Adoecimento Mental em Policiais: Identificação de Preditores para o Desenvolvimento de um Sistema de Intervenção Precoce

Abstract

This dissertation, comprising two studies, aimed to identify predictors of mental illness in order to propose selection criteria for the development of an Early Intervention System (EIS) in mental health. Both studies were conducted using de-identified administrative records provided by a police institution in the Midwest region of Brazil for this purpose. Study 1 conducted an epidemiological analysis of mental and behavioral disorders (MBD) among the institution's officers from 2014 to 2022, recording 6,298 health-related leaves for 1,944 employees, with an average of 3.24 leaves per employee. The prevalence of MBD ranged from 2.73% to 5.03%, while incidence varied from 2.82% to 5.28%. Neurotic, stress-related, and somatoform disorders were the most frequent, accounting for 42.22% to 55.59% of occurrences, followed by mood disorders, which ranged from 46.67% to 50.84%. These data reveal that 30% of medical leaves were associated with MBD, highlighting the need for preventive strategies. Study 2 identified predictors of MBD through the training of machine learning models. The trained models demonstrated high accuracy: random forest (96.09%), decision tree (95.56%), and logistic regression (89.06%). An integrated interpretation of the three models' results was sought. Recognition at work, such as commendations and awards, acted as a protective factor, while poor performance during the probationary period, occurrences of other illnesses, and job relocations increased the risk of MBD. Finally, we argue that the results provide evidence that interventions focused on recognition and job stability can prevent MBD among officers, and we advocate for the establishment of an EIS in mental health.