What predicted stress, anxiety, and depression among Filipinos during COVID-19

The surge of COVID-19 cases in the Philippines came with a surge in mental health problems. Plenty was said about it, but little of it was measured with data. Carlos Alfonso V. Palattao, Geoffrey A. Solano, Cherica A. Tee, and Michael L. Tee used machine learning to find which factors contributed to stress, anxiety, depression, and adverse psychological impact among the general population.

Approach

Data came from an online survey with three parts: questions the researchers wrote on demographics, symptoms and physical health, contact history, knowledge and beliefs about COVID-19, precautions taken in the past 14 days, and pre-existing conditions; the shortened Depression, Anxiety, and Stress Scale (DASS-21); and the Impact of Event Scale-Revised (IES-R), which measures the psychological impact of an event. The respondents were 2,119 students and alumni of the University of the Philippines and their friends and acquaintances.

The survey answers were the predictors and the four scale scores (stress, anxiety, depression, and impact of event) were the targets. Three approaches were compared. Approach A used only the researchers’ questions as the initial feature set. Approaches B and C added the DASS-21 and IES-R responses. Each approach ran its own feature selection (correlation-based feature subset selection with best-first search for A and C, information gain with a ranker for B) before training random forest, naive Bayes, support vector machine, and logistic regression classifiers, scored on average accuracy, precision, and recall.

Findings

Approaches B and C scored significantly higher than A on accuracy, precision, and recall. From this the authors suggest that stress, anxiety, depression, and impact of event are good predictors of one another.

The final feature sets name specific contributors. For stress: how a person rates their own health and whether they have a pre-existing neuropsychiatric condition. For depression: the same two, plus age. For anxiety: actively seeking information on the symptoms of the virus, being at home longer, and having symptoms such as body pain and difficulty breathing. For adverse psychological impact: seeking pandemic information on social media, wanting to know about an effective medicine, concern for family, feeling discriminated against, and having a cough.

What they recommend

The authors present the method as a way to find candidate contributing factors, not a final answer. Psychologists and researchers should validate and describe those factors with further study, so that stakeholders can write policies and guidelines that prevent more distress.

Source: Palattao, C. A. V., Solano, G. A., Tee, C. A., and Tee, M. L. (2021). Determining factors contributing to the psychological impact of the COVID-19 Pandemic using machine learning. 2021 International Conference on Artificial Intelligence in Information and Communication (ICAIIC). doi:10.1109/ICAIIC51459.2021.9415276

Originally shared on Facebook.