A retrospective study was done to assess the effect on in-hospital neonatal mortality of a series of interventions in neonatal care in the highlands of Papua New Guinea. Between 1995 and 1997, prior to the interventions, the mortality among neonates admitted to the Goroka Hospital Special Care Nursery was 18% and two-thirds of very low birthweight (1-1.5 kg) neonates died. The interventions began in December 1997 and were aimed at reducing mortality among all neonates and particularly among those with very low birthweight. Compared to the 30-month period prior to the interventions, the in-hospital neonatal mortality in the 30-month period after the interventions began was 44% lower (relative risk (RR) 0.56, 95% confidence interval (CI) 0.45-0.69). After adjustment for a higher number of neonates <1500 g in the pre-intervention period, the relative risk was 0.59 (0.48-0.74). The mortality in the intervention phase for very low birthweight babies was 56% lower (RR 0.44, 95% CI 0.30-0.65) and for moderate low birthweight (1.5-2 kg) 50% lower (RR 0.50, 95% CI 0.28-0.90). Mortality was also significantly lower in the intervention phase in neonates with a diagnosis of septicaemia or pneumonia (RR 0.36, 95% CI 0.19-0.67), but there were no differences in mortality from birth asphyxia, meconium aspiration or extremely low birthweight (<1 kg). We estimate that in the 30 months after beginning the interventions 82 neonatal deaths that would previously have occurred were avoided. The costs of the improvements in technology described are estimated at US$445 (K1000) per life saved, but substantial training and improved supervision of staff and other human factors may have been more important than equipment. Apnoea monitors were the single most important technology introduced. A similar evaluation of the effect of minimal standards should be done without the use of incubators and overhead heaters, as these are costly and may be dangerous when used by less experienced operators. The 33 neonatal deaths that we estimate were avoided each year because of the interventions represent less than 10% of all neonatal deaths in the province. Although this study provides justification for increasing the technology for supportive neonatal care and training in medium-sized hospitals in rural areas in developing countries, estimates of cost-effectiveness must be compared with other interventions that will effectively lower neonatal mortality, both in and out of hospitals.