Page 208 - GHES-3-3

Global Health Economics and
Sustainability
Neonatal mortality in Pakistan

Table 1. Detailed definition of all variables that NM is a dichotomous variable, the logistic model
is an appropriate statistical technique for analyzing the
Variables Definition likelihood of survival or death in newborns based on
Dependent variable delivery location and other explanatory factors. Unlike
Neonatal mortality 1: If the neonatal died within 28 days of birth linear regression models, which assume a continuous
0: Otherwise dependent variable, logistic regression models are
Independent variables designed for binary outcomes. In this case, the dependent
Place of delivery It is a dichotomous variable: variable — NM — is classified as “1” if the child survives
1: If delivered within the facility beyond the neonatal period (first 28 days of life) and “0” if
0: Otherwise the child dies within this period. The model predicts the
Birth order Birth order is classified into four categories: probability of neonatal survival as a function of various
1: 1 independent variables, including place of delivery and
2: 2 – 3
3: 4 – 6 other sociodemographic factors.
4: 7 – 15 A key advantage of using logistic regression is the
Education The education of women respondents has four introduction of the odds ratio (OR), which quantifies
categories: the strength and direction of the association between
0: No education
1: Primary education independent variables (e.g., place of delivery) and the
2: Secondary education dependent variable (NM). The OR provides an intuitive
3: Higher education interpretation of risk: An OR >1 indicates that the variable
Employment Women’s employment status is categorized into increases the likelihood of NM, whereas an OR >1 suggests
two categories: a protective effect.
1: Employed
0: Unemployed Mathematically, the logistic model is represented as:
Size of child Size of the child at the time of birth is p = E(Y|X) = β β X (I)
categorized into three categories: i i 1+ 2 i
1: If the size of the child is small (<2.5 kg) where, pis the probability of success, E(Y|X) represents
i
i
2: If the size of the child is average (2.5 kg – 4 kg) the expected value of the dependent variable Y given the
3: If the size of the child is large (more than 4 kg) value of the independent variable X, whereas β is the
Sex of child The gender of the child is a categorical variable: intercept and β is the slope. 1
1: If girl 2
0: Otherwise 2.3. Logistic regression model
Region Region is categorized into five possible A large number of studies that use data from household
categories:
1: If the respondent lived in Punjab surveys seek to characterize, comprehend, and evaluate
2: If the respondent lived in Sindh the link between independent and outcome variables.
3: If the respondent lived in Khyber However, traditional estimation methods such as ordinary
Pakhtunkhwa least squares regression are inappropriate when the
4: If the respondent lived in Balochistan dependent variable is discrete and categorical. Categorical
5: If others (AJK, GB, and ICT) dependent variables, especially binary variables, need
Residence The residence of respondents is a categorical specific modeling strategies to ensure accurate estimation
variable measured in two categories:
0: Rural and meaningful interpretation, in contrast to continuous
1: Urban variables, which can have any value. Consequently, logit
Mother’s age The age of mothers is a continuous variable or probit models — which are frequently employed for
categorized into seven categories: analyzing binary outcomes — are frequently relied upon
1: 15 – 19 by researchers.
2: 20 – 24
3: 25 – 29 The binary logistic regression model, a member of the
4: 30 – 34 generalized linear model family, is one of the most widely
5: 35 – 39 used models for this purpose. It is particularly useful when
6: 40 – 44 the dependent variable has only two possible outcomes,
7: 45 – 49
Note: Variables are taken from the Pakistan Demographic and Health frequently categorized as “success” or “failure,” “yes” or
“no,” or “1” and “0.” This makes the model particularly
Survey 2017 – 2018.
Abbreviations: AJK: Azad Jammu Kashmir; GB: Gilgit Baltistan; applicable to the analysis of events such as healthcare
ICT: Islamabad Capital Territory. access (received vs. did not receive care), employment

Volume 3 Issue 3 (2025) 200 https://doi.org/10.36922/ghes.5089