Background: Our experience has suggested that not all the drug dose put into an IV infusion reached the bloodstream of neonates. Aim: To use sparse-data modelling to estimate how much of a continuously infused dose of caffeine, gentamicin, phenobarbitone and vancomycin was delivered. To use the results to develop a modified infusion protocol to minimise drug losses. Method: Various drug infusions were prepared in duplicate (test, control). The test set was delivered through an in vitro infusion system identical to that used in the neonatal intensive care nursery. Samples were collected after flushing the system twice with 1.5 mL of normal saline. Each experiment was replicated 4 times for each drug. Drug concentrations were measured by enzyme immunoassay. Recovery and variability during flushing was described as a function of the normal saline flushing volume using an asymptotic mono-exponential accumulation model. Results: 53% of phenobarbitone, 60% of caffeine, 92% of gentamicin, and 99% of vancomycin doses were recovered. There was considerable variability in the recovery of phenobarbitone. Conclusion: Current infusion protocols appear inadequate for delivery of drugs to neonates and may result in suboptimal therapy. It is recommended that the administration set be completely primed with the drug and any additional volume of normal saline required, then flushed with a further 2 mL of normal saline to obtain a minimum 90% recovery of these drugs.