SCH: Wearable Augmented Prediction of Burnout in Nurses: A Synergy of Engineering, Bioethics, Nursing

ABSTRACT: The 21st century workforce is experiencing increasing job demands while employers optimize job resources to meet regulatory, fiscal and productivity standards. This is perhaps most apparent in today’s healthcare system, wherein the workforce is under constant stress to cope with rapidly changing care delivery approaches, widespread adoption of electronic health records, and increased reliance on publicly reported quality metrics. In May 2019, the World Health Organization defined burnout as an occupational phenomenon. Unfortunately, burnout is underrecognized by those who suffer from it, and it typically goes undetected until employees’ performance deteriorates or catastrophes occur in workplace. Therefore, this project’s overarching goal is to develop a data-driven technology for predicting impending burnout before its effects on health and work performance become manifest. As a case study, this project will establish predictability of burnout in registered nurses (RNs). In hospital settings, 35%-45% of RNs report burnout primarily driven by increased work demands (higher patient acuity), work inefficiencies, interpersonal conflict, moral distress, and low level of control over decisions that affect their work. Burnout in RNs is associated with poor patient outcomes (increased risk of medical errors, hospital-acquired infections), lower quality of care, increased absenteeism and poor patient satisfaction. Within this context, the proposed project’s vision and aims are presented. This project’s vision is to develop a technology to predict burnout in RNs (as a case study) by combining workplace, psychological, and physiological factors, and exploring the barriers to adopting such a technology. This effort focuses on the following aims: Aim1. To create a unique, open- access, de-identified dataset that transforms the science of burnout internationally and informs the interaction of continuous physiological measures (measured from smart watches) and repeated (quarterly) psychological (measured using validated rating scales) and work-related factors (administrative databases) for predicting burnout (Aim 2) in RNs at Mayo Clinic’s Florida (Cohorts-A&B) and Rochester (Cohort-C) sites. Aim 2. To develop an analytical framework combining probabilistic graphical models (PGMs) and multitask learning (MTL) to derive interpretable predictions of burnout. PGMs addresses the challenge of inherent stochasticity of burnout manifestation across individuals, and MTL will identify common burnout factors predictive of burnout risks (high, medium and low). Predictability established using Cohort-A will be validated in Cohorts-B&C. Aim 3. Explore barriers (bioethics and administrative) to adopting burnout prediction technologies by assessing perspectives of RNs, nurse supervisors and hospital administrators.