A facility-based retrospective cross-sectional study was conducted from 1 February to 30 April 2018, in the Tikur Anbessa Specialized Hospital found in Addis Ababa the capital city of Ethiopia. Tikur Anbessa is Ethiopia's largest specialized and referral public hospital. According to the 2007 statistical report of the population and housing census of Ethiopia, Addis Ababa has a total population of 3,384,569 (17). The health service coverage of the city is 52.2, and 82% of deliveries take place in public health facilities (18). There are 17 public and 25 private hospitals (19). The hospital provides diagnoses and treatment for approximately 370,000–400,000 patients per year in all wards. The diabetes center is one of the departments in the hospital services. Of a total of 800 beds, 80 were used by the department of obstetrics and gynecology during the survey (20). In the city, around 4,600 deliveries were attended each year of which 60% are operative deliveries (8).

Sample size was determined using a single population proportion formula. The following assumptions were used while calculating the sample size; 95% confidence level, a margin of error (0.05) and 50% anticipated population proportion was taken since there is no published paper on assessing cesarean-section among mothers with diabetes in Ethiopia. Therefore, 422 sample size was planned to use for this study with a 10% non-respondent rate. However, since the numbers of diabetic mothers' cards with complete records were less than the planned sample size, all cards with complete records were included.

Tikur Anbessa Specialized Hospital was selected purposively since it is the largest public hospital with maternal health service and diabetes center/department. The required data were extracted from patient charts. First, the health management and information system (HMIS) delivery registration book, postnatal registrations, admission registration, and gestational diabetes registration in diabetes mellitus center card numbers were obtained and documents of all delivered mothers who had DM during the planned study period at obstetrics ward was searched and checked for completeness of the data. Then, cards of mothers with complete records were separated and counted. Since the numbers of cards with complete records were less than the planned sample size, all cards of the diabetic mothers with completed records were included whereas incomplete documents were excluded.

A structured checklist adapted from published studies with certain modifications was used (16, 21–23). The checklist includes information on socio-demographic characteristics, obstetric characteristics, feto-maternal outcome, and diabetes mellitus. After the card numbers were obtained, the documents of all delivered mothers who had DM during the planned study period at the obstetrics and gynecologic ward were searched and checked for completeness of the data. The data were collected through document review in the obstetrics ward from the patient chart, delivery registration book, post-natal registrations, duty report registration books, and operation logbooks. The data were extracted by two record office staff, three midwives working in the obstetrics unit, two supervisors, and the principal investigator.

To ensure the quality of the data before the actual data collection, a pretest was done on 5% of patient records at Zewditu Memorial Hospital. Appropriate modifications were made to the checklist and procedures after analyzing the pretest result. One-day training was given to data collectors on how to collect the data. The supervisors and the principal investigator coordinated and checked the data collection process, and daily supervision was done to ensure the completeness and consistency of the gathered information.

The collected data was entered to Epi data version 4.2, cleaned, and transported to SPSS version 23.0 for data analysis. Descriptive statistics, frequency tables, figures, and percentages were used to summarize the data. Bi-variable analysis and multiple logistic regressions were used to test for the association between dependent and independent variables. Variables that showed an association in the bi-variable analysis with p-value < 0.25 were entered into a multiple logistic regression model. At last, the multiple logistic regression models were used after controlling for confounding factors using regression. Hosmer-Lemeshow's goodness of fit test was used to assess whether the necessary assumptions for the application of multiple logistic regression had been fulfilled. Multi-co-linearity was assessed by using standard error, and the variables were entered into the multiple models without multi-co-linearity. Adjusted odds ratios (AOR) were calculated with 95% CIs, and statistical significance was declared at p-value < 0.05.

Ethical clearance was obtained from the ethical review committee of Addis Ababa University, School of Nursing and Midwifery. The official letter of cooperation was taken from Addis Ababa University to Tikur Anbessa Specialized Hospital. Informed and written consent was obtained from the medical director and NICU head. To keep confidentiality, their patients were not documented; rather a code was given for each card.

In this study, a total of 346 records were reviewed. Of a total of the respondents, the majority (40.2%) were in the 30–34 years age group. Almost all (98.6%) mothers were married. Addis Ababa was the dominant residence of the mothers, accounting for 93.1%. Regarding the occupational status of the mothers, almost half of them (50.2%) were housewives (Table 1).

Regarding the parity of diabetic women, the majority (80.9%) were multipara. Nearly one out of five, (20.8%) of mothers had a history of abortion and CS (20.2%), respectively. Around (13.6%) of mothers had experienced a stillbirth in their previous pregnancy (Table 2).

Nearly half, (51.2%) of the labors occurred spontaneously and (30.3%) were elective CS. Pregnancy-induced hypertension (PIH) accounts for the major parts of the complications the mother developed from pregnancy to postpartum period (26.0%). The majority, (82.1%) of mothers gave birth at term, and 97.4% of delivered newborns were alive (Table 3).

Regarding mode of delivery, the majority (57.8%) of mothers gave birth by CS whereas the rest (42.2%) of the mothers gave childbirth by spontaneous vaginal delivery (Figure 1).

In the bivariate logistic regression analysis, marital status, occupational status, residence, history of stillbirth, parity, history of abortion, PIH, birth weight, and fetal distress were factors significantly associated with CS. However, in multiple logistic regression mothers who have PIH were three times [AOR: 3.35, (95% CI: (1.22, 9.20)] more likely to undergo CS. Also, mothers who have a history of CS were 1.62 times [AOR: 1.62, (95% CI: (2.54, 4.83)] more likely to undergo CS than their counterparts. Mothers who have fetal distress were nearly four times [AOR: 4.36, (95% CI: 1.30, 14.62)] more likely to give birth by CS (Table 4).