Nursing turnover is a leading source of inefficiency in health care delivery but studying nursing turnover is difficult given data limitations. This project will harness data from one of the largest home health companies in the country to identify drivers of nursing turnover in order to predict which nurses are at highest risk of turnover. Because home health nurses work one-on-one with most of their patients, we can also examine the deeply confounded relationship between nursing turnover and patient care. The specific aims include:
Aim 1: Identify the mechanisms that drive turnover among home health nurses, including scheduling, patient acuity, team composition, and state-level scope of practice laws
Aim 2: Using the identified factors, construct an algorithm that can flag nurses at high risk for turnover.
Aim 3: Estimate the impact of nursing turnover on patient outcomes, including readmissions and mortality.
High levels of turnover in professions that rely heavily on skill acquisition is a major source of inefficiency due to productivity loss, termination costs, vacancy, orientation and training, and other spillovers. Turnover among nurses, a bedrock in health care, is a particular concern given the amount of direct care they provide patients - indeed, compromised nursing care has been associated with longer lengths of stay and higher rates of adverse outcomes, including mortality. To date, most studies of nursing turnover have focused on inpatient nurses in hospitals using self-reported survey data and are therefore limited in scope. No projects have been able to cleanly connect nurses to patients, making it impossible to understand the full impact of nursing turnover. This project will leverage a collaborative agreement between the University of Pennsylvania and Encompass Home Health, which is one of the largest home health agencies in the country, currently operating in 30 states. We will acquire a host of different datasets from Encompass to answer two main questions: first, which factors drive nursing turnover in home health; and second, how does nursing turnover in home health affect patient outcomes such as hospitalizations and mortality? Using variables constructed from human resources, payroll, visit logs, and other secondary data sources, we will include a host of potential factors that may drive turnover, e.g., driving time, patient acuity, and outside labor market opportunities, in order to construct an algorithm that can flag nurses at high risk of turnover. This will allow for more tailored personnel interventions to improve retention among home health nurses at Encompass. Next, we will use these unique data to explore the confounded relationship between nursing turnover and patient outcomes. Because home health nurses work alone or in tandem with other provider types, we are able to connect individual nurses to individual patients. That, along with more rigorous statistical approaches, gives us an unprecedented opportunity to pin down the causal pathway between nursing turnover and patient care.