Discovering Latent Psychological Structures from Self-Report Assessments of Hospital Workers

Hospitals are high-stress environments where workers face a high risk of occupational burnout due to a mix of imbalanced schedules, understaffing, and emotional stress. In this paper, we propose a computational framework to infer the latent psychological makeup and traits of hospital workers. We apply machine learning models to psychometric data obtained from a suite of psychological survey instruments, collected as a part of TILES, a ten-week research study carried out in a large Los Angeles hospital. The study population represents over 200 hospital employees, including nurses and those in administrative positions. A computational framework that combines clustering and non-negative matrix factorization was used to extract the latent interplay between psychological constructs along dimensions of health, affect, personality, cognitive ability, and job performance. We illustrate how the proposed framework can help reveal the latent psychological structures related to occupational burnout.