Anxiety symptoms during public health crises are associated with adverse psychiatric outcomes and impaired health decision-making. Therefore, remotely identifying the severity of short-term anxiety symptoms in the population during lockdown measures is an important public health agenda.
This study compared two machine learning models for predicting clinical anxiety based on the Generalized Anxiety Disorder 7-item scale from time-series data of communication and social networking app usage, and anxiety-associated clinical survey variables, including cohabitation with essential workers, worries about life instability, changes in social interactions, and health status.
The data was collected from a sample of psychiatric outpatients in Madrid, Spain, before and during the mandatory COVID-19 lockdown. Our first pipeline was based on a hidden Markov model (HMM), while in the second model, we opted for a recurrent neural network (RNN) for temporal data processing. Both architectures model the distribution of a sequence of random observations from a set of latent variables; however, in RNN, the latent variable is deterministically deduced from the current observation and the previous latent variable, while, in HMM, the set of (random) latent variables is a Markov chain. The evaluation was performed using 10-fold cross-validation due to limited data. The model accuracy and area under the receiver operating curve (AUROC) mean and standard deviation are reported.
Published in JMIR Mental Health:
- Ryu J, Sükei E, Norbury A, H Liu S, Campaña-Montes J, Baca-Garcia E, Artés A, Perez-Rodriguez M. Shift in Social Media App Usage During COVID-19 Lockdown and Clinical Anxiety Symptoms: Machine Learning–Based Ecological Momentary Assessment Study. DOI: 10.2196/30833