxiety symptoms. During visit 1, the test participants described more anxiety than during the third visit. The positive estimate suggests that the SEV group in particular reported more symptoms than the CTL group.

Table 16 Parameter estimation of fixed effects: Model 4.
Full size table

Relationship between somatic complaints and the COVID burden

Hypothesis 5:

Association between somatic complaints and COVID burden.

To examine Hypothesis 5, we assessed the association between somatic complaints and COVID burden (clinical severity group). Several linear mixed models were calculated, and predictors were added stepwise to identify the most suitable model. As shown in Table 17, Model 1 was selected for further analysis because it yielded the lowest AIC value (AIC = 479.65).

Table 17 Overview table of the fixed effects and AICs of all sub-models for Model 6: Somatic complaints, group, time point (N = 127).
Full size table

. Findings of QQ-Plots indicated significant deviations from normality in both cases. Although deviations from normality were detected, linear mixed-effects models are generally robust to moderate violations of this assumption. Nonetheless, the results should be interpreted with caution and regarded as exploratory.

The constant term is not significant in the estimation of the fixed parameters, which indicates that there is no significant baseline value for somatic complaints without the influence of the variable of the study group. As can be seen from Table 18, the SEV study group showed significant differences with regard to the occurrence of somatic complaints. Participants in this group reported more somatic complaints than the CTL group. However, the ASY group did not differ from the control participants, who served as the reference population in this context. All other parameters included in the analysis were not significant.

Table 18 Parameter estimation of fixed effects: Model 1.
Full size table

Relationship between psychological distress and working memory performance

Hypothesis 6:

Association between psychological distress and working memory performance.

The results provide partial support for Hypothesis 6, as the association between psychological distress and working memory performance differed across the two assessment points. At Visit 1, psychological distress (ISR total raw score) was significantly associated with visuospatial working memory performance in the Corsi task. In contrast, no significant association was observed for auditory working memory performance (Digit Span backward). Taken together, these findings indicate that Hypothesis 6 is only partially supported.

Data Availability

The datasets generated and/or analysed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Conclusion

Finally, the key results of the statistical analysis are explained. The gender distribution showed no differences between the groups. The SEV group showed higher values for psychological distress, anxiety symptoms and somatic complaints at both time points. Anxiety was more pronounced than general psychological distress or somatic complaints in all groups and at both time points. In addition, the SEV group showed poorer cognitive performance than the ASY and CTL groups. The ASY group achieved better results than the CTL group in the CORSI test at visit 1, while control participants performed better in other cognitive tests. At visit 1, mental discomfort correlated negatively with CORSI performance, but not with the Digit Span backwards. The latter showed a positive correlation with education at both visits.

At visit 3, there were no correlations between mental stress and cognitive performance. Mixed linear models showed that age and mental stress were associated with lower visual-spatial memory performance, while auditory working memory performance was influenced by education but not by mental stress. Individuals with more years of education performed better on the Digit Span task. Participantsinfluence of COVID burden (group membership) on working memory performance was then investigated. Initially, age was found to have a significant negative influence on cognitive visual-spatial performance, but the group classification was not statistically significant. When analysing auditory working memory performance, COVID exposure was also not a statistically significant factor. In contrast, the years of education parameter was identified as a significant positive factor and the time of the survey as a statistically significant negative factor. At the first survey time, the test participants performed worse in the test—Digit Span backwards—than at visit 3. Furthermore, the relationship between COVID burden (group membership) and mental complaints was analysed. The time of the examination and the group were identified as significant positive factors for mental complaints. Participants who described more psychological complaints were also more affected by a severe course of COVID. The investigation of anxiety in connection with the COVID burden showed that the group assignment and the time of the survey also had a significant positive relationship with the anxiety symptoms. The participants reported more anxiety at the earlier time of the survey. The SEV group reported more anxiety than the CTL group. The SEV study group showed considerable differences with regard to the occurrence of somatic complaints. The participants in this group reported more somatic complaints compared to the CTL group.

Overall, the results show that COVID burden (group membership) and psychological complaints influence cognitive and mental health. The findings are categorised and compared with existing research in the following discussion section.

Discussion

It is important to note that COVID burden in the present study reflects illness severity based on group classification rather than directly assessed perceived COVID-related stress. Therefore, findings should be interpreted in this conceptual framework. Our data show that higher levels of COVID-19 burden among participants were associated with higher psychological distress, as assessed by ISR total raw scores. When a stressor exceeds the perceived resources available to cope, psychological distress may result15. The SEV group, already physically burdened by severe COVID-19, likely had limited resources to manage additional stressors, thereby explaining their increased psychological distress. Compared to the reference value at Visit 3, participants had higher ISR scores during the first visit. Over time, habituation to the extraordinary circumstances of the pandemic likely occurred, as people adapted to the new lifestyle. Bendau et al.16 similarly demonstrated that anxiety symptoms tend to decrease over the course of epidemics and pandemics. Furthermore, we found a significant correlation between anxiety symptoms and COVID-19 burden. These findings align with the meta-analysis by17, demonstrating that COVID-19 patients with severe cases more frequently reported elevated anxiety levels. Furthermore, our mixed linear model calculations revealed a significant positive effect of the timing of assessments on anxiety symptoms. Table 16 shows the estimation of the fixed parameters of the model for identifying influencing variables on anxiety symptoms. At visit 1, the participants described more anxiety than at the third visit (B = 0.152; p = 0.045). Finally, the SEV group differed significantly from the reference group in terms of psychosomatic complaints. As can be seen from Table 18, the SEV study group showed significant differences with regard to the occurrence of somatic complaints. Participants in this group reported more somatic complaints than the CTL group (B = 0.711; p < 0.001). However, the ASY group did not differ from the control participants, who served as the reference population in this context (B = 0.278; p 0.104).

In light of the exploratory nature of the present analyses and the fact that several model assumptions were not fully satisfied, the interpretation of the findings must be approached with particular caution. Specifically, the assumption of normally distributed residuals was violated in the models examining psychological distress, anxiety, and somatic complaints. Although mixed-effects models are generally considered robust to moderate deviations from normality, these violations nonetheless constrain the strength and generalizability of the inferences that can be drawn. Consequently, the observed patterns should be regarded as preliminary and indicative rather than as definitive effects. The conclusions presented here therefore emphasize emerging trends and potential associations, while refraining from strong claims regarding underlying mechanisms or causal interpretations.

The focus of our study was to examine whether working memory performance in COVID-19 patients was associated with psychological distress. At Visit 1, psychological distress showed a significant association with performance in the Corsi task. (Another finding is that aged individuals performed worse in the CORSI task. The negative influence of age on visual working memory performance has been reported elsewhere and may be attributed to degenerative processes in the hippocampus (Raz et al., 2005).) Studies by18 show that mental stress can impair the executive control of the prefrontal cortex, especially in complex visual-spatial tasks. Our study directly compared the effects of psychological distress on both visuospatial and auditory working memory, providing a more differentiated perspective on cognitive impairments in COVID-19 patients. Furthermore, our longitudinal design allowed us to track changes over time, indicating that the association of psychological distress on visual working memory decreases as distress levels decrease.

Importantly, no significant association between psychological distress and visual working memory performance was observed at Visit 3. This could be attributed to the reduction in psychological stress over the investigation period, as shown by the decrease in ISR total raw scores between Visits 1 and 3.

Mental stress decreased significantly in all three study groups during the course of the study. A reduction in mental stress could have had a positive effect on cognitive functions, so that no correlation between mental stress and visual-spatial working memory performance could be identified at the third visit, which can be seen in Table 8 (τ = -0.043; p > 0.05).

Bendau et al.16 similarly reported a decline in mental distress, such as anxiety, over the course of the pandemic. This aligns with data from10—the authors demonstrated that COVID-related anxiety was associated with poorer performance on working memory tasks, even after controlling factors such as state and trait anxiety, age, gender, and education. The influence of anxiety on working memory has been reported elsewhere. For instance, patients with generalized anxiety disorder or social anxiety disorder exhibit poorer working memory performance19. As indicated in the meta-analysis by9, self-reported anxiety symptoms are associated with poorer working memory performance, both in complex and simple working memory tasks. Similarly, research by20 showed that people with social anxiety disorder exhibit lower activity in the dorsolateral prefrontal cortex, an important area for working memory. Accordingly, participants performed worse in the n-back task11.

Auditory memory, specifically the ability to recall number sequences in reverse order, was unaffected by psychological distress. However, participants with higher levels of education showed better test results. Based on these results, it can be inferred that auditory working memory may be less susceptible to performance declines due to stress. Grenfell-Essam et al.21 demonstrated that visually presented information is more vulnerable to stress-induced disruptions than auditory information. This may help explain why psychological distress (e.g., stress or anxiety) is more strongly associated with visual-spatial memory performance, given that such symptoms can coincide with increased demands on cognitive control processes. . Regarding the influence of education years, the cognitive reserve model can be used as an explanatory framework. This concept refers to the brain’s ability to compensate for impairments, which can be influenced by factors such as years of education or lifestyle. These reserves allow individuals to approach tasks in alternative ways and prevent performance limitations. In this context, the positive effect of education on the backward Digit Span task can be explained. Similar results were observed regarding verbal fluency, measured by the Wiener Object Naming Test (WOBT). Verbal fluency correlated with educational years but not with psychological distress. Therefore, education, independent of acute stressors, can be considered a crucial factor in maintaining cognitive abilities. Across multiple models, the clinical severity groups (SEV, ASY, CTL) did not emerge as significant predictors of working memory performance. Consequently, any apparent group-level differences should be regarded as descriptive rather than statistically supported effects.

Limitations

Despite the comprehensive and careful planning of the COVIMMUN-Clin study, certain limitations must be considered when interpreting the results. A major drawback is that the ISR (Item-Symptom Rating Scale) was not administered during the second visit. This data gap could influence the study’s findings in several areas. Without ISR data for Visit 2, it is challenging to develop precise longitudinal models that capture the nuanced trajectory of psychological distress. Consequently, it remains unclear whether the difference between Visits 1 and 3 represents a linear change or whether intermediate values might have predicted a more complex progression, such as temporary increases or decrases. Due to the observational longitudinal design and the absence of pre-pandemic baseline data, causal inferences cannot be drawn. All findings should therefore be interpreted as associational. Another concern is the heterogeneity within the SEV group, which includes patients with both short and long hospital stays as well as those requiring intensive care. This results in symptom variability related to illness. The study did not account for individual treatments received in the intensive care unit. The lack of pre-pandemic baseline data raises another issue. It is unclear whether the observed changes in the COVID-19 cohort are directly attributable to the infection. Pre-existing memory issues in COVID-19 patients cannot be ruled out. The study aimed, in part, to assess cognitive deficits using self-report questionnaires, asking participants to document cognitive changes perceived since the acute phase of COVID-19. To enable comparisons with healthy participants, the CTL group was also surveyed about potential changes in their cognitive abilities. This approach allows the study to identify cognitive changes in healthy participants that are unrelated to a COVID-19 infection. Methodological factors may also play a disadvantageous role. For example, learning effects could occur as the same tests were used at different time points. Participants may become familiar with the tasks through repeated administration, which could influence the results. Lastly, the risk of infection among healthy participants must be acknowledged. Prior to statistical analysis, supposedly healthy participants who contracted COVID-19 during the study were excluded. Infection status was determined using blood tests (IgG and IgM antibodies) and self-reports. However, due to the subjective nature of participant reports, reinfections during the study cannot be ruled out. When interpreting group differences over time, it is important to consider that the control group at Visit 3 was relatively small. This limits the robustness of longitudinal group comparisons and reduces the precision of estimates for the control group at the final assessment point. Therefore, group-based changes across visits should be interpreted with appropriate caution. The interpretation of the findings is limited by multiplicity, as conducting multiple statistical tests increases the risk of false-positive results. Given the large number of statistical tests conducted in this study (including ANOVAs, correlation analyses, and multiple mixed-effects models), the issue of multiplicity warrants consideration. No formal correction for multiple comparisons (e.g., Bonferroni or false discovery rate adjustments) was applied, as the analyses were primarily exploratory in nature and aimed at identifying potential patterns rather than testing a narrowly defined set of confirmatory hypotheses. We therefore report unadjusted p-values and interpret the results with appropriate caution, emphasizing effect patterns and consistency across models When interpreting the observed improvements in working memory performance from Visit 1 to Visit 3, it is important to consider potential practice or learning effects resulting from repeated testing. Tasks such as the Corsi Block-Tapping Test and Digit Span are known to show performance gains simply through increased familiarity with the test structure. Therefore, part of the observed improvement may reflect procedural learning rather than genuine cognitive change. This possibility should be taken into account when interpreting longitudinal effects.

Outlook

The results of this study highlight several points to be addressed in future research to close existing gaps and deepen the understanding of COVID-19’s impact on cognitive and mental health.

Firstly, the psychological burden within the SEV group should be examined in greater detail. This could help identify risk factors that exacerbate psychological distress and enable the development of targeted prevention programs. Additionally, a more comprehensive investigation of all psychological health characteristics in COVID-19 patients over an extended period is necessary. This would allow for the assessment of long-term psychological trajectories and the implementation of tailored intervention programs. The effectiveness of specific rehabilitation programs aimed at reducing psychological distress in COVID-19 patients should also be explored. Future studies could evaluate these programs to establish evidence-based treatment strategies for individuals affected by COVID-19. Due to their high psychological burden, including symptoms of anxiety, the SEV group requires intensive psychological support. Stress-reduction techniques, self-help groups, or guided psychotherapeutic group sessions could be implemented in public facilities. Interacting with peers could help participants exchange strategies for managing mental health challenges and foster mutual support.

Given the pronounced negative relationship between of psychological distress on visuospatial but not auditory working memory, studies should delve deeper into these differences on a neurobiological, structural, and functional level. Investigations could address why visuospatial working memory is particularly vulnerable to psychological changes and how it can be protected proactively. Focusing on the SEV group, memory strategies such as mnemonic techniques or the use of external memory aids could facilitate daily life and compensate for cognitive deficits. These methods could be tailored to individuals and integrated into rehabilitation programs to ease everyday challenges. In an interdisciplinary rehabilitation framework, strategies for coping with daily life could be developed to alleviate physical and mental complaints.

Longitudinal studies are essential to understanding the long-term cognitive effects of COVID-19. It remains unclear which cognitive impairments are transient, and which may become chronic. Multi-year studies could identify factors that promote complete recovery. Moreover, cognitive rehabilitation programs targeting improvements in visuospatial working memory among COVID-19 patients should be a starting point for future research.

Given the high prevalence of anxiety symptoms across all study groups, programs to reduce anxiety should be prioritized in treatment plans. Cognitive-behavioural therapy, breathing techniques, or progressive muscle relaxation could help alleviate anxiety symptoms. Online programs or apps teaching stress-reduction techniques could empower individuals to better manage uncertainties and stressors in the future.

For the future, institutions must adopt a transparent approach to infection-reduction measures, ensuring their necessity is clear to all demographics. Offering a sense of control to affected individuals could help reduce anxiety. When people can understand and exert a degree of control over a situation, it can diminish fear and foster a sense of self-efficacy. Expanding psychosocial counselling services and implementing universal programs for the prevention of mental health issues should also be a priority.

Conclusion

This study aimed to examine the association between psychological distress and working memory in individuals with varying levels of COVID-19 burden over a 12-month period. . This concluding chapter recaps the study, summarizes the main findings, and reflects on them in the context of the research question.

The data for this research was drawn from the COVIMMUN-Clin study, which investigated the long-term cognitive and pulmonary effects of a COVID-19 infection. This longitudinal study included three data collection points, with the analyses in this thesis focusing on the results from Visit 1 (baseline) and Visit 3 (follow-up after 12 months). The evaluation concentrated on auditory memory performance (assessed with the backward Digit Span test) and visuospatial components (measured using the CORSI block-tapping test). Psychological distress, anxiety-specific symptoms, and somatic complaints were assessed using the Item Symptom Rating Scale (ISR). Groups were categorized based on the severity of COVID-19 illness, with control participants who contracted the virus during the study being excluded.

The analysis of mixed linear models yielded three central findings:

  1. 1.

    Psychological burden in the SEV group

    Participants in the SEV group—those who had experienced a severe course of COVID-19—reported significantly higher levels of psychological distress compared to the ASY and CTL groups. This increased burden was evident across multiple domains, including anxiety symptoms, somatic complaints, and general psychological distress, and remained consistently elevated at both assessment points (Visit 1 and Visit 3).

  2. 2.

    Anxiety as the most prominent symptom

    Across all three groups, anxiety symptoms emerged as the most pronounced dimension of psychological distress, surpassing both somatic complaints and general psychological stress scores. This pattern was stable over time, indicating that anxiety represented a particularly persistent psychological response in the aftermath of COVID-19, regardless of disease severity.

  3. 3.

    Domain-specific effects of psychological distress on working memory

    Psychological distress—as measured by the total score on the ISR—was found to negatively influence visuospatial working memory performance, as assessed by the CORSI Block-Tapping Test. In contrast, auditory working memory, assessed via the Digit Span backward task, showed no such association. This suggests a domain-specific effect of psychological distress, with visuospatial memory being more susceptible to the impact of mental burden than auditory memory.

Educational attainment and age were significant covariates influencing cognitive performance:

More years of education correlated with better auditory and visuospatial working memory performance. Furthermore, older participants performed worse on the CORSI block-tapping test, reflecting age-related neurodegenerative processes.

These findings align with existing literature that highlights the link between mental distress and COVID-19 burden and suggests that psychological distress can directly or indirectly alter working memory performance. Specifically, visuospatial abilities appear to be more sensitive to mental distress, while auditory abilities remain stable. This vulnerability of visuospatial performance has also been noted in other studies. The concept of cognitive reserve may explain why individuals with more years of education achieved better results in cognitive tests.

The study emphasizes the importance of rehabilitation programs for COVID-19 patients. Such programs could reduce psychological distress and potentially have a moderating effect on cognitive performance. Preventive measures aimed at reducing mental health issues are equally critical. These interventions, particularly those targeting anxiety symptoms, could indirectly protect visuospatial performance from distress-related declines.

Future research should explore the impact of psychological distress on other cognitive parameters in COVID-19 patients. Longitudinal studies over several years are needed to identify late-emerging tertiary effects. Monitoring the trajectory of anxiety symptoms is also crucial to implement timely interventions and foster a sense of security and self-efficacy among the public during challenging times.