Cookies on this website

We use cookies to ensure that we give you the best experience on our website. If you click 'Accept all cookies' we'll assume that you are happy to receive all cookies and you won't see this message again. If you click 'Reject all non-essential cookies' only necessary cookies providing core functionality such as security, network management, and accessibility will be enabled. Click 'Find out more' for information on how to change your cookie settings.

<jats:p>Background: Using UK Biobank data from baseline to the imaging study, our aim was to investigate whether participants with comorbid psychiatric disorders suffer from higher risk of impaired cognition over time compared to those without. There are several conventional statistical techniques to examine whether this pattern indicates the increased risk of comorbidity of the aforementioned, however, they carry certain inherent limitations. Machine learning (ML) has shown specific advantages in examining potential predictors simultaneously in an unbiased manner, especially its ability to identify patterns of information within useful features for the prediction of an outcome of interest. We applied ML techniques on UK Biobank data to examine whether anxiety and/or depression are important longitudinal predictors of impaired cognition. Methods: We used data from UK Biobank (n = 1.158) across three time waves to longitudinally assess the effects of comorbidity (anxiety and depression, measured through self-report scales) on cognition. Theta (Θ) measures for each mental health scale were computed using item response theory (IRT) and intraindividual variability of reaction time performance (IIV) - raw standard deviation - was used as a measure of cognitive performance. First, comorbidity information was summarized to show the variation of impaired cognition for participants with and without psychiatric disorders. A machine learning (ML) approach was then applied to examine if psychiatric disorders and other putative covariate markers may be important predictors of long-term cognitive decline. Results: For participants with anxiety, the percentage of having impaired cognition constantly increases through time from 35.09% to 42.15%; in contrast, an increase with less magnitude is found for participants without anxiety from 36.10% to 40.03%. Likewise, for participants with depression, the percentage of having impaired cognition constantly increases through time from 35.49% to 41.67%; in contrast, an increase with less magnitude is found for participants without depression from 35.64% to 40.52%. Using the area under the Receiver Operating Characteristic (ROC) curve it was observed that the anxiety model achieved the best performance among all models, with an Area Under the Curve (AUC) of 0.68, followed by the depression model with an AUC of 0.64. The cardiovascular and diabetes model and the demographics model had relatively weak performance in predicting cognition, with an AUC of 0.60 and 0.57, respectively. Conclusions: Using data from UK Biobank, this study provides empirical evidence which suggests that psychiatric disorders are important comorbidities of cognitive decline. Furthermore, when other comorbidities were included in the model these were not as important on long-term effect. When the recurrent neural networks were trained for the psychiatric disorder features (anxiety and depression) they showed improved performance in predicting cognition in comparison with cardiovascular disease, diabetes and demographic factors. These findings suggest that mental health disorders (anxiety and depression) have a deleterious effect on long-term cognition, and may be considered an important comorbid disorder of cognitive decline. The implications of this work is that the important predictive effect of poor mental health on longitudinal cognitive decline should be considered in both research and clinical settings.</jats:p>

Original publication




Journal article


Cold Spring Harbor Laboratory

Publication Date