Abstract
The risk factors for post-COVID-19 cognitive impairment have been poorly described. This study aimed to identify the sociodemographic, clinical, and lifestyle characteristics that characterize a group of post-COVID-19 condition (PCC) participants with neuropsychological impairment. The study sample included 426 participants with PCC who underwent a neurobehavioral evaluation. We selected seven mental speed processing and executive function variables to obtain a data-driven partition. Clustering algorithms were applied, including K-means, bisecting K-means, and Gaussian mixture models. Different machine learning algorithms were then used to obtain a classifier able to separate the two clusters according to the demographic, clinical, emotional, and lifestyle variables, including logistic regression with least absolute shrinkage and selection operator (LASSO) (L1) and Ridge (L2) regularization, support vector machines (linear/quadratic/radial basis function kernels), and decision tree ensembles (random forest/gradient boosting trees). All clustering quality measures were in agreement in detecting only two clusters in the data based solely on cognitive performance. A model with four variables (cognitive reserve, depressive symptoms, obesity, and change in work situation) obtained with logistic regression with LASSO regularization was able to classify between good and poor cognitive performers with an accuracy and a weighted averaged precision of 72%, a recall of 73%, and an area under the curve of 0.72. PCC individuals with a lower cognitive reserve, more depressive symptoms, obesity, and a change in employment status were at greater risk for poor performance on tasks requiring mental processing speed and executive function. Study registration: www.ClinicalTrials.gov, identifier NCT05307575.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
Post-COVID-19 condition (PCC) develops in people previously infected with the SARS-CoV-2 virus, approximately 3 months after infection. These individuals exhibit symptoms that persist for no less than 2 months and cannot be accounted for by another diagnosis [1]. Typical symptoms include, though are not exhaustively limited to, fatigue, difficulty breathing, and cognitive impairment, which tend to adversely affect daily life activities.
The prevalence of PCC varies according to the study cohort, the methodology, the SARS-CoV-2 variant, and the vaccination status [2]. Nevertheless, according to a prevalence study that accounted for confounding variables, approximately one in eight individuals infected with COVID-19 could develop PCC [3]. Several mechanisms have been suggested to explain the pathophysiology of PCC, including organ damage occurring during the acute phase of infection, a chronic inflammation state, persistent SARS-CoV-2 viral antigens, and reactivation of latent herpesviruses [4, 5].
Risk factors for PCC are heterogeneous and remain poorly characterized. In a recent meta-analysis, Tsampasian et al. [6] reported that female gender, advanced age, and obesity were associated with an increased risk of developing PCC. Other comorbidities, such as anxiety and/or depression, asthma, chronic obstructive pulmonary disease, diabetes, ischemic heart disease, and immunosuppression, were also substantially associated with an increased risk of PCC. In addition, patients who required hospitalization or those in intensive care during the acute phase of COVID-19 infection had a risk of developing PCC that was more than double that of those who did not [6]. In contrast, a healthy lifestyle before infection is associated with a substantially lower risk of PCC, according to a prospective cohort study involving 32,249 women [7]. The risk of presenting persistent neurological symptoms (fatigue, headache, altered smell/taste, and cognitive complaints) was found to be increased in women, those of Hispanic ethnicity, low socioeconomic status, and hospitalization for COVID-19 [8].
The concept of post-COVID cognitive dysfunction has emerged as a potential form of cognitive impairment that occurs at least 3 months after acute infection with COVID-19. This condition is characterized by a distinct cognitive profile, primarily marked by deficits in attention and processing speed. In some cases, these deficits may be accompanied by impairments in executive function and episodic memory [9,10,11]. Mental processing speed and executive function are highly important for mental and physical health and in achieving success in academic and personal goals [12]. Mental processing speed refers to an individual’s capacity to process information efficiently and rapidly, and it is highly correlated with higher order cognitive functions [13]. Executive function regulates behavior and other cognitive processes, including attention and memory [14]. Alterations in executive function and speed processing can have significant adverse effects on daily functioning [15, 16].
Several factors are associated with post-COVID cognitive dysfunction. The severity of COVID-19 has been thoroughly investigated, yielding mixed results [17]. Further risk factors for cognitive impairment include age [9, 18, 19], level of education [9, 20,21,22], sex [19, 23], and the presence of comorbidities, particularly obesity [19]. On the other hand, cognitive reserve (CR) is the most significant protective factor for post-COVID cognitive impairment [24,25,26,27]. However, the existing body of research on CR has primarily focused on post-COVID patients who have undergone treatment in the intensive care unit (ICU) [24,25,26,27] or during the subacute phase [24, 26].
Machine learning techniques utilized across diverse fields of health can be categorized into unsupervised and supervised learning. Unsupervised learning refers to a learning algorithm that does not rely on labeled data and aims to partition data points into homogeneous groups to find a dataset’s natural structure and underlying patterns using a measure of similarity [28]. Several studies have categorized patients into groups, and distinct forms of cognitive dysfunction have been identified solely based on the patient’s data through unsupervised clustering techniques [29,30,31,32]. Supervised learning is used to train classification and prediction models using examples or outputs given a set of labels or a continuous response [33]. The utilization of supervised machine learning algorithms for disease prediction has become increasingly popular [34]. Cognitive and neuroimaging data have proven valuable in predicting several neurological and psychiatric conditions [35,36,37,38].
Unsupervised machine learning clustering techniques have been tested to identify post-COVID-19 cognitive phenotypes. Although data from five different cognitive domains were used, only two clusters emerged that differed in cognitive impairment according to the study of Matias-Guiu et al. [9]. In the present study, we focus on tests that measure mental processing speed and executive function, which we had found to be altered in our previous research [39]. We propose that an unsupervised partition of participants based on cognitive performance could be obtained using clustering algorithms. Then using a new set of variables and a supervised learning algorithm, this partition can be characterized by analyzing the characteristics of different groups and the relevance of the selected variables in the group separation.
The aim of this study was to identify the sociodemographic, clinical, and lifestyle characteristics that characterize a group of PCC participants with neuropsychological impairment.
Materials and methods
Participants
We conducted an observational, exploratory multicenter study of patients enrolled from neuropsychology and COVID-19 units across 19 hospitals in Catalonia, Madrid, Canarias (Spain), and Andorra, coordinated by the Consorci Sanitari de Terrassa (Barcelona, Spain) (ClinicalTrials.gov ID NCT05307575). We included preliminary data on 426 participants with PCC recruited from June 2021 through to July 2023. The mean age of the participants was 50.29 years (SD = 9.5), the mean level of formal education was 13.94 years (SD = 3.33), and 285 of them (70%) were female. Participants with PCC were classified into three groups according to the WHO clinical progression scale [40]: severe-intensive care unit (ICU) (6 to 9 points), hospitalized (4 and 5 points), and mild (2 and 3). According to the severity of the disease, 182 (42.7%) participants were hospitalized, of which 96 (22.5% of total sample) were admitted to the intensive care unit (ICU). The remaining 244 (57.3%) individuals with PCC were outpatients and had a mild illness at home. The mean time since onset to assessment was 362 (SD = 229) days. The age, sex, level of education, and severity of COVID-19 did not differ substantially among participants recruited at each center; however, the duration between the onset of illness and the time of assessment exhibited greater heterogeneity (please refer to Supplementary Table 1).
The inclusion criteria for the PCC group were as follows: (a) confirmed diagnosis of COVID-19 according to World Health Organization criteria, with signs and symptoms of the disease during the acute phase; (b) a post-infection period of at least 12 weeks; and (c) aged between 18 and 65 years. The exclusion criteria were as follows: (a) established diagnosis of psychiatric, neurological, or neurodevelopmental disorders, or systemic pathologies known to cause cognitive deficits, prior to COVID-19 infection; and (b) motor or sensory alterations that could impede neuropsychological examinations. In addition, patients who scored < 14 points on the Montreal Cognitive Assessment (MoCA) [41] as a general cognitive screening tool and/or < 85 points on the Word Accentuation Test (TAP) [42] as an estimate of premorbid IQ were excluded. All participants were native Spanish speakers.
Participant anonymity and confidentiality were guaranteed. The study was conducted with the approval of the Drug Research Ethics Committee (CEIm) of Consorci Sanitari de Terrassa (CEIm code: 02–20-107–070) and the Ethics Committee of the University of Barcelona (IRB00003099). The investigation followed the latest version of the Declaration of Helsinki.
Procedure
Before participation, written informed consent was obtained from all participants. Participation in the study was voluntary, and all participants followed a standardized protocol to achieve their participation requirements.
In the first session, we collected data on sociodemographic characteristics: age, sex, education, income, and change in employment status. The variable change in employment status was collected dichotomously through responses to the following question: Has your current employment situation changed compared to before the illness? We also asked about the employment situation at the time of the evaluation (disability, medical leave, change of job, etc.). In addition, we gathered data on the professions they held before the onset of the illness and classified them into occupational categories following the International Standard for Occupational Classification [43]. We also obtained clinical information: previous comorbidities, COVID-19 symptoms (how they experienced COVID-19), and post-COVID-19 symptoms. For each of 29 typical acute COVID-19 symptoms, the current presence/absence was detailed. Also, we asked participants to report any other symptoms they had been experiencing and had not been covered in the interview. A post-symptom score was calculated by adding the number of persistent symptoms at the time of evaluation and dividing by the total number of possible symptoms. Emotional symptoms were excluded to avoid overlap with the variables collected in the mental health questionnaires. Participants were then scheduled for a second visit in which the neuropsychological examination was performed, as described in Ariza et al. [39]. For this study, we used the following neuropsychological tests: parts A and B of the Trail Making Test (TMT) [44] (visual scanning, motor speed and attention, and mental flexibility); the Controlled Oral Word Association Test (COWAT) [45] (verbal fluency); Digit Symbol Coding Test (WAIS-III) [46] (visual scanning, tracking, and motor speed), and the Stroop Color and Word Test (SCWT) [47] that measured processing speed (SCWT word and color) and cognitive inhibitory control (SCWT word-color). These instruments are recommended for evaluating post-COVID cognitive impairment [48], and prior research has demonstrated their sensitivity to changes in executive function and mental processing speed [17, 49, 50]. All evaluations were performed by trained neuropsychologists.
Finally, participants were provided with all the questionnaires to complete online or on paper to assess different variables. Anxiety symptoms were assessed with the Generalized Anxiety Disorder (GAD-7) [51], depressive symptoms with the Patient Health Questionnaire (PHQ-9) [52], posttraumatic stress symptoms with the Posttraumatic Stress Disorder Checklist for DSM-5 (PCL-5) [53], and fatigue with the Chalder Fatigue Questionnaire (CFQ) [54]. The International Physical Activity Questionnaire (IPAQ) [55] was used to assess physical activity levels, the UCLA Loneliness Scale for loneliness [56], and the Cognitive Reserve Questionnaire (CRC) [57] for CR.
Statistical and machine learning analyses
Categorical variables were presented as numbers with percentages and continuous variables were presented as means and standard deviations. Differences between clusters were determined using the independent Student’s t test and Pearson's Chi-squared test, as appropriate. Odds ratios (ORs) for logistic regression were reported. Analyses were conducted using Python language with the following libraries: Pandas, Scikit-learn, Statsmodels, and SciPy.
The flow chart for data analysis is shown in Fig. 1. The first analysis was performed on a subset of the collected data. We focused on tests that measured processing speed and executive function, cognitive domains in which we had previously observed differences between individuals with PCC and controls without COVID-19 [17, 39]. Seven neuropsychological variables were selected to obtain a partition of the data according to their joint distribution.
Before data analysis, standard preprocessing of the data was applied. The missing values preprocess was applied and those participants with missing values were removed from the analysis. Outlier participants were detected using the local outlier factor (LOF) algorithm [58]. Data were standardized before the analysis and partitioning was performed using partitional clustering. Different clustering algorithms were applied to the data, including K-means, bisecting K-means, and Gaussian mixture models. The number of clusters was assessed using three cluster validity measures (Calinski–Harabasz, Davis–Bouldin, and Silhouette index). Principal component analysis (PCA) was also used to visually assess the validity of the clusters. Logistic regression was computed using the cluster partition as labels to assess the separability of the clusters. Data were divided into training and test sets (70% and 30%), and logistic regression was computed using least absolute shrinkage and selection operator (LASSO) regularization.
For the second analysis, the sociodemographical, clinical, and lifestyle variables were selected. Similar preprocessing of the data was applied using missing values imputation when possible. Data were split into the training and test sets (80% and 20%). Different machine learning algorithms were used to obtain a classifier able to separate the two clusters according to the selected variables, including logistic regression with LASSO (L1) and Ridge (L2) regularization, support vector machines (linear/quadratic/ radial basis function (RBF) kernels), and decision tree ensembles (random forest/gradient boosting trees). Hyper-parameters of the models were adjusted using Bayesian optimization search and tenfold cross-validation. The final model was selected according to accuracy, recall, precision, and the area under receiver operating characteristic curve (AUC). Recursive backward elimination was applied, which reduced the variables of the best model according to the order of the significance using the z-test, and keeping variables with p < 0.05. Interpretation of the final model was performed by computing the Shapley values using the SHAP algorithm. This is a well-established interpretation procedure for machine learning methods. Shapley values [59] is a method for assessing variables’ relevance in a model based on the coalitional game theory. It determines the fair payout that corresponds to the participant according to their contribution to the result. In the case of a machine learning model, the value for each variable corresponds to the average contribution of the variable to the classification of the examples in the dataset by the model.
Results
Sociodemographic and clinical characteristics
Four-hundred five individuals were employed before acquiring the illness. “Accounting, administrative, and other office employees” (22.7%) comprised the largest occupational category, while “technicians and scientific and intellectual health and teaching professionals” (17.9%) ranked second. “Elementary occupations: unskilled workers in services” and “directors and managers” followed with comparable percentages (9.8% and 9.1%, respectively). In total, 188 PCC individuals (44.7%) underwent employment status changes due to the disease. No significant variations were observed in the proportion based on occupational classification. However, it should be noted that the category “drivers and mobile-plant operators” showed the highest rate of job change (82%), and “qualified workers in agriculture, forestry, and fishing industry” showed the lowest rate of (20%), although both comprised a small number of workers (see Supplemental Table 2 and 3).
Table 1 shows symptoms reported by participants with PCC at the time of assessment. The most frequently reported general symptoms following COVID-19 were fatigue, headache, and pain. On the other hand, the neuropsychiatric symptoms that were reported most frequently were cognitive complaints and manifestations of depression and anxiety.
Clustering analyses
For the first analysis, after applying preprocessing to the data, the number of participants included in the dataset to be analyzed was 378. Exploration of the range of the possible number of clusters (2–10) with all the selected clustering algorithms resulted in an agreement on the detection of only two clusters in the data. The obtained partitions were similar; the one obtained with K-means was chosen as it was more natural for computing centroids and obtaining an interpretation of the clusters.
The results were also assessed visually by applying PCA to the data; the first two principal components (Fig. 2) corresponded to 69.7% of the total variance of the data. In the figure, two overlapping data clouds can be observed with different distributions. According to the variables’ centroids, we labeled the clusters as good performance (GP) and bad performance (BP). Table 2 shows the means and standard deviations of raw scores of neuropsychological variables corresponding to each cluster. Figure 3 shows a cluster boxplot for each standardized neuropsychological variable.
As the variables did not show non-linear dependencies and the groups were clearly visible using PCA, we applied logistic regression analysis. The accuracy was 99%, with a weighted average 99% recall, 99% precision, and an AUC of 1.0. The weights assigned to each one of the attributes are shown in Fig. 4. The obtained logistic regression was finally used to label the original dataset and to define the partition for further study.
Classification analyses
For the second analysis, a new dataset with sociodemographical, clinical, and lifestyle variables was defined (Table 2). The original dataset was reduced to 343 participants. The means and standard deviations of the two clusters using the selected variables are shown in Table 3.
All the different machine learning models that were trained had similar results; hence, we choose logistic regression with LASSO regularization for simplicity and interpretability, since it obtains a sparse solution that discards attributes that are not predictive. The accuracy of the model on the test dataset was 74% with a weighted average precision of 74%, recall of 74%, and an AUC of 0.74 (CI 95%: 0.63–0.84). The regularization discarded eight of the chosen variables (coefficients with zero value) and only four remaining variables had coefficients that were significative (p < 0.05 according to a z-test for the coefficients). The discarded variables were eliminated, and a recursive feature elimination procedure was applied, eliminating variables with less significant coefficients until all the remaining variables were significant (p < 0.05). The final model had four of the original variables with an accuracy of 72%, a weighted average precision of 72%, a recall of 73% and an AUC of 0.72 (CI 95%: 0.62–0.83) on the test dataset. This is not a perfect model, but it can be considered reasonably good, explaining a large amount of the behavior of the groups. Table 4 presents the ORs of the variables selected in the logistic regression model.
CRC Cognitive Reserve Questionnaire, PHQ-9 Patient Health Questionnaire
We applied PCA to the dataset reduced to the four selected variables. The representation of the first two components, which account for 58.6% of the variance, shows a clear separation among the examples from the two classes (Fig. 5). A first interpretation of the logistic regression model can be obtained from the model weights in Fig. 6.
Further interpretation was performed by computing the Shapley values with the SHAP algorithm. Figure 7 represents the mean Shapley values computed on the test set for each variable in the model with corresponding confidence intervals (95%) computed by bootstrapping. This represents the mean effect in absolute value of each variable, that is, the expected change from the mean prediction (probability 0.5) for each variable. This assigns to the CRC total score variable the largest effect from all the selected variables, followed by PHP-9, Obesity and Change in employment status all with a similar importance.
A detailed analysis of the Shapley values for individual examples in the test data according to the magnitude of the variables is presented in Fig. 8. A higher CRC total score has a positive effect, moving examples toward the GP class, the effect seems distributed equally in both directions depending on the magnitude of its value, except for an extreme case. For the obesity variable, the impact of being obese has a larger effect on the BP class than not being obese. A similar trend can be observed for the change in employment status variable, although smaller. For the PHQ-9 total score variable, larger values lean the classification to the BP class, but the distribution of the effect is uneven, with higher values having more impact than lower values.
A more detailed analysis of the relevance of the variables was performed by partitioning the dataset according to the sex and COVID severity variables, as represented in Figs. 9 and 10. There is no significant difference in the importance of the variables disaggregating the decisions by sex. For the severity disaggregation, there is a larger difference on the CRC total score and obesity variables. For the most severe cases (hospital and ICU), the CRC variable was more important. The obesity variable has greater importance for ICU cases.
Discussion
The aim of this study was to identify the sociodemographic, clinical, and lifestyle characteristics that characterize a group of PCC participants with neuropsychological impairment. We focused on the performance of mental processing and executive dysfunction because alteration of these cognitive functions has been widely demonstrated in objective assessments [9, 50, 60,61,62], and in our previous research [17, 39].
Although cognitive complaints are one of the most common persistent symptoms of PCC [63], not all individuals confirm cognitive impairment. In many cases, other symptoms predominate. Matias-Guiu et al. [9] previously described the clustering of their sample based on the severity of cognitive impairment rather than cognitive phenotypes. The results of our cluster analysis revealed the presence of two distinct groups based on their different performance levels.
Based on the standardized results, it is evident that our BP group would fall within the cognitive impairment category. This classification is supported by the fact that all the tests, except Trail making Test (TMT) A, exhibit scores that are 1–1.5 standard deviations below (over in the case of TMT B) those of the GP group [9, 50]. Nevertheless, based on the PCA visualization, there is some degree of overlap between the groups. Consequently, while logistic regression can establish a boundary between the two classes, the separation may be rather ambiguous. Notwithstanding the oversimplification inherent in assuming linear separation between the two groups, the main trends in the relationship between variables should be valid. In addition, it was noted that all cognitive variables were of comparable importance in delineating these groups.
Our study revealed that individuals diagnosed with PCC who exhibit lower CR, more depressive symptoms, obesity, and changes in employment status were more at risk for experiencing BP in tasks related to mental speed processing and executive function.
CR was the variable with the most weight in predicting performance in mental processing speed and executive function. CR is a theoretical concept that refers to the adaptability and flexibility of cognitive processes and explains why cognitive abilities or functions are susceptible to deterioration with age, disease, or brain injury [64]. Variations in CR can be attributed to the characteristics of functional brain processes, which are shaped by the interaction between innate and environmental influences throughout life. As a theoretical construct, CR has often been assessed using lifestyle proxies, such as educational level, employment position, or participation in intellectually stimulating activities. Nevertheless, it is important to note that each proxy variable has a distinct impact on CR [64, 65], indicating that the approximations used do not fully encompass the entirety of the CR. One potential approach for addressing this issue involves the utilization of a comprehensive measure that considers several parameters, such as the CR questionnaire employed in our study [57].
CR is a protective factor in susceptibility to dementia [66, 67] and cognitive impairment resulting from acquired brain damage [68,69,70] or mental illnesses [71, 72]; a higher CR indicates a more robust capacity to cope with brain injury [73]. Our findings align with previous studies showing that CR is also the most significant protective factor for post-COVID cognitive impairment [24,25,26,27]. These studies have been performed with patients who required ICU care during COVID-19 infection. Navarra-Ventura’s study [24] found that a higher level of CR was positively correlated with improved working memory performance, and Fernandez-Gonzalo et al. [26] found that the cognitively normal cluster had a higher CR based on educational attainment and estimated IQ, 1 month after ICU discharge. In addition, Costas-Carrera et al. [27] reported that a higher CR was related to a reduced risk of developing verbal memory or executive function impairment 6 months after discharge, and CR emerged as a protective factor against objective cognitive deficit 12 months after ICU discharge according to the study by Godoy-González et al. [25]. Other authors have found similar results using the CR-proxy education attainment in participants during the short-time following ICU discharge [21]. Studies that included individuals of all severities found that a higher education level had a protective effect against cognitive decline 6 months after COVID-19 infection [20, 22] and in long-COVID individuals [9, 19]. The participants in our study underwent evaluation, on average, 11 months after contracting COVID-19. Those individuals within the BP cluster presented with a significantly lower CR than those in the GP cluster. It should be noted that their educational attainment and estimated IQ scores were also comparatively lower, so they were removed from the regression variables.
It is worth highlighting the difference in the contribution of CR segregated by severity of COVID-19 to the model. According to the data, CR was more important for those with PCC who were hospitalized (ICU and hospitalized) than those with mild COVID-19. This is unsurprising since brain insult in the acute phase is likely greater in more severe cases. Brain injury during COVID-19 has multiple causes, including direct virus damage to the brain vasculature (SARS-CoV2 can enter the brain through the blood–brain barrier damaged by endothelial injury or cytokines and chemokines generated during infection) [74, 75], vasodilation, edema, and even cerebral ischemia produced by severe hypoxia secondary to systemic inflammation [76] or coagulopathy [77], and severe hypoxia from respiratory failure or sepsis, which is frequently observed in critical patients [78].
The second factor that helped to shape our model was depressive symptoms. Post-COVID patients exhibit elevated levels of depression [79, 80]. Even though some studies have found only weak [61, 62] or nonexistent relationships [81], most of the previous research has demonstrated a clear connection between depressive symptomatology and cognitive impairment [39, 82,83,84,85,86]. Reported depressive symptoms are the variable that gave the best predictive performance in attention and information speed processing [83, 86] and executive function [83, 85].
The causal relationship between depression and cognitive impairment is currently unknown. People with severe depressive disorder frequently exhibit cognitive impairments [87], particularly in executive dysfunction [88]. As previous studies [83, 85, 86] and the present study have shown, depressive symptoms predict mental processing speed and executive function in people with PCC. However, several characteristic markers of people with PCC, such as increased cortisol levels or increased pro-inflammatory factors [5], are associated with depression [89]. Furthermore, the implication of chronic inflammation on the development of cognitive impairment seems increasingly plausible [90]. According to inflammatory models of neurodegeneration, systemic inflammation increases neuro-inflammation, which in turn affects cognition [91]. Consequently, the same inflammatory process that causes post-COVID syndrome is responsible for cognitive dysfunction and depressive symptoms.
Finally, it cannot be ruled out that post-COVID symptoms, such as cognitive challenges, may cause depression. More research is needed to draw accurate conclusions regarding the connection between depression and cognitive deficits.
In our model, at body mass index (BMI) cutoff of 30 kg/m2, that is, being obese or not, was also a significant factor in classifying patients with PCC according to their mental processing speed and executive function performance. There is a significant connection between obesity and the increased severity and mortality rate among COVID-19 patients [92]. Furthermore, neuropsychological studies have linked obesity with cognitive impairments; obesity increases the likelihood of developing mild cognitive impairment [93]. Executive function has been shown to be reduced in obese subjects [94], and greater visceral adiposity has been related to slower mental processing speed [95]. Neuroimaging studies reported altered connections across brain regions, structural abnormalities, and task-related prefrontal cortex dysfunction [96]. Hypoactivation of the prefrontal cortex has also been observed during executive function performance [97]. In addition, obese participants exhibited altered functional connectivity strength in the salience network (putamen nucleus), which is associated with a slower mental processing speed [98].
It is currently unclear how obesity affects cognition; however, it is thought to be through inflammatory mechanisms. Obesity is characterized by chronic systemic inflammation, which involves the aberrant production of cytokines, heightened levels of acute-phase reactants, and activation of a network of inflammatory signaling pathways [99]. Research suggests that low-grade systemic inflammation can mediate BMI and executive function deficits [100,101,102].
Obesity increases the likelihood of developing PCC [6]. Recent evidence indicates that adipose tissue appears to be one of the primary sources of increased levels of chemokines in PCC individuals [102] and that chronic inflammation may be an etiopathogenic factor of PCC (5). Our results also show that obesity is a risk factor for post-COVID cognitive impairment. That risk is probably increased by the additive effect of chronic inflammation to an inflammatory disease, such as COVID-19 [103]. As expected, the contribution of obesity to the severity segregation pattern of COVID-19 was more important in ICU cases, where the inflammatory processes are more severe.
Finally, change in employment status was also a significant predictor of poorer cognitive performance. Employment status is part of the social determinants of health, which are factors related to the conditions around an individual’s birth and living situations that influence their overall health [104]. An individual’s socioeconomic status is widely recognized as a significant determinant of their overall health. Individuals with lower socioeconomic status are more susceptible to experiencing premature mortality, developing disability, and dementia [105, 106]. It is also associated with good brain health and better aging. In a population sample of men without dementia, those with socioeconomic advantages demonstrated superior psychomotor function, visual learning, and general cognition [107].
Unemployment has been reported to be associated with COVID-19 mortality in Chilean [108] and North American [109] cohorts. Lower socioeconomic status is associated with a higher risk of developing persistent neurological symptoms [8]. Our results show that a change in employment status is important in classifying people with PCC based on their performance on tests assessing the processing speed and executive function. Furthermore, this change in the employment situation affected all labor groups, not just those with lower qualifications. People in intellectually demanding occupations experienced difficulties in continuing a satisfactory performance, became less productive, and felt obliged to resign or accept lower-skilled positions. People who are unable to return to work full-time usually undergo disease exacerbations, which necessitate an extended period of medical leave. These individuals’ incapability to restore themselves to their prior level of activity carries significant economic and social consequences.
In contrast to previous research [9, 22, 110, 111], our model revealed no effect of age or gender. In fact, the evidence for the direction of the relationship of age and gender with respect to cognitive impairment is inconclusive. Some authors reported that the risk of cognitive impairment increased with age [22, 111], whereas others found that the risk was greatest for the youngest individuals [9, 112]. Regarding gender, it has been determined that both males [22] and females [19] are at risk for cognitive impairment. In our participants, the duration between the onset of the disease and the evaluation was highly variable. However, in this case, the two clusters for this variable were comparable, meaning it had no explanatory value. Conversely, the cross-sectional design limits the optimal implementation of this analysis. A further investigation should be conducted into the influence of time on the recovery process from cognitive impairment following COVID-19.
Our model explains a substantial portion of the variance in cognitive performance between groups; however, it is not ideal. Structural and functional brain abnormalities in people with post-COVID cognitive complaints, likely of neuroinflammatory origin, have been linked to alterations in attention, processing speed, and working memory after 1 year [113]. Conversely, chronic inflammation, persistent SARS-CoV-2 viral antigens, and reactivation of inactive herpesviruses have been proposed as factors that contribute to the persistence of symptoms [5]. A further investigation should be conducted incorporating plasma markers of inflammation and structural and functional neuroimaging data, to acquire a more precise profile of the factors predicting cognitive impairment following COVID-19.
Our study’s limitations and strengths must be considered when forming conclusions. First, the cross-sectional nature of our study design prevented us from determining causal relationships with precision. Our partition cannot ensure that the low-performance cluster is entirely made up of people with psychometric impaired cognition, because the clusters overlap. We lost several cases that had to be excluded from the analyses due to missing values that could not be imputed. Finally, although this study was conducted across multiple centers, all of them maintained a consistent cultural environment. Replication in other samples should be performed to confirm the generalizability of these findings.
However, we recruited a large sample size, including patients from several health centers, representing the full spectrum of severity of COVID-19. In addition, the selection of the sample was performed by ruling out comorbidities that could cause cognitive impairment. Participants underwent a comprehensive assessment with multiple levels, including clinical, neuropsychological, emotional, and lifestyle dimensions.
To conclude, despite methodological limitations, we found that PCC individuals with a lower CR, greater depressive symptoms, obesity, and a change in employment status were at greater risk for poor performance on tasks requiring mental processing speed and executive function. This information allows for the identification of at-risk individuals and the development and implementation of early intervention programs. Some of these factors are modifiable. Promoting a healthy lifestyle, avoiding a sedentary lifestyle, and increasing physical activity, as well as adequate nutrition, engaging in novel and stimulating activities, and cognitive remediation are ways to increase CR, emphasizing the neuroplasticity-influencing factors in participants with PCC.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
Soriano JB, Murthy S, Marshall JC et al (2022) A clinical case definition of post-COVID-19 condition by a Delphi consensus. Lancet Infect Dis 22:e102–e107. https://doi.org/10.1016/S1473-3099(21)00703-9/ATTACHMENT/EF4FD06B-88FA-4A0C-B837-DCFEE13E82D7/MMC1.PDF
ECDC (2022) Prevalence of post COVID-19 condition symptoms: a systematic review and meta-analysis of cohort study data, stratified by recruitment setting. Europe Centre For Disease Prevention And Control
Ballering AV, van Zon SKR, olde Hartman TC, Rosmalen JGM (2022) Persistence of somatic symptoms after COVID-19 in the Netherlands: an observational cohort study. The Lancet 400:452–461. https://doi.org/10.1016/S0140-6736(22)01214-4
Proal AD, VanElzakker MB (2021) Long COVID or Post-acute sequelae of COVID-19 (PASC): an overview of biological factors that may contribute to persistent symptoms. Front Microbiol. https://doi.org/10.3389/FMICB.2021.698169
Klein J, Wood J, Jaycox J et al (2023) Distinguishing features of Long COVID identified through immune profiling. Nature. https://doi.org/10.1038/s41586-023-06651-y
Tsampasian V, Elghazaly H, Chattopadhyay R et al (2023) Risk factors associated with post-COVID-19 condition a systematic review and meta-analysis. JAMA Intern Med 183:566–580. https://doi.org/10.1001/jamainternmed.2023.0750
Wang S, Li Y, Yue Y et al (2023) Adherence to healthy lifestyle prior to infection and risk of post-COVID-19 condition. JAMA Intern Med 183:232–241. https://doi.org/10.1001/jamainternmed.2022.6555
Ch’en PY, Gold LS, Lu Q et al (2023) Exploring risk factors for persistent neurocognitive sequelae after hospitalization for COVID-19. Ann Clin Transl Neurol. https://doi.org/10.1002/acn3.51801
Matias-Guiu JA, Herrera E, González-Nosti M et al (2023) Development of criteria for cognitive dysfunction in post-COVID syndrome: the IC-CoDi-COVID approach. Psychiatry Res. https://doi.org/10.1016/j.psychres.2022.115006
Ceban F, Ling S, Lui LMW et al (2022) Fatigue and cognitive impairment in Post-COVID-19 Syndrome: a systematic review and meta-analysis. Brain Behav Immun 101:93–135. https://doi.org/10.1016/J.BBI.2021.12.020
Quan M, Wang X, Gong M et al (2023) Post-COVID cognitive dysfunction: current status and research recommendations for high risk population. Lancet Reg Health West Pac 38:100836. https://doi.org/10.1016/j.lanwpc.2023.100836
Diamond A (2013) Executive functions. Annu Rev Psychol 64:135–168. https://doi.org/10.1146/annurev-psych-113011-143750
Kail R, Salthouse TA (1994) Processing speed as a mental capacity. Acta Psychol (Amst) 86:199–225. https://doi.org/10.1016/0001-6918(94)90003-5
Stuss DT (2011) Functions of the frontal lobes: relation to executive functions. J Int Neuropsychol Soc 17:759–765. https://doi.org/10.1017/S1355617711000695
Chung CS, Pollock A, Campbell T et al (2010) Cognitive rehabilitation for executive dysfunction in patients with stroke or other adult non-progressive acquired brain damage. Cochrane Database Syst Rev. https://doi.org/10.1002/14651858.cd008391
Vance DE, Heaton K, Fazeli PL, Ackerman ML (2010) Aging, speed of processing training, and everyday functioning: Implications for practice and research. Act Adapt Aging 34:276–291. https://doi.org/10.1080/01924788.2010.523867
Ariza M, Cano N, Segura B et al (2023) COVID-19 severity is related to poor executive function in people with post-COVID conditions. J Neurol. https://doi.org/10.1007/s00415-023-11587-4
PHOSP-COVID Collaborative Group (2022) Clinical characteristics with inflammation profiling of long COVID and association with 1-year recovery following hospitalisation in the UK: a prospective observational study. Lancet Respir Med. https://doi.org/10.1016/S2213-2600(22)00127-8/ATTACHMENT/7D899F10-EAF8-4CB5-AFB5-49B44810EC9B/MMC1.PDF
Miskowiak KW, Pedersen JK, Gunnarsson DV et al (2023) Cognitive impairments among patients in a long-COVID clinic: Prevalence, pattern and relation to illness severity, work function and quality of life. J Affect Disord 324:162–169. https://doi.org/10.1016/j.jad.2022.12.122
Liu YH, Wang YR, Wang QH et al (2021) Post-infection cognitive impairments in a cohort of elderly patients with COVID-19. Mol Neurodegener 16:1–10. https://doi.org/10.1186/s13024-021-00469-w
Basagni B, Abbruzzese L, Damora A et al (2023) Applied neuropsychology: adult cognition in COVID-19 infected patients undergoing invasive ventilation : results from a multicenter retrospective study. Appl Neuropsychol Adult. https://doi.org/10.1080/23279095.2023.2181083
Hartung TJ, Neumann C, Bahmer T et al (2022) Fatigue and cognitive impairment after COVID-19: A prospective multicentre study. EClinicalMedicine 53:101651. https://doi.org/10.1016/j.eclinm.2022.101651
Asadi-Pooya AA, Akbari A, Emami A et al (2022) Long COVID syndrome-associated brain fog. J Med Virol 94:979–984. https://doi.org/10.1002/jmv.27404
Navarra-Ventura G, Gomà G, de Haro C et al (2021) Virtual reality-based early neurocognitive stimulation in critically ill patients: a pilot randomized clinical trial. J Pers Med 11:1–14. https://doi.org/10.3390/jpm11121260
Godoy-González M, Navarra-Ventura G, Gomà G et al (2023) Objective and subjective cognition in survivors of COVID-19 one year after ICU discharge: the role of demographic, clinical, and emotional factors. Crit Care 27:1–11. https://doi.org/10.1186/s13054-023-04478-7
Fernández-Gonzalo S, Navarra-Ventura G, Bacardit N et al (2020) Cognitive phenotypes 1 month after ICU discharge in mechanically ventilated patients: a prospective observational cohort study. Crit Care 24:1–11. https://doi.org/10.1186/s13054-020-03334-2
Costas-Carrera A, Sánchez-Rodríguez MM, Cañizares S et al (2022) Neuropsychological functioning in post-ICU patients after severe COVID-19 infection: the role of cognitive reserve. Brain Behav Immun Health 21:100425. https://doi.org/10.1016/J.BBIH.2022.100425
Rajoub B (2020) Chapter 3 - Supervised and unsupervised learning. In: Zgallai W (ed) Developments in biomedical engineering and bioelectronics. Biomedical signal processing and artificial intelligence in healthcare. Academic Press, pp 51–89. ISSN 25897527, ISBN 9780128189467
Rankin D, Black M, Flanagan B et al (2020) Identifying key predictors of cognitive dysfunction in older people using supervised machine learning techniques: observational study. JMIR Med Inform 8:1–23. https://doi.org/10.2196/20995
Mukerji SS, Petersen KJ, Pohl KM et al (2023) Machine learning approaches to understand cognitive phenotypes in people with HIV. J Infect Dis 227:S48–S57. https://doi.org/10.1093/infdis/jiac293
Kim SY, Lim TS, Lee HY, Moon SY (2014) Clustering mild cognitive impairment by mini-mental state examination. Neurol Sci 35:1353–1358. https://doi.org/10.1007/s10072-014-1711-y
Lewandowski KE, Sperry SH, Cohen BM, Öngür D (2014) Cognitive variability in psychotic disorders: a cross-diagnostic cluster analysis. Psychol Med 44:3239–3248. https://doi.org/10.1017/S0033291714000774
Habehh H, Gohel S (2021) Machine learning in healthcare. Curr Genom 22:291–300. https://doi.org/10.2174/1389202922666210705124359
Uddin S, Khan A, Hossain ME, Moni MA (2019) Comparing different supervised machine learning algorithms for disease prediction. BMC Med Inform Decis Mak 19:1–16. https://doi.org/10.1186/s12911-019-1004-8
Abós A, Baggio HC, Segura B et al (2017) Discriminating cognitive status in Parkinson’s disease through functional connectomics and machine learning. Sci Rep 7:1–13. https://doi.org/10.1038/srep45347
Ambrosen KS, Skjerbæk MW, Foldager J et al (2020) A machine-learning framework for robust and reliable prediction of short- and long-term treatment response in initially antipsychotic-naïve schizophrenia patients based on multimodal neuropsychiatric data. Transl Psychiatry. https://doi.org/10.1038/s41398-020-00962-8
Stamatis CA, Batistuzzo MC, Tanamatis T et al (2021) Using supervised machine learning on neuropsychological data to distinguish OCD patients with and without sensory phenomena from healthy controls. Br J Clin Psychol 60:77–98. https://doi.org/10.1111/bjc.12272
Shahid S, Iftikhar S (2022) Machine learning models to predict neuropsychiatric disorders in various brain tumors. Curr Med Res Opin 38:687–696. https://doi.org/10.1080/03007995.2022.2043654
Ariza M, Cano N, Segura B et al (2022) Neuropsychological impairment in post-COVID condition individuals with and without cognitive complaints. Front Aging Neurosci. https://doi.org/10.3389/fnagi.2022.1029842
Marshall JC, Murthy S, Diaz J et al (2020) A minimal common outcome measure set for COVID-19 clinical research. Lancet Infect Dis 20:e192–e197. https://doi.org/10.1016/S1473-3099(20)30483-7
Ojeda N, del Pino R, Ibarretxe-Bilbao N et al (2016) Montreal Cognitive Assessment Test: normalization and standardization for Spanish population. Rev Neurol 63:488–496. https://doi.org/10.33588/rn.6311.2016241
Gomar JJ, Ortiz-Gil J, McKenna PJ et al (2011) Validation of the Word Accentuation Test (TAP) as a means of estimating premorbid IQ in Spanish speakers. Schizophr Res 128:175–176. https://doi.org/10.1016/j.schres.2010.11.016
ISCO - International Standard Classification of Occupations. https://www.ilo.org/public/english/bureau/stat/isco/. Accessed 23 Nov 2023
Reitan RMM (1958) Validity of the Trail Making Test as an indicator of organic brain damage. Perceptual Motos Skills 8:271–276. https://doi.org/10.2466/PMS.8.7.271-276
Peña-Casanova J, Quiñones-Úbeda S, Gramunt-Fombuena N et al (2009) Spanish Multicenter Normative Studies (NEURONORMA Project): norms for verbal fluency tests. Arch Clin Neuropsychol 24:395–411. https://doi.org/10.1093/ARCLIN/ACP042
Wechsler D (2001) Wais III. Escala de inteligencia de wechsler para adultos. Manual de aplicación. TEA Ediciones. Departamento I+D., Barcelona
Golden CJ (2005) Test de colores y palabras (Stroop). Madrid
Cysique LA, Łojek E, Cheung TCK et al (2021) Assessment of neurocognitive functions, olfaction, taste, mental, and psychosocial health in COVID-19 in adults: recommendations for harmonization of research and implications for clinical practice. J Int Neuropsychol Soc. https://doi.org/10.1017/S1355617721000862
Vakani K, Ratto M, Sandford-James A et al (2023) COVID-19 and cognitive function: Evidence for increased processing speed variability in COVID-19 survivors and multifaceted impairment with long-COVID symptoms. Eur Psychiatry. https://doi.org/10.1192/j.eurpsy.2023.25
Becker JH, Lin JJ, Twumasi A et al (2023) Greater executive dysfunction in patients post-COVID-19 compared to those not infected. Brain Behav Immun 114:111–117. https://doi.org/10.1016/j.bbi.2023.08.014
García-Campayo J, Zamorano E, Ruiz MA et al (2010) Cultural adaptation into Spanish of the generalized anxiety disorder-7 (GAD-7) scale as a screening tool. Health Qual Life Outcomes 8:8. https://doi.org/10.1186/1477-7525-8-8
Diez-Quevedo C, Rangil T, Sanchez-Planell L et al (2001) Validation and utility of the patient health questionnaire in diagnosing mental disorders in 1003 General Hospital Spanish Inpatients. Psychosom Med 63:679–686
Blevins CA, Weathers FW, Davis MT et al (2015) The posttraumatic stress disorder checklist for DSM-5 (PCL-5): development and initial psychometric evaluation. J Trauma Stress 28:489–498. https://doi.org/10.1002/jts
Jackson C (2015) The Chalder Fatigue Scale (CFQ 11). Occup Med (Chic Ill) 65:86–86. https://doi.org/10.1093/OCCMED/KQU168
Román Viñas B, Ribas Barba L, Ngo J, Serra Majem L (2013) Validación en población catalana del cuestionario internacional de actividad física. Gac Sanit 27:254–257. https://doi.org/10.1016/j.gaceta.2012.05.013
Russell DW (1996) UCLA Loneliness Scale (Version 3): reliability, validity, and factor structure. J Pers Assess 66:20–40. https://doi.org/10.1207/s15327752jpa6601_2
Rami L, Valls-Pedret C, Bartrés-Faz D et al (2011) Cognitive reserve questionnaire. Scores obtained in a healthy elderly population and in one with Alzheimer’s disease. Rev Neurol 52:195–201
Breunig MM, Kriegel H-P, Ng RT, Sander J (2000) LOF: Identifying Density-Based Local Outliers. In: Proceedings of the 2000 ACM SIGMOD International Conference on management of data. pp 93–104
Lundberg SM, Lee SI (2017) A unified approach to interpreting model predictions. Adv Neural Inf Process Syst 2017-Decem:4766–4775
Becker JH, Lin JJ, Doernberg M et al (2021) Assessment of Cognitive Function in Patients After COVID-19 Infection. JAMA Netw Open 4:e2130645–e2130645. https://doi.org/10.1001/JAMANETWORKOPEN.2021.30645
García-Sánchez C, Calabria M, Grunden N et al (2022) Neuropsychological deficits in patients with cognitive complaints after COVID-19. Brain Behav. https://doi.org/10.1002/brb3.2508
Delgado-Alonso C, Valles-Salgado M, Delgado-Álvarez A et al (2022) Cognitive dysfunction associated with COVID-19: a comprehensive neuropsychological study. J Psychiatr Res 150:40–46. https://doi.org/10.1016/j.jpsychires.2022.03.033
Davis HE, Assaf GS, McCorkell L et al (2021) Characterizing long COVID in an international cohort: 7 months of symptoms and their impact. EClinicalMedicine 38:101019. https://doi.org/10.1016/J.ECLINM.2021.101019/ATTACHMENT/499C606A-AE36-49F5-87DD-09E3B87369C9/MMC1.DOCX
Stern Y, Arenaza-Urquijo EM, Bartrés-Faz D et al (2020) Whitepaper: Defining and investigating cognitive reserve, brain reserve, and brain maintenance. Alzheimer’s Dement 16:1305–1311. https://doi.org/10.1016/j.jalz.2018.07.219
Opdebeeck C, Martyr A, Clare L (2016) Cognitive reserve and cognitive function in healthy older people: a meta-analysis. Aging Neuropsychol Cogn 23:40–60. https://doi.org/10.1080/13825585.2015.1041450
Wang HX, MacDonald SWS, Dekhtyar S, Fratiglioni L (2017) Association of lifelong exposure to cognitive reserve-enhancing factors with dementia risk: a community-based cohort study. PLoS Med 14:1–17. https://doi.org/10.1371/journal.pmed.1002251
Ding D, Zhao Q, Wu W et al (2020) Prevalence and incidence of dementia in an older Chinese population over two decades: the role of education. Alzheimer’s and Dementia 16:1650–1662. https://doi.org/10.1002/alz.12159
Ojala-Oksala J, Jokinen H, Kopsi V et al (2012) Educational history is an independent predictor of cognitive deficits and long-term survival in postacute patients with mild to moderate ischemic stroke. Stroke 43:2931–2935. https://doi.org/10.1161/STROKEAHA.112.667618
Kesler SR, Adams HF, Blasey CM, Bigler ED (2010) Premorbid intellectual functioning, education, and brain size in traumatic brain injury: an investigation of the cognitive reserve hypothesis. Appl Neuropsychol 10:153–162. https://doi.org/10.1207/S15324826AN1003_04
Grafman J, Jonas BS, Martin A et al (1988) Intellectual function following penetrating head injury in Vietnam veterans. Brain 111(Pt 1):169–184. https://doi.org/10.1093/BRAIN/111.1.169
Amoretti S, Verdolini N, Mezquida G et al (2021) Identifying clinical clusters with distinct trajectories in first-episode psychosis through an unsupervised machine learning technique. Eur Neuropsychopharmacol 47:112–129. https://doi.org/10.1016/j.euroneuro.2021.01.095
Lin X, Lu D, Zhu Y et al (2020) The effects of cognitive reserve on predicting and moderating the cognitive and psychosocial functioning of patients with bipolar disorder. J Affect Disord 260:222–231. https://doi.org/10.1016/j.jad.2019.09.019
Stern Y (2012) Cognitive reserve in ageing. Lancet Neurol 11:1006–1012. https://doi.org/10.1016/S1474-4422(12)70191-6.Cognitive
Paniz-Mondolfi A, Bryce C, Grimes Z et al (2020) Central nervous system involvement by severe acute respiratory syndrome coronavirus -2 (SARS-CoV-2). J Med Virol. https://doi.org/10.1002/jmv.25915
Sankowski R, Mader S, Valdés-Ferrer SI (2015) Systemic inflammation and the brain: novel roles of genetic, molecular, and environmental cues as drivers of neurodegeneration. Front Cell Neurosci. https://doi.org/10.3389/FNCEL.2015.00028
Lee M-H, Perl DP, Nair G et al (2021) Microvascular injury in the brains of patients with covid-19. N Engl J Med 384:481–483. https://doi.org/10.1056/nejmc2033369
Wang H, Tang X, Fan H et al (2020) Potential mechanisms of hemorrhagic stroke in elderly COVID-19 patients. Aging 12:10022–10034. https://doi.org/10.18632/aging.103335
Heming N, Mazeraud A, Verdonk F et al (2017) Neuroanatomy of sepsis-associated encephalopathy. Crit Care 21:65
Taquet M, Geddes JR, Husain M et al (2021) 6-month neurological and psychiatric outcomes in 236 379 survivors of COVID-19: a retrospective cohort study using electronic health records. Lancet Psychiatry. https://doi.org/10.1016/S2215-0366(21)00084-5
Mazza MG, Palladini M, Villa G et al (2023) Prevalence of depression in SARS-CoV-2 infected patients: an umbrella review of meta-analyses. Gen Hosp Psychiatry 80:17–25. https://doi.org/10.1016/j.genhosppsych.2022.12.002
Radmanesh D, Powell E, Trinh H (2023) Too tired to think: Relationship between post-COVID-19 fatigue and cognition in a veteran sample. Neuropsychol Rehabil. https://doi.org/10.1080/09602011.2023.2244159
Whiteside DM, Basso MR, Naini SM et al (2022) Outcomes in post-acute sequelae of COVID-19 (PASC) at 6 months post-infection part 1: cognitive functioning. Clin Neuropsychol 36:806–828. https://doi.org/10.1080/13854046.2022.2030412
Poletti S, Palladini M, Mazza MG et al (2022) Long-term consequences of COVID-19 on cognitive functioning up to 6 months after discharge: role of depression and impact on quality of life. Eur Arch Psychiatry Clin Neurosci 272:773–782. https://doi.org/10.1007/s00406-021-01346-9
Brown LA, Ballentine E, Zhu Y et al (2022) The unique contribution of depression to cognitive impairment in Post-Acute Sequelae of SARS-CoV-2 infection. Brain Behav Immun Health 22:100460. https://doi.org/10.1016/J.BBIH.2022.100460
Saucier J, Jose C, Beroual Z et al (2023) Cognitive inhibition deficit in long COVID-19: an exploratory study. Front Neurol 14:1–14. https://doi.org/10.3389/fneur.2023.1125574
Mazza MG, Palladini M, De Lorenzo R et al (2021) Persistent psychopathology and neurocognitive impairment in COVID-19 survivors: effect of inflammatory biomarkers at three-month follow-up. Brain Behav Immun 94:138–147. https://doi.org/10.1016/J.BBI.2021.02.021
Knight MJ, Baune BT (2018) Cognitive dysfunction in major depressive disorder. Curr Opin Psychiatry 31:26–31. https://doi.org/10.1097/YCO.0000000000000378
Snyder HR (2013) Major depressive disorder is associated with broad impairments on neuropsychological measures of executive function: a meta-analysis and review. Psychol Bull 139:81–132. https://doi.org/10.1037/a0028727
Pariante CM (2017) Why are depressed patients inflamed? A reflection on 20 years of research on depression, glucocorticoid resistance and inflammation. Eur Neuropsychopharmacol 27:554–559. https://doi.org/10.1016/j.euroneuro.2017.04.001
McAfoose J, Baune BT (2009) Evidence for a cytokine model of cognitive function. Neurosci Biobehav Rev 33:355–366. https://doi.org/10.1016/j.neubiorev.2008.10.005
Glass CK, Saijo K, Winner B et al (2010) Mechanisms underlying inflammation in neurodegeneration. Cell 140:918–934. https://doi.org/10.1016/j.cell.2010.02.016
Singh R, Rathore SS, Khan H et al (2022) Association of obesity with COVID-19 severity and mortality: an updated systemic review, meta-analysis, and meta-regression. Front Endocrinol (Lausanne) 13:1–18. https://doi.org/10.3389/fendo.2022.780872
Hassing LB, Dahl AK, Pedersen NL, Johansson B (2010) Overweight in midlife is related to lower cognitive function 30 years later: a prospective study with longitudinal assessments. Dement Geriatr Cogn Disord 29:543–552. https://doi.org/10.1159/000314874
Yang Y, Shields GS, Guo C, Liu Y (2018) Executive function performance in obesity and overweight individuals: a meta-analysis and review. Neurosci Biobehav Rev 84:225–244. https://doi.org/10.1016/j.neubiorev.2017.11.020
Anand SS, Friedrich MG, Lee DS et al (2022) Evaluation of adiposity and cognitive function in adults. JAMA Netw Open. https://doi.org/10.1001/jamanetworkopen.2021.46324
Garcia-Garcia I, Jurado MA, Garolera M et al (2012) Functional connectivity in obesity during reward processing. Neuroimage 66C:232–239. https://doi.org/10.1016/j.neuroimage.2012.10.035
Hendrick OM, Luo X, Zhang S, Li CSR (2012) Saliency processing and obesity: a preliminary imaging study of the stop signal task. Obesity 20:1796–1802. https://doi.org/10.1038/oby.2011.180
García-García I, Jurado MÁ, Garolera M et al (2013) Alterations of the salience network in obesity: a resting-state fMRI study. Hum Brain Mapp 34:2786–2797. https://doi.org/10.1002/hbm.22104
Wellen KE, Hotamisligil GS (2005) Inflammation, stress, and diabetes. J Clin Invest 115:1111–1119. https://doi.org/10.1172/JCI25102
Bourassa K, Sbarra DA (2017) Body mass and cognitive decline are indirectly associated via inflammation among aging adults. Brain Behav Immun 60:63–70. https://doi.org/10.1016/j.bbi.2016.09.023
Lasselin J, Magne E, Beau C et al (2016) Low-grade inflammation is a major contributor of impaired attentional set shifting in obese subjects. Brain Behav Immun 58:63–68. https://doi.org/10.1016/j.bbi.2016.05.013
Torres-Ruiz J, Lomelín-Gascón J, Lira Luna J et al (2023) Novel clinical and immunological features associated with persistent post-acute sequelae of COVID-19 after six months of follow-up: a pilot study. Infect Dis 55:243–254. https://doi.org/10.1080/23744235.2022.2158217
Xiang M, Wu X, Jing H et al (2023) The intersection of obesity and (long) COVID-19: Hypoxia, thrombotic inflammation, and vascular endothelial injury. Front Cardiovasc Med 10:1–13. https://doi.org/10.3389/fcvm.2023.1062491
World Health Organization (2011) Social determinants of health. https://www.who.int/health-topics/social-determinants-of-health#tab=tab_1. Accessed 25 Oct 2023
Cadar D, Lassale C, Davies H et al (2018) Individual and area-based socioeconomic factors associated with dementia incidence in England: evidence from a 12-year follow-up in the English longitudinal study of ageing. JAMA Psychiat 75:723–732. https://doi.org/10.1001/jamapsychiatry.2018.1012
Minkler M, Fuller-Thomson E, Guralnik JM (2006) Gradient of disability across the socioeconomic spectrum in the United States. N Engl J Med 355:695–703. https://doi.org/10.1056/nejmsa044316
Corney KB, Pasco JA, Stuart AL et al (2023) Social determinants of health and cognitive function: A cross-sectional study among men without dementia. Brain Behav. https://doi.org/10.1002/brb3.3235
Núñez-Cortés R, Ortega-Palavecinos M, Soto-Carmona C et al (2021) Determinantes sociales de la salud asociados a la severidad y mortalidad en pacientes con COVID-19. Gac Med Mex 157:263–270. https://doi.org/10.24875/GMM.M21000556
Ramírez IJ, Lee J (2020) COVID-19 emergence and social and health determinants in Colorado: a rapid spatial analysis. Int J Environ Res Public Health 17:1–15. https://doi.org/10.3390/ijerph17113856
Miskowiak KW, Johnsen S, Sattler SM et al (2021) Cognitive impairments four months after COVID-19 hospital discharge: pattern, severity and association with illness variables. Eur Neuropsychopharmacol 46:39. https://doi.org/10.1016/J.EURONEURO.2021.03.019
Group P-CC, Evans RA, McAuley H et al (2021) Physical, cognitive and mental health impacts of COVID-19 following hospitalization—a multi-centre prospective cohort study. Cold Spring Harbor Laboratory Press
Prabhakaran D, Day GS, Munipalli B et al (2023) Neurophenotypes of COVID-19: Risk factors and recovery outcomes. Brain Behav Immun Health. https://doi.org/10.1016/j.bbih.2023.100648
Díez-Cirarda M, Yus M, Gómez-Ruiz N et al (2023) Multimodal neuroimaging in post-COVID syndrome and correlation with cognition. Brain 146:2142–2152. https://doi.org/10.1093/brain/awac384
Acknowledgements
NAUTILUS—Project Collaborative Group list of affiliation: Jose A. Bernia, Servei d’Anestesia Reanimació i Clinica del Dolor, Consorci Sanitari de Terrassa (CST) (Terrassa, Barcelona, Spain). Vanesa Arauzo, Servei de Medicina Intensiva. Consorci Sanitari de Terrassa (CST) (Terrassa, Barcelona, Spain). Marta Balague-Marmaña, Hospital Sant Joan Despí Moisès Broggi, Consorci Sanitari Integral (Sant Joan Despí, Spain). Pérez-Pellejero, Cristian, Hospital Sant Joan Despí Moisès Broggi, Consorci Sanitari Integral (Sant Joan Despí, Barcelona, Spain). Silvia Cañizares. Hospital Clinic de Barcelona (Barcelona, Spain). Jose Antonio Lopez Muñoz. Occupational Health Care Service, Hospital Clínic (Barcelona, Spain). Jesús Caballero, Hospital Universitari Arnau de Vilanova (Lleida, Spain). Anna Carnes-Vendrell, Hospital Universitari de Santa Maria (Lleida, Spain). Gerard Piñol-Ripoll, Hospital Universitari de Santa Maria (Lleida, Spain). Ester Gonzalez-Aguado, Consorci Sanitari Alt Penedès-Garraf (Vilafranca de Penedés, Barcelona, Spain). Mar Riera-Pagespetit, Consorci Sanitari Alt Penedès-Garraf (Vilafranca de Penedés, Barcelona, Spain). Eva Forcadell-Ferreres, Hospital Verge de la Cinta, (Tortosa, Tarragona, Spain). Silvia Reverte-Vilarroya, Hospital Verge de la Cinta, (Tortosa, Tarragona, Spain). Susanna Forné, Fundació Sant Hospital de la Seu d’Urgell (La Seu d’Urgell, Lleida, Spain). Jordina Muñoz-Padros, Consorci Hospitalari de Vic (Vic, Barcelona, Spain). Anna Bartes-Plan, Consorci Hospitalari de Vic (Vic, Barcelona, Spain). Jose A. Muñoz-Moreno, Servei de Malalties Infeccioses, Fundació Lluita contra les Infeccions—Hospital Universitari Germans Trias i Pujol (Badalona, Barcelona, Spain). Anna Prats-Paris, Servei de Malalties Infeccioses, Fundació Lluita contra les Infeccions—Hospital Universitari Germans Trias i Pujol (Badalona, Barcelona, Spain). Inmaculada Rico Pons, Hospital Universitari de Bellvitge (L’Hospitalet de Llobregat, Barcelona, Spain). Judit Martínez Molina, Hospital Universitari de Bellvitge (L’Hospitalet de Llobregat, Barcelona, Spain). Laura Casas-Henanz, Hospital Universitari Mútua Terrassa (Terrassa, Barcelona, Spain). Judith Castejon, Hospital Universitari Mútua Terrassa (Terrassa, Barcelona, Spain). Maria José Ciudad Mas, Badalona Serveis Assistencials (Badalona, Barcelona, Spain). Anna Ferré Jodrà, Badalona Serveis Assistencials (Badalona, Barcelona, Spain). Manuela Lozano, Institut d’Assistència Sanitària (Girona, Spain). Tamar Garzon, Institut d’Assistència Sanitària (Girona, Spain). Marta Cullell , Fundació Salut Empordà (Figueres, Girona, Spain). Sonia Vega, Fundació Salut Empordà (Figueres, Girona, Spain). Sílvia Alsina, Fundació Hospital de Puigcerdà (Puigcerdà, Girona, Spain). Maria J. Maldonado-Belmonte, Hospital Universitario Central de la Cruz Roja San José y Santa Adela (Madrid, Spain). Susana Vazquez-Rivera, Hospital Universitario Central de la Cruz Roja San José y Santa Adela (Madrid, Spain). Eloy García-Cabello. Clínica Universitaria de Psicología. Facultad de Ciencias de la Salud. Universidad Fernando Pessoa (La Palmas de Gran Canaria, Islas Canarias, Spain). Yaiza Molina. Clínica Universitaria de Psicología. Facultad de Ciencias de la Salud. Universidad Fernando Pessoa (La Palmas de Gran Canaria, Islas Canarias, Spain). Sandra Navarro, Servei Andorrà d’Atenció Sanitària (SAAS) (Andorra). Eva Baillès, Servei Andorrà d’Atenció Sanitària (SAAS) (Andorra).
Funding
Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This research was supported by the Agency for Management of University and Research Grants (AGAUR) from the Generalitat de Catalunya (Pandemies, 2020PANDE00053), the La Marató de TV3 Foundation (202111–30-31–32), the Ministerio de Ciencia e Innovación (TED2021-130409B-C55).
Author information
Authors and Affiliations
Consortia
Contributions
Conceptualization: MA, JB, MG, CJ, and BS; data curation: MA, JB, and CB; formal analysis: JB and CB; funding acquisition: MG, MA, CJ, BS, CUC, JB; investigation: NC and NAUTILUS Project Collaborative group; methodology: MA and JB; project administration: MG, CJ, and BS; software: JB and CB; supervision: NC; validation: MA, MG, CJ, and BS; visualization: JB and MA; writing—original draft: MA and JB; writing—review and editing: all authors.
Corresponding authors
Ethics declarations
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Additional information
The members of the NAUTILUS Project Collaborative Group are listed in the Acknowledgments.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Ariza, M., Béjar, J., Barrué, C. et al. Cognitive reserve, depressive symptoms, obesity, and change in employment status predict mental processing speed and executive function after COVID-19. Eur Arch Psychiatry Clin Neurosci (2024). https://doi.org/10.1007/s00406-023-01748-x
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00406-023-01748-x