Criterion-based audit
The Donabedian framework for assessing quality of care by evaluating structure, process and outcome was applied to ensure a broader understanding of quality than simple birth outcomes18.
The audit criteria were selected from previous Tanzanian studies, which successfully applied them,15,16 and were modified to suit the MSF guidelines for maternal and newborn care and resemble locally agreed essential practices.17 Only indicators that were recorded and retrievable from case files were included. Structural resources were assessed by staff numbers, number of labouring women and available supplies, as well as by available background characteristics of the labouring women. Key criteria for the process of intrapartum care delivery included partograph use, intrapartum monitoring of maternal blood pressure, foetal heart rate and labour progress, use of oxytocin augmentation and indications for caesarean section. Birth outcome criteria included rates of stillbirths, intrafacility neonatal deaths and Apgar scores <7.
All case files were retrieved from the hospital storage where they were filed by month and mode of delivery. Case files of births before gestational age of 28 weeks were excluded. Birth outcomes and mode of delivery were assessed on all case files in the two months studied. However, due to limitations in time and resources, 250 case files of vaginal births were randomly selected from each month to assess the process of care delivery (case files were piled, the total number was divided by 250 and according to this number every third to fifth file was selected). In addition, all women giving birth by caesarean section were included from both study months to assess the indications for the surgeries. In accordance with the pre-selected audit criteria, the first author and a Yemeni research assistant extracted data from the case files, which were double-entered into a data collection form in Excel. Any practice not recorded was assumed not being done14.
For data analysis, the statistical software R was used. Frequencies and percentages were calculated for background characteristics of the labouring women, process of care delivery and outcome data. Comparisons between the two periods (high- and low-volume month) were done using Chi-square tests (Fisher’s exact test was used when any category cell count was less than five). Additionally, we calculated relative risks (RR) with 95% confidence intervals (CI) for process and outcome data, considering the exposure to be admission in the high-volume month (August 2017).