Cognitive reserve, depressive symptoms, obesity, and change in employment status predict mental processing speed and executive function after COVID-19

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.

Keywords: Clustering; Executive function; Logistic regression; Machine learning; Mental speed processing; Post-COVID-19 condition.

Fig. 1

Flowchart for data analysis

Fig. 2

Principal components of the clusters obtained by K-means on the neurophysiological variables

Fig. 3

Boxplots grouped by group for each neuropsychological variable. The values ​​of the cognitive tests are standardized with respect to a healthy control group. BP bad performance, GP good performance, SCWT Stroop Color and Word Test, TMT = Trail Making Test

Fig. 4

Weights assigned to each one of the attributes in the logistic regression. All variables are used by the logistic regression and have a similar range of weights, with a sign according to the influence on the corresponding classes. SCWT Stroop Color and Word Test, TMT = Trail Making Test

Fig. 5

Principal components of the clusters obtained by K-means on the sociodemographic, clinical, and lifestyle characteristics selected by the logistic regression model

Fig. 6

Weights assigned to the selected variables by the logistic regression model. The sign of the weights represents the impact of the variable on the prediction of each of the classes (positive leans to GP class, negative leans to BP class). GP good performance cluster, BP bad performance cluster, CRC Cognitive Reserve Questionnaire, PHQ-9 Patient Health Questionnaire-9

Fig. 7

Mean Shapley values computed on the test set for each variable in the model. This represents the mean effect in the absolute value of each variable, that is, the expected change from the mean prediction (probability 0.5) of each variable. CRC Cognitive Reserve Questionnaire, PHQ-9 Patient Health Questionnaire-9. The Shapley values for the model compute the mean of the influence of each variable for changing the class from the mean prediction (same probability for both classes). A higher value corresponds to a higher influence on the final model answer

Fig. 8

Distribution of the Shapley values of the test examples with respect to the magnitude of the data variables. CRC = Cognitive Reserve Questionnaire; PHQ-9 = Patient Health Questionnaire-9

Fig. 9

Shapley mean values ​​calculated for variables segregated by sex for each variable in the model. CRC Cognitive Reserve Questionnaire, PHQ-9 Patient Health Questionnaire-9

Fig. 10

Shapley mean values ​​calculated for variables segregated by severity for each variable in the model. CRC Cognitive Reserve Questionnaire, PHQ-9 Patient Health Questionnaire-9