To develop a nurse staffing prediction model and evaluate deviation from predicted nurse staffing as a contributor to patient outcomes.
Secondary data collection conducted 2017‐2018, using the California Office of Statewide Health Planning and Development and the California Perinatal Quality Care Collaborative databases. We included 276 054 infants born 2008‐2016 and cared for in 99 California neonatal intensive care units (NICUs).
Repeated‐measures observational study. We developed a nurse staffing prediction model using machine learning and hierarchical linear regression and then quantified deviation from predicted nurse staffing in relation to health care‐associated infections, length of stay, and mortality using hierarchical logistic and linear regression.
We linked NICU‐level nurse staffing and organizational data to patient‐level risk factors and outcomes using unique identifiers for NICUs and patients.
An 11‐factor prediction model explained 35 percent of the nurse staffing variation among NICUs. Higher‐than‐predicted nurse staffing was associated with decreased risk‐adjusted odds of health care‐associated infection (OR: 0.79, 95% CI: 0.63‐0.98), but not with length of stay or mortality.
Organizational and patient factors explain much of the variation in nurse staffing. Higher‐than‐predicted nurse staffing was associated with fewer infections. Prospective studies are needed to determine causality and to quantify the impact of staffing reforms on health outcomes.