Machine learning-ready mental health datasets for evaluating psychological effects and system needs in Mexico city during the first year of the COVID-19 pandemic

Abstract

The prevalence of mental health problems constitutes an open challenge for modern societies, particularly for low and middle-income countries with wide gaps in mental health support. With this in mind, five datasets were analyzed to track mental health trends in Mexico City during the pandemic's first year. This included 33, 234 responses to an online mental health risk questionnaire, 349, 202 emergency calls, and city epidemiological, mobility, and online trend data. The COVID-19 mental health risk questionnaire collects information on socioeconomic status, health conditions, bereavement, lockdown status, and symptoms of acute stress, sadness, avoidance, distancing, anger, and anxiety, along with binge drinking and abuse experiences. The lifeline service dataset includes daily call statistics, such as total, connected, and abandoned calls, average quit time, wait time, and call duration. Epidemiological, mobility, and trend data provide a daily overview of the city's situation. The integration of the datasets, as well as the preprocessing, optimization, and machine learning algorithms applied to them, evidence the usefulness of a combined analytic approach and the high reuse potential of the data set, particularly as a machine learning training set for evaluating and predicting anxiety, depression, and post-traumatic stress disorder, as well as general psychological support needs and possible system loads.