Modelling Time to Death or Discharge in Neonatal Care: An Application of Competing Risks

Background Understanding length of stay for babies in neonatal care is vital for planning services and for counselling parents. While previous work has focused on the length of stay of babies who survive to discharge, when investigating resource use within neonatal care, it is important to also incorporate information on those babies who die while in care. We present an analysis using competing risks methodology which allows the simultaneous modelling of babies who die in neonatal care and those who survive to discharge. Methods Data were obtained on 2723 babies born at 24-28 weeks gestational age in 2006-10 and admitted to neonatal care. Death and discharge alive are two mutually exclusive events and can be treated as competing risks. A flexible parametric modelling approach was used to analyse these two competing events and obtain estimates of the absolute probabilities of death or discharge. Results The absolute probabilities of death or discharge are presented in graphical form showing the cause-specific cumulative incidence over time by gender, gestational age and birthweight. The discharge of babies alive generally occurred over a longer time period for babies of lower gestational age and smaller birthweight than for bigger babies. Conclusion This study has presented a useful statistical method for modelling the length of stay where there are significant rates of in-unit mortality. In health care systems that are increasingly focusing on costs and resource planning, it is essential to consider not only length of stay of survivors but also for those patients who die before discharge.