Background and aims High oxygen exposure has adverse long-term medical consequences in premature infants. We hypothesised that a quality improvement initiative decreasing oxygen saturation limits in NICU-admitted infants would affect neuropsychological outcomes at age 3.

Methods We studied 83 NICU-admitted participants born at 23–27 weeks; n = 37 (b.2009–2010) with targeted supplemental oxygen rates of 83–93% until 32 weeks, 85–95% until 35 weeks, and >95% after 35 weeks (O₂ Reduced); 8% laser eye surgery; 65% ROP compared with n = 46 (b.2007–2008) with higher target rates (89–100%; O₂ Higher); 28% laser eye surgery; 83% ROP. Group mortality rates (<10%) did not differ significantly, χ²(n = 83, df=1)= 0.16, p = 0.900. ANCOVAs and regressions were conducted for 19 neuropsychological outcomes.

Results O₂ Reduced performed significantly better than O₂ Higher in visual attention, controlling GA (p = 0.042; d=.47); O₂ Reduced was significantly correlated with visual attention (r = 0.288, p < .05); ROP was not (r=-0.183, p > 0.05). 18/19 outcome scores favoured O₂ Reduced, low-to-moderate effect sizes (d range:12–0.47). GA was positively related to all outcomes. A meaningful interaction effect between GA and O₂ Reduced for auditory attention and manual dexterity explained 3%-4% unique variance, respectively. Interaction showed O₂ Reduced’s effect on attention was greatest at 23–24 weeks, and at 27 weeks for manual dexterity.

Conclusions Systematic reduction of saturation parameters that decreased retinopathy and laser surgery rates in infancy resulted in favourable three-year-old outcomes suggesting cognitive as well as medical advantages for O₂ Reduced. Moderate effect sizes demonstrated important associations between reduced oxygen saturation rates and 18/19 sensitive performance-based measures. Lowered targeted oxygen saturation rate had favourable yet differential effects across GA in extremely premature survivors.