Caesarean delivery and its correlates in Northern Region of Bangladesh: application of logistic regression and cox proportional hazard model

Study area

The study sample comprised of 1142 women who had delivery either by caesarean or non-caesarean
delivery at four private and four public hospitals maternity wards in the northern
region of Bangladesh during the period of January to March 2010. Among the 1142 delivery
cases, 652 were caesarean and the remaining 490 were non-caesarean. The northern region
is the part of north in Bangladesh, where the hospitals were situated. The hospitals
involved in the study are Islamic bank hospital, Shapla, Rangpur and city clinic which
are the private hospitals, while Rajshahi, Bogra, Rangpur and Dinajpur medical college
hospitals are expressed as public hospitals. Also, the terms private and public patients
refer to respondents, who were admitted in maternity wards for safe delivery in the
respective hospitals.

Study population

Pregnant women in the Northern region of Bangladesh.

Sampling design

The study followed a cross-sectional design where data were collected by direct interviews.
Before delivery, the participants were selected by simple random sampling and proportion
to the estimated load of deliveries, which accounted for 60 % of all deliveries during
the survey period. Most of the questions were close-ended and the answers chosen by
the respondents were indicated by the tic mark. The response rate was 100 %.

Measurement of variables

Dependent variables

The dependent variables considered as (i) the types of delivery coded as dichotomous
(caesarean?=?1, non-caesarean?=?0) and (ii) duration of time (that is, after marriage
to getting one child or from previous child to current child) to event (that is, mode
of delivery).

Independent variables

The maternal variables included prolonged labour (more than 12 h), fetal distress
(it is commonly used to describe fetal hypoxia that is low oxygen levels in the fetus,
which can result in fetal damage if the fetus is not promptly delivered), previous
c-section, breathing difficulty, child aborted around delivery, multiple births; head
circumference, length and weight of babies.

For the analysis of data, the category related to prolonged labour, fetal distress,
previous C-section, breathing difficulty, child aborted around delivery and multiple
births were assessed as yes or no. The head circumference of newborns was classified
into two categories: 32 cm and more than 32 cm. The length and weight of baby were
categorized into: 45 cm or more than 45 cm and 2.5 kg or more than 2.5 kg respectively.
The socio-demographic variables included maternal age at birth, age at marriage, parity
(order of birth), and maternal educational level. Maternal age was categorized into
four broad groups (years): 20, 20–24, 25–29 and more than 30. The age at marriage
was classified into three categories: 18 years, 18–22 years and 23 years and above.
The parity was divided into three groups: 1, 2, and???3. Education status is the highest
level of schooling attained, measured as primary and below (0–5 years), secondary
(6–10 years) and higher (11 years and above). Place of residence and duration of taking
balance diet (it refers to milk, fish, egg, fruit and vegetables that contains adequate
amounts of all the necessary nutrients required for healthy growth and activity and
those diets were taken a woman in pregnancy period) were also considered as the other
related variables in the study. Additionally, place of residence was classified as
rural verses urban and duration of taking balance diet was measured as a categorical
variable: often, once a week and rarely.

Statistical analysis

An initial bivariate analysis was performed to identify significant associations between
types of delivery (caesarean vs. non-caesarean) and a series of independent variables.
Dichotomous variables were analysed by the ?²
test or Fisher exact test, where appropriate. To determine the risk factors which
are associated with the C-section, based on the different criteria, two multivariate
techniques were used. They are logistic regression model and Cox proportional hazard
model. Logistic Regression and Cox proportional hazard models are the most frequently
used for analysing data in epidemiological and clinical studies 38]. The logistic regression is analogous to multiple linear regressions where the dependent
measure is dichotomous in nature (coded by the values 0 and 1); whereas the Cox proportional
regression model assumes that the effects of the predictor variables (names of variables
that we expect to predict survival time) are constant over time. For both techniques,
maternal, socio-demographic and other relevant variables were treated as independent
variables, while the dependent variables were already mentioned in the above section.
The most influential risk factors were estimated separately for overall, public and
private hospital by stepwise selection. The value of P??0.05 was considered statistically significant. Finally, to identify and measure
the risk factors for caesarean delivery, that is, how well the model fits the data,
Akaike Information Criterion (AIC) is used. Generally, the AIC formula is ?2 log(L)?+?2 k,
where, L is the maximized value of the likelihood function for the estimated model
and k is the number of parameters in the statistical model. Lower AIC indicates a
better likelihood.

Ethical clearance

We obtained informed verbal consent from the respondents before conducting the interview.

The study was approved by the ethical board and research review committee of the Dept.
of Population Science Human Resource Development, University of Rajshahi, Bangladesh.