Detection of determinants of PHQ 9 score for major depressive symptoms in health workers during the COVID-19 pandemic using machine learning techniques

Luciano Battioni, Consejo de Insuficiencia Cardiaca e Hipertensión Pulmonar, Buenos Aires, Argentina.
Cristhian E. Scatularo, Sanatorio de la Trinidad, Buenos Aires, Argentina
Sebastián Bellia, Consejo de Aspectos Psicosociales. Sociedad Argentina de Cardiología. Buenos Aires, Argentina
Adrián Lescano, Consejo de Insuficiencia Cardiaca e Hipertensión Pulmonar, Sociedad Argentina de Cardiología, Ciudad Autónoma de Buenos Aires, Argentina
Stella M. Pereiro, Consejo de Insuficiencia Cardiaca e Hipertensión Pulmonar, Sociedad Argentina de Cardiología, Ciudad Autónoma de Buenos Aires, Argentina
Julio Giorgini, Consejo de Aspectos Psicosociales. Sociedad Argentina de Cardiología. Buenos Aires, Argentina

Background: SARS-COV2 pandemic has generated deleterious psychological and social effects as reported in our survey IMPPACTS-SAC.20. Objective: To determine which domains of the Patient Health Questionnaire (PHQ 9) have the biggest influence in the diagnosis of major depression, also, to identify subpopulations with high prevalence of this disease. Method: IMPPACTS-SAC.20 survey analysis. Unsupervised machine learning techniques were used to perform a factorial analysis and to create groups of similar cases according to their performance in PHQ 9. Results: 1221 participants that took the PHQ 9 questionnaire were included. Factorial analysis showed that two main dimensions (neurasthenia and negative self-perception) accounted for 67.2% of the questionnaire variance (KMO test 0.911; Bartlett p < 0.001). The combination of both dimensions in hierarchical analysis generated nine clusters. Groups 5, 4, 2 and 1 explained 93% of the major depression cases. Groups 5 and 4 presented high neurasthenia values, and groups 2 and 1 high negative self-perception. Groups 6, 7 and 8 combined, presented a prevalence of major depression of 0.6%. Conclusions: The implementation of machine learning techniques detected two dimensions within the PHQ 9 score, neurasthenia and negative self-perception. Subgroups with a high prevalence of major depression were found, whose main clinical characteristics were female sex, alcohol consumption, smoking and suicidal intention.

Keywords: Depression. Patient Health Questionnaire. Neurasthenia. Self-concept. Healthcare workers. COVID-19.