Data
The present study uses the latest round of Demographic Health Survey (DHS) data from Pakistan 2012-13 (PDHS-3). DHS is considered a nationally representative sample as it covers samples from across the country with a well-specified sampling procedure. All the DHS uses multi-stage stratified sampling for sample selection. Survey collected information on reproductive and child health, family planning, fertility, water and sanitation, nutrition, lifestyle, violence, and other topics using the prescribed format of the questionnaire by DHS with some country-specific modifications.
The sample size of households interviewed was; 12,943 households with 13,558 eligible women and 3,134 men interviewed in Pakistan. For the current study, only household and eligible women's schedules will be used for the analysis.
In the survey, the anthropometric parameters (height and weight) of women and children were also collected. In Pakistan, PDHS measured height and weight of all children aged less than 5 years, however, for women aged 15-49 years, height and weight were measured in every third household selected for male interview.
Variable description:
Dependent variable: The present study is divided into two sections: nutritional status of women aged 15-49 years and nutritional status of children aged less than 5 years in Pakistan. The nutritional status of women is examined through the body mass index (BMI). It is categorised into four categories: Under-weight, Normal weight, Overweight and Obese. However, the nutritional status of children is examined through Z-scores of three parameters: Stunting, Wasting, and Under-weight & Over-weight for age. Stunting is categorised into two categories (Severe and Moderate); Wasting is categorised into three categories (Severe, Moderate, Over-weight); Weight for age is categorised into three categories (Severe underweight, Moderate underweight, and Overweight).
Women nutritional status: Cut-off limit for BMI (Weight for Height:
Underweight: BMI<18.5 kg/m2, Normal weight: BMI>=18.5 & <24.9 kg/m2, Overweight: BMI>=30.0 & <29.9 kg/m2, Obese: BMI>=30.0 kg/m2
Children nutritional status: Cut-off limit for Stunting (Height-for-Age):
Severely Stunted: Z-score <-3.0 SD below mean
Moderately Stunted: Z-score <-2.0SD below mean
Cut-off limit for Wasting (Weight-for-Height):
Severely Wasted: Z-score <-3.0 SD below mean
Moderately Wasted: Z-score <-2.0 SD below mean
Overweight: Z-score > +2.0 SD below mean
Cut-off limit for Weight-for-age:
Severely underweight: Z-score < -3.0 SD below mean
Moderately underweight: Z-score < -2.0 SD below mean
Overweight: Z-score > +2.0 SD below mean
Independent variables: Socio-economic and demographic characteristics of women and children are considered to understand the nutritional status by selected background characteristics. The selected socio-economic and demographic characteristics include: the place of residence (Rural, Urban), religion (Hindu, Muslim, Others), age (continuous), marital status (Married, Unmarried, Others), educational attainment, working status (Yes, No), Source of drinking water, type of cooking fuel, type of toilet facility, and wealth index. Apart from the above-mentioned variables, the number of children and dietary pattern (food composition) of women are considered. For children, the number of siblings, sex, breastfeeding pattern, immunization, and dietary pattern are considered.
Methods
Descriptive statistics and bivariate analysis have been used to understand the nutritional status among women age 15-49 and children aged 0-59 months in Pakistan.
To determine factors associated with Body Mass Index among women, multinomial logistic regression model was used. This allowed us to assess the independent effect of background characteristics in determining the prevalence of BMI. Multinomial logistic regression is an expansion of logistic regression in which one equation is set up for each logit relative to the reference outcome. BMI consist of four categories: normal, underweight, overweight, and obese. For a dependent variable with four categories, this requires the estimation of three equations, one for each category relative to the reference category (not related), to describe the relationship between the dependent and the independents variables:
ln [{P(Yi = 2)|Xi}/{P(Yi = 1)|Xi}] = α2 + β12X1 . . . βk2Xik ………1
ln [{P(Yi = 3)|Xi}/{P(Yi = 1)|Xi}] = α3 + β13X1 . . . βk3Xik …………..2
ln [{P(Yi = 4)|Xi}/{P(Yi = 1)|Xi}] = α4 + β14X1 . . . βk4Xik……..3
Where α2, α3, and α4 are the intercepts for the category underweight, overweight, and obese, respectively, and βk2, βk3, and βk4 are the slope coefficient of the Xi variables for respective category of the dependent variable.
We also used binary logistic regression to determine the factors associated with severe and moderate stunting, severe and moderate wasting, and severe and moderate underweight among children aged 0-59 months. In this analysis, the response variable ‘no’ was recoded as 0 if the child was not malnourished and, and 1 if with child was malnourished:
loge [P(Yi = 1| Xi) / 1 – P(Yi = 1| Xi)] = = α + β1Xi1, . . . , βk Xik …..4
Where Yi is the binary response variable; Xi is the set of explanatory variables, such as sociodemographic characteristics as mentioned in case of multinomial model; and β1, . . . , βk are the coefficients of the Xi variables.