Difficulty focusing and following conversations is the most accurate characterization of so-called brain fog, according to a study in 25,796 participants.

Brain fog has gained increased attention since the COVID-19 pandemic, but very little is known about this condition. It is a broad, subjective term with diverse symptoms across various conditions. “Given this inherent heterogeneity, the best way to characterize brain fog is to capture subjective descriptions in a large population sample. This data was collected through Mindstep, a validated UK-only smartphone application for remote data collection,” explained Mohammad Mahmud, MBBS, who is affiliated with Mindstep, at EAN 2023.

The app includes a broad range of questions and tests, many of which have been validated. The researchers systematically studied the associations between 29 variables with self-reported brain fog via univariate and machine-learning methods. Studied variables included clinical comorbidities, lifestyle factors, symptoms, functional deficits, and cognitive scores.

The researchers screened 25,796 participants, 7,280 (28.2%) of whom reported experiencing brain fog. Dr. Mahmud and colleagues focused on 29 variables in four groups: demographics, comorbidities, symptoms, and lifestyle. Of demographics, older age (35.7 vs 32.8 years; P<0.0001), female sex (OR, 1.2; P<0.001), and higher BMI (28.8 vs 27.9; P<0.0001) were associated with brain fog. Of comorbidities, long COVID-19 (OR, 3.8; P<0.0001) and concussion (OR, 2.4; P<0.0001) had the strongest association with brain fog. Associated symptoms were difficulty focusing or concentrating (OR, 3.3; P<0.0001) and difficulty following conversations (OR, 2.2; P<0.0001). Associated lifestyle factors were less exercise (0.21 vs 0.22) and reduced sleep quality (Sleep Quality Scale: 4.1 vs 4.4; P<0.0001).

Their study resulted in a predictive model with these top five features by XGBoost feature importance (“Gain”):

1. Difficulty following conversations
2. Migraine Disability Assessment Questionnaire (MIDAS)
3. Long COVID-19
4. Difficulty focusing/concentrating
5. Difficulty remembering appointments

Extreme gradient boosting algorithms (an ensemble of decision trees) achieved a training accuracy of 84% with a fivefold cross-validated accuracy of 74% and could be used in the future.

Copyright ©2023 Medicom Medical Publishers