Study Site
The study was conducted in Kembata Tembaro Zone which is located 357kms south of Addis Ababa, the capital of Ethiopia. During the study period, there were about 31,103 pregnant mothers residing in the zone. There were several health service facilities, including: 1 general hospital, 3 primary hospitals, 31 health canters and 136 health posts. Among the hospitals, only Shenshecho primary and Durame general hospitals have standardized neonatal intensive care unit.
Study Design
Institution based unmatched case control study design was conducted from October 1, 2018 to February1, 2019. Controls represent women who gave live full term births, i.e., births of babies with at least 37weeks and less than 43 weeks of gestational age during this study period in Shenshecho primary and Durame general hospitals in Kembata Tembaro Zone. Whereas the cases represent mothers who gave live preterm neonates, i.e., births of babies’ less than 37 and above 24 weeks of gestational age in the selected hospitals during the study period based on the following criteria (Julie-Anne Quinn, 2016).
The following criteria was used for identifying controls and cases irrespective of mode of delivery; Gestational age of neonates which was determined by LNMP, review of medical records for fundal height or ultra-sound of mothers in second trimester who gave live births were assigned as mothers who gave full term births or controls with pregnancy duration of >=37 weeks and less than 43 weeks irrespective of the neonates weight and as mothers who gave preterm births or cases with pregnancy duration of 25 to less than 37 weeks 1516.
Small or Large for gestational age with pregnancy duration of 25 to less than 37 weeks were also assigned as mothers who gave preterm births or cases17. Mothers who were referred to the selected hospitals for this study from nearby hospitals, health centres and health posts and gave live births during the study period assigned as cases and controls corresponding to their location based on the above criteria.
Source Population
All mothers who gave live births during the data collection period in Kembata Tembaro Zone, public hospitals having neonatal intensive care unit in Southern Ethiopia was considered as source population for both case and controls.
Study population
All selected mothers who gave live births during the data collection period at Shenshecho primary and Durame general hospitals were considered as a study population for both case and control.
Inclusion and Exclusion Criteria
Exclusion criteria:
- Mothers who gave live births during the data collection period but referred to other health institution for better service
- Mothers who had postpartum psychosis, unable to hearing and speaking that gave live births during the data collection period
- Mothers who gave more than one live births at a time or multiple pregnancy
Sample size determination
Sample size was estimated based on a previous finding of AOR of 2.22 and 17% of controls exposed for previous history of abortion (p) which were used to assess factors associated with preterm birth 18. So that by using 95% CI, 80% power, 1:3 case to control ratio, and Epi info version 7, a total of 394 mothers who gave live births with 99 cases and 295 controls were calculated as a minimum sample size required for the study. Finally, by considering 5% of non-response rate, the optimum sample size calculated was 414 mothers with 104 cases and 310 controls (Table 1).
Table 1: Sample size calculation of the study conducted in Kembata Tembaro Zone, Southern Ethiopia

Sampling Method
Durame General and Shenshecho Primary Hospitals which had the neonatal intensive care unit were purposively selected. Based on total live births of the hospitals over the past four months from review of delivery registry book we allocated the sample size for both selected hospitals by proportional allocation based on number of births in each hospital. Then, mothers who gave live preterm births (cases) and mothers who gave live full term births (controls) were included in the study, three controls for one case were consecutively included till the desired sample size of each hospital was attained. Total mothers who had live births over four months in both hospitals were 1312 of which 864 were in Durame General and 448 in Shenshecho Primary Hospitals (Figure 1).
Data Collectors
Data was collected by trained BSc nurses who were fluent speakers of Amharic language. They were given training on methods of data collection and pretest was done for consecutive five days prior to actual data collection.
Data Collection tools
Data was collected by using structured and pretested questionnaire from the selected mothers and checklists were used to collect data from medical records of preterm neonates. The questionnaires were first developed in English and then translated to Amharic language. The Amharic version was then back translated into English by another language expert to check for consistency of two versions. The questionnaires were adapted from EDHS 2011and FAO that was developed for DHS project to reflect the population and health issues relevant to Ethiopia19.
Anthropometric measurement
Mid upper arm circumference (MUAC) of mothers who gave live births was also taken with level of accuracy nearer to 0.1cm. Maternal body mass index (BMI) was calculated from measurements of height and weight with the accuracy nearer to 0.1cm and 0.1Kg, respectively after delivery. Height and weight of neonates were taken immediately after birth with new born measuring scale. Duplicate measurements were performed with the same measuring instrument and measurer with calibration of measurement instrument in order to enhance precision and accuracy of required value.
Food frequency questions
Dietary habit of the mothers was assessed using food frequency questionnaire. The food frequency questions were developed based on the studied population dietary intake habit and commonly consumed food items with respect to their culture. Thirty-four food items were included in the questionnaire and aggregated them into nine food groups; 1) cereals and tubers; 2) pulses; 3) organ vegetables; 4) Organ fruits; 5) meat and fish; 6) poultry and eggs; 7) milk and milk products; 8) other fruits and 9) other vegetables.
These were collected from mothers who gave live births during the study period about their past twelve months’ dietary frequency intake. Recall bias was tried to overcome by using corresponding food pictures included their products as much as possible while data collecting. Then, this was converting to the standard seven days dietary frequency intake by using conversion factors and summed up 2021. Next, by using standard food groups’ weight for each, we calculated respondents’ household food consumption score 21. Finally, based on food consumption score threshold as poor (0-21.4), borderline (21.5-35) and acceptable (>35), we measured the association of preterm birth with caloric intake and dietary quality using logistic regression (Table 2).
Table 2: Food groups and their weight for food consumption score, Kembata Tembaro Zone Public hospitals, 2019.

Adopted from food security indicator for integrating nutrition and food security programming for emergency response workshop 21.
Estimation of gestational age (GA)
Estimation of gestational age could be carried out based on different methods such as menstrual periods, date of conception, fetal ultra-sound and physical parameters using the new Ballard score 16. According to this study almost 87% of gestational ages of live births were determined by using normal last menstrual period and aiding local events in order to increase recall ability of mothers in addition to review medical records. Eleven percent of live births gestational ages were estimated assisted by first or second try-minister pregnancy fetal ultra-sound. The rest 2% were determined by new methods of Ballard score for neonates of mothers who had no ultra-sound and couldn’t recall their last normal menstrual period by any means. The Ballard score is based on the neonate's physical and neuromuscular maturity and can be used up to 4 days after birth (in practice, the Ballard score is usually used within 24 hours after delivery). The neuromuscular components are more consistent over time because the physical components mature quickly after birth. However, the neuromuscular components can be affected by illness and drugs such as magnesium sulfate given during labor. Because the Ballard score is accurate only within plus or minus 2 weeks, it should be used to assign gestational age only when there is no reliable obstetrical information.
Study variables
Dependent Variable
Independent Variables
- Identification of the preterm birth:
- Maternal socio-demographic characteristics:
- Gynaecologic-obstetric related factors:
- Medical disorders & infection in mother:
- Maternal physical factors:
- Nutrition and life style factors
Data Quality Control
The adopted and developed tool was evaluated by experienced researchers. Pretest was employed on 5% of the sample size with structured questionnaire in Durame General Hospital two weeks prior to the actual study to check quality of data collector, questioner and usually recorded variables on the patient’s folder. Daily evaluation of the data for completeness and encountered difficulties on the time of data collection was attended accordingly.
Data processing and analysis
Completeness of the questionnaire was rechecked preceding data entry. Following this, data were coded, entered, cleaned, recoded and analysis was accomplished by using IBM SPSS version 20. Chi-square test and bivariate logistic regression analysis was done after dichotomizing the dependent variables by coding with ‘0’for full term birth and ‘1’ for preterm birth to measure association of independent variables with outcome variable. After checking associations of the variables, those with p<0.25 in the bivariate model were selected for multivariate logistic regression analysis using enter method to control for confounding factors in the process of evaluating the strength of association. P-value of <0.05 was used to declare statistical significance of the variables. Text, table of frequency and graphs were used to present result of the study.