Predictive roles of cognitive biases in health anxiety: A machine learning approach.

Abstract

Prior work suggests that cognitive biases may contribute to health anxiety. Yet there is little research investigating how biased attention, interpretation, and memory for health threats are collectively associated with health anxiety, as well as the relative importance of these cognitive processes in predicting health anxiety. This study aimed to build a prediction model for health anxiety with multiple cognitive biases as potential predictors and to identify the biased cognitive processes that best predict individual differences in health anxiety. A machine learning algorithm (elastic net) was performed to recognise the predictors of health anxiety, using various tasks of attention, interpretation, and memory measured across behavioural, self-reported, and computational modelling approaches. Participants were 196 university students with a range of health anxiety severity from mild to severe. The results showed that only the interpretation bias for illness and the attention bias towards symptoms significantly contributed to the prediction model of health anxiety, with both biases having positive weights and the former being the most important predictor. These findings underscore the central role of illness-related interpretation bias and suggest that combined cognitive bias modification may be a promising method for alleviating health anxiety.