Objective: To assess trends in place of death for children with a life-limiting condition and the factors associated with death at home or hospice rather than hospital.Design: Observational cohort study using linked routinely collected data.Setting: England.Patients: Children aged 0-25 years who died between 2003 and 2017.Main Outcome Measures: Place of death: hospital, hospice, home. Multivariable multinomial logistic regression models.Results: 39 349 children died: 73% occurred in hospital, 6% in hospice and 16% at home. In the multivariable models compared with dying in a hospital: neonates were less likely, and those aged 1-10 years more likely, than those aged 28 days to <1 year to die in hospice. Children from all ethnic minority groups were significantly less likely to die in hospice, as were those in the most deprived group (RR 0.8, 95% CI 0.7 to 0.9). Those who died from 2008 were more likely than those who died earlier to die in a hospice.Children with cancer (RR 4.4, 95% CI 3.8 to 5.1), neurological (RR 2.0, 95% CI 1.7 to 2.3) or metabolic (RR 3.7, 95% CI 3.0 to 4.6) diagnoses were more likely than those with a congenital diagnosis to die in a hospice.Similar patterns were seen for clinical/demographic factors associated with home versus hospital deaths.Conclusions: Most children with a life-limiting condition continue to die in the hospital setting. Further research on preferences for place of death is needed especially in children with conditions other than cancer. Paediatric palliative care services should be funded adequately to enable equal access across all settings, diagnostic groups and geographical regions. [ABSTRACT FROM AUTHOR]