We implemented the Robson ten group classification system and assessed the proportion of each group in the obstetric population, the contribution of CS in each group to the overall CS rate and the CS rate within each group.
In our study, Group 1 and Group 3 represented the two largest groups presenting for labor and delivery. This finding is consistent with a study done in India where Group 1 and Group 3 contributed to 24.2% and 19.4% of all deliveries respectively . Similarly, studies done in Brazil, Italy, and Tanzania showed Group 3 and Group 1 were the two most represented obstetric groups [13–15].
Group 10 was found to be the third-largest obstetric group. The contribution of this group to the overall CS rate depends on its size . As such, Group 10 made the highest contribution to the CS rate accounting for nearly one in five CS deliveries. This is in sharp contrast with a study done at a university hospital in Eastern Ethiopia, where Group 10 was the 6th place contributor to the overall CS rate, accounting for 6% of CS deliveries . This variation can be explained by the significant difference in the obstetric population served by the two facilities. Our study was done in a tertiary referral hospital with a dedicated maternal-fetal medicine unit. A significant proportion of care is given to mothers with major obstetric and medical comorbidities who may require interventions, increasing the likelihood of iatrogenic prematurity. This can account for the higher proportion of Group 10 and its contribution to the CS rate in our setting. In fact, our finding is consistent with studies conducted in other tertiary care facilities [7, 8]. A study from a tertiary unit in Italy also showed Group 10 was the second largest contributor to the CS rate .
About 20% of women in our study did not have a milestone to ascertain gestational age. This is not unusual as the national guideline does not include routine dating ultrasound during antenatal care follow up  and has not yet adopted the WHO recommendation of at least one ultrasound scan before 24 weeks of gestation . In this study, birth weight was used as an indirect estimate of gestational age for those women we could not ascertain gestational age. This adaptation has also been used in other studies that implemented the RTGCS in low-resource settings. For example, Tura AK et al  and Schantz C et al  used birth weight of ≥ 2,500 g as proxy to term when data on gestational age was not available whereas other studies applied the method to their entire dataset and defined gestational age solely based on birth weight [14, 20, 24].
Using neonatal birth weight as a proxy indicator of gestational age can however result in misclassification of growth-restricted newborns as preterm and potentially increase the relative proportion of Group 10. To test for this, we performed a separate analysis after excluding the 839 women with unknown gestational age. However, there was no change to the relative proportion of Group 10 and its contribution to the CS rate, and the group remained the third largest group and the leading contributor to the overall CS rate.
In addition, the CS rate within Group 10 was found to be 41%, which indicates the possibility of a high rate of pre-labor CS (Table 3) . Subgroup analysis shows one-third of women in this group underwent CS before the onset of labor. Further examination of the indications for CS can help us understand and design tailored interventions to reduce the CS rate in this group. This is especially relevant as interventions to reduce the CS rate within this particular group were found to be successful in other facilities .
Group 2 and Group 4 were also important contributors to the overall CS rate, accounting for one-third of CS deliveries. The CS rate within each group was also about 70%. Existing evidence suggests a high pre-labor CS rate at a particular institution if the CS rate within Group 2 and Group 4 is more than 35% and 20%, respectively . Subdividing these groups into induced labor and CS before labor provides useful information regarding the proportion of pre-labor CS and the success of induction (Table 3). This is particularly important as women in these groups are considered low risk. Our subgroup analysis showed a large proportion of women in both groups underwent pre-labor CS. This calls for further investigation of the indications for pre-labor CS.
Several studies across different settings identified Group 5 as the leading contributor to the CS rate [13, 14, 16]. In our study, Group 5 was the third-largest contributor to the overall CS rate and its relative size to the obstetric population was less than 10%. These findings are suggestive of relatively low CS rate in the previous years . Indeed, in 2016 the national institutional and population-based CS rates were 4% and 2.7% respectively .
Though the safety and long-term benefits of vaginal birth after cesarean (VBAC) are well established, 62% of women in Group 5 underwent repeat CS (Table 3). Thus, there is a need to evaluate the proportion of women who were offered a trial of labor and the success rate of VBAC. This will enable the design and implementation of antenatal counseling strategies and labor management protocols, reducing the number of repeat CS.
The contribution of Group 3 to the overall CS rate was small. However, the high rate of CS within this group is an alarming finding. The group represents low-risk women and the CS rate within this group is not expected to be higher than 3%. Auditing this group is an effective means to assess how an institution manages labor . Thus, evaluation of labor management protocols in our institution is warranted.
Since the publication of the term breech trial, there is a global shift towards CS among women with breech presentation [25, 26]. Consequently, several studies showed a high rate of CS in Group 6 and 7 [13, 15]. However, in our study, nearly half of breech presentations both in nulliparous and multiparous women were delivered vaginally (Table 3). This is similar to the 40% vaginal breech delivery observed in another teaching hospital in southwest Ethiopia . A more liberal national and institutional protocols that allow assisted vaginal breech delivery in selected women can explain these observations . Though we find this practice encouraging, further analysis should be done to assess maternal and perinatal outcomes among these groups.
The strengths of this study include the large sample size and availability of complete data for analysis. The results of this study can serve as baseline data to monitor trends of CS rate over time in our institution, as well as to compare our practice with that in other institutions. We also believe results from this study can be generalized to other tertiary facilities within the country and to some extent within other low-income countries.
This study also has some limitations. Our definition of fetal viability based on gestational age of 28 weeks or birth weight of ≥ 1,000 g may affect the rate of CS and the relative size of Robson’s groups. This in turn can impact the generalizability of our findings to other countries. Findings from RTGCS are only a starting point and should be viewed as a means, not an end. We now have a clear insight about "who" is having CS but not "why" the CS is being performed. Crucial variables such as indications, maternal and perinatal outcomes, are not incorporated, limiting the extent to which conclusions can be drawn from our study.