Aims We aimed to identify determinants of variation in central line use between neonatal units in England.

Methods We obtained data from the National Neonatal Research Database at Imperial College London for 112 neonatal intensive care and high dependency care units in England for the period 2010 to 2015, as part of the PREVAIL Generalisability Study (ISRCTN: 81931394). The data contained daily care records, including the presence of central lines, and birth details. We defined central line days as days when a peripherally inserted cutaneous long line, surgical central venous catheter, umbilical venous catheter or umbilical arterial catheter was present. We used multi-level Poisson regression to model line days per admission, adjusting for gestation, birthweight, age at admission, surgery and days on respiratory support.

Results Of 62 948 admissions with gestation <32 weeks, 58 550 (93.3%) had a central line compared with 244,916/368,065 (66.5%) admissions with a gestation 32 weeks. The median line days per admission by unit ranged from 1 to 16 for babies with gestation <32 weeks and from 0 to 5 days for babies with gestation 32 weeks. Fewer line days were associated with later gestation (incidence rate-ratio (IRR) for each week of gestation 0.90; 95% confidence interval (CI) 0.89 to 0.90) and older age at admission (IRR for each day of age 0.99; 95% CI 0.99 to 0.99). Longer line duration was associated with surgery (IRR 2.27; 95% CI 2.24 to 2.30) and days on respiratory support (IRR 1.02; 95% CI 1.02 to 1.02). Multi-level Poisson regression demonstrated significant variation in line days between units (0.66), of which 93% could be attributed to differences between babies.

Conclusion Nearly all variation in line days between units was explained by differences in baby case-mix. We suggest when using line days as a denominator for comparing rates of bloodstream infection between neonatal units, case mix should be taken into account.