81 Regional variations in childbirth interventions and their correlations age, ethnic background, socioeconomic position and degree of urbanisation did not lead to considerable changes in ORs. Therefore, it is unlikely that regional variation in underlying morbidity, including variation in indications for interventions, would explain entirely the variations observed. Reporting bias is an issue for datasets based on national registers (30) but we have no indication that misclassifications are different across regions. The number of missing values was very low; there were no outcome variables with more than 1.5% of values missing, and no characteristic variables with more than 2.5% of values missing. Additionally, the correlation coefficients calculated are only a crude indicator of the relevant and significant correlations between variables, since it was not possible to account for confidence intervals by calculating Spearman’s correlation coefficients of adjusted ORs. A major strength of this study was access to data of the total population of women in the Netherlands who gave birth in the four-year study period. Furthermore, we were able to differentiate between outcomes for women receiving midwife-led and obstetrician-led care, at the onset of labour and at the time of birth. In this manner, the distinction could be made between groups of women at a low and those at a higher risk of complications. This makes confounding by indication less likely. While not all twelve regions have a university hospital, the analyses were based on the women’s residential postal code instead of the postal code of the place of birth, thus the presence or absence of a university hospital in a region will have had a limited influence on the results. However, other confounders, such as distance to a hospital, may have had an influence upon the outcomes and may explain, in part, some variations. Correlations which were investigated in this study cannot be interpreted as causal relationships. Further research is needed to examine whether causality exists for the associations found. Interpretation and further research Interpretation in view of previous literature Previous studies on regional variations in other countries, showed a variety in intervention rates across studies. When focusing on the subjects of our study, we found lower rates and less variation in rates of AROM, and, similar to our results, large variations in episiotomy rates. Studies were found on regional variations in episiotomy and AROM in Brazil, Ireland, Canada, and France, but there is a lack of literature on regional variations in care processes. A Brazilian study, including 23,940 women, found rates of AROM varying from between 32% to 48%, and episiotomy rates varying from 49% to 69% across five regions, although variation in episiotomy rates was only significantly different in the region with the highest incidence (32). The episiotomy rate in a study in Ireland, including 323,588 births, varied from between 19% to 27% across four regions, and the rate of AROM varied from between 5% to 9% (33). A study on variation across 13 regions in Canada, including 8,244 3
RkJQdWJsaXNoZXIy MTk4NDMw