Learning the Mental Health Impact of COVID-19 in the United States With Explainable Artificial Intelligence: Observational Study

  1. Indra Prakash Jha
  2. Raghav Awasthi
  3. Ajit Kumar
  4. Vibhor Kumar
  5. Tavpritesh Sethi

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

The COVID-19 pandemic has affected the health, economic, and social fabric of many nations worldwide. Identification of individual-level susceptibility factors may help people in identifying and managing their emotional, psychological, and social well-being.

Objective

This study is focused on learning a ranked list of factors that could indicate a predisposition to a mental disorder during the COVID-19 pandemic.

Methods

In this study, we have used a survey of 17,764 adults in the United States from different age groups, genders, and socioeconomic statuses. Through initial statistical analysis and Bayesian network inference, we have identified key factors affecting mental health during the COVID-19 pandemic. Integrating Bayesian networks with classical machine learning approaches led to effective modeling of the level of mental health prevalence.

Results

Overall, females were more stressed than males, and people in the age group 18-29 years were more vulnerable to anxiety than other age groups. Using the Bayesian network model, we found that people with a chronic mental illness were more prone to mental disorders during the COVID-19 pandemic. The new realities of working from home; homeschooling; and lack of communication with family, friends, and neighbors induces mental pressure. Financial assistance from social security helps in reducing mental stress during the COVID-19–generated economic crises. Finally, using supervised machine learning models, we predicted the most mentally vulnerable people with ~80% accuracy.

Conclusions

Multiple factors such as social isolation, digital communication, and working and schooling from home were identified as factors of mental illness during the COVID-19 pandemic. Regular in-person communication with friends and family, a healthy social life, and social security were key factors, and taking care of people with a history of mental disease appears to be even more important during this time.

Article activity feed

  1. Version published to 10.2196/25097
  2. Version published to 10.2196/preprints.25097
  3. Version published to 10.1101/2020.07.19.20157164v2 on medRxiv
  4. ScreenIT

    SciScore for 10.1101/2020.07.19.20157164: (What is this?)

    Please note, not all rigor criteria are appropriate for all manuscripts.

    Table 1: Rigor

    NIH rigor criteria are not applicable to paper type.

    Table 2: Resources

    Software and Algorithms
    SentencesResources
    Datasets: We extracted data of 17764 adults (https://www.covid-impact.org/)8 from weekly surveys of the U.S. adult household population nationwide for 18 regional areas including 10 states (CA, CO, FL, LA, MN, MO, MT1, NY, OR, TX) and 8 Metropolitan Statistical Areas (Atlanta, Baltimore, Birmingham, Chicago, Cleveland, Columbus, Phoenix, Pittsburgh).
    Phoenix
    suggested: (Phoenix, RRID:SCR_003163)
    Different supervised machine learning models - Random-Forest(RF), Support vector machine (SVM), logistic, naive-Bayes, were learned for predicting the response to mental health indicators using the Scikit-learn library 16in Python.
    Scikit-learn
    suggested: (scikit-learn, RRID:SCR_002577)
    Python
    suggested: (IPython, RRID:SCR_001658)

    Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).

    Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:
    Our study has a few limitations. The dataset used for training the model is cross-sectional and we could not comment upon the temporality and persistence of the discovered effects. Secondly, our results were currently limited to only one geography, i.e. the United States. However, the relatively large sample size and multi-ethnic involvement in the survey allowed us to construct a robust network with ensemble averaging and majority voting from 101 bootstrapped networks. Therefore, the discovered influences are likely to hold true in the United States. Further, we believe that this is the first attempt to quantify the impact of social factors on mental health through an explainable AI model and many of our findings are intuitive. The ranking of features and quantification of this impact couldn’t have been intuitively achieved without modeling and this study will provide a basis for many further studies and design of effective social interventions to mitigate the mental health impact of natural health disasters.

    Results from TrialIdentifier: No clinical trial numbers were referenced.

    Results from Barzooka: We did not find any issues relating to the usage of bar graphs.

    Results from JetFighter: We did not find any issues relating to colormaps.

    Results from rtransparent:
    • Thank you for including a conflict of interest statement. Authors are encouraged to include this statement when submitting to a journal.
    • Thank you for including a funding statement. Authors are encouraged to include this statement when submitting to a journal.
    • No protocol registration statement was detected.

    About SciScore

    SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore checks for the presence and correctness of RRIDs (research resource identifiers), and for rigor criteria such as sex and investigator blinding. For details on the theoretical underpinning of rigor criteria and the tools shown here, including references cited, please follow this link.

  5. Version published to 10.1101/2020.07.19.20157164v1 on medRxiv