Statistical analysis
First, we studied trends over time in indicators of birth asphyxia. We calculated the number and rate of births with the outcome measurements for each year of the study. To analyse possible trends over time, we performed Cochrane-Armitage trend tests. The absolute difference was the outcome in year 2019 minus 2010. The relative difference (in percentage) was the absolute difference divided by the outcome in year 2010.
Second, we investigated risk factors for a low 5-minute Apgar score, both <7 and <4, by calculating risk per 100 births and performing logistic regression analysis. We assessed crude and adjusted odds ratios (ORs) with 95% confidence intervals (CI). The Population Attributable Risk (PAR), which takes the number of exposed individuals into account, was calculated for the variables with statistically significant ORs for Apgar score, both <7 and <4.
Thirdly, we studied trends in prespecified obstetric interventions (induction of labour, use of epidural analgesia, planned and emergency caesarean section, and instrumental delivery) and in level of care. In subgroups based on these obstetric interventions, we also separately assessed trends in low Apgar scores.
Finally, we performed a multivariate logistic regression analysis to adjust for the co-variates. Collinearity diagnostics was performed to find possible interaction of factors. Data of confounders were missing in less than 0.5% of the cases, except for arterial umbilical cord blood pH’s. Arterial umbilical cord blood pH’s are not routinely collected and were available in 9.6% of all infants. Missing data of co-variates were imputed. All analyses were complete case analyses. All statistical analyses were carried out with SAS software version 9.4 (SAS Institute Cary, North Carolina USA). A p- value of less than 0.05 was considered statistically significant.