Geospatial inequalities in women’s malnutrition in Pakistan

Abstract

Background

In developing countries, regional disparities in maternal malnutrition are a major deterrent to development. Inadequate nutrition and poor health among women not only affect their quality of life but also the well-being of their children, risking the future generation of the country. This study examines the spatial distribution of malnutrition at the extreme quantiles of Body Mass Index—severe thinness and underweight at the lower quantile and over-weight and obese at the upper quantile— and associated risk factors among women in Pakistan using Bayesian additive quantile regression.

Methods

A sample of 5,252 of the currently non-pregnant and non-lactating married women aged 15–49 was taken from Pakistan Demographic and Health Survey 2017–18. The response variable was the women’s nutritional status measured in body mass index (weight in kilograms/height in meters squared) of women. Following WHO guidelines, we used four indicators of BMI, as follows: severe thinness (BMI < 16 kg/m2); underweight (BMI < 18.5 kg/m2); Overweight (BMI > 24 kg/m2); and obese (BMI >  = 30 kg/m2). A set of explanatory variables comprising women’s characteristics and household related variables were used to assess their association with the likelihood of various forms of malnutrition. The structured Bayesian Geo-additive Quantile regression approach was employed to examine the association of the explanatory variables with the entire conditional distribution of the response variable.

Results

The sizable regional variation was found in malnutrition among reproductive age women. Women living in urban areas are more likely to become overweight (mean: 0.3; 95% CI: 0.06, 0.58) than their rural counterparts. Working women are less prone to obesity (mean: -0.51; 95% CI: -0.79, -0.23). Women with unimproved toilet are more likely to become overweight (mean: 0.7; 95%CI: 0.34., 1.04) and obese (mean: 0.90; 95%CI: 0.48, 1.33).

Conclusions

Findings underscore the need for targeted interventions to address the complex and varied challenges posed by women’s malnutrition.

Background

Malnutrition, as defined by the World Health Organization (WHO), involves deficiencies, excesses, or imbalances in the intake of energy and nutrients. It encompasses two main categories of conditions. The first is undernutrition, which includes stunting (inadequate height for age), wasting (inadequate weight for height), underweight (inadequate weight for age), and deficiencies in essential vitamins and minerals. The second category includes overweight and obesity, as well as diet-related non-communicable diseases such as heart disease, stroke, diabetes, and cancer. The prevalence of malnutrition among women and children is a major deterrent to the development and well-being of an individual imagn developing countries. Women and children are more vulnerable to nutritional deficiencies due to their unique nutritional requirement [1]. The prevalence of malnutrition among women has been found ranging 10% to 40% in most Low- Middle-Income Countries (LMICs) [2].

Malnutrition remains a primary concern, particularly among reproductive age women in Pakistan, demanding on special attention by researchers and policymakers. A strong association was observed between a mother’s nutritional status and pregnancy outcomes such as low birth weight, infections, growth challenges, and child development delays. Various factors contribute to malnutrition among women, such as women of reproductive age are most vulnerable to low dietary intake, unequal household food distribution across regions, inappropriate and improper food storage and preparation, dietary taboos, prevalence of infectious diseases, and inadequate health care facilities [3]. Sociodemographic factors such as place of residence, literacy, wealth, sanitation, and source of drinking water are also associated with maternal malnutrition in developing countries [4,5,6]. In the urban women’s population, the exposure to a sedentary lifestyle such as eating junk food and lacking physical exercise are some risk factors for overweight or obesity [4].

In Pakistan, about 46% women of reproductive age had normal BMI, 14.5% were underweight, 24% were overweight, and 13.9% were obese, as per estimates by the National Nutrition Survey, 2018. The proportion of underweight women decreased significantly over the past two decades, from 25% according to the 1993–94 National Health Survey of Pakistan (NHSP) to 16% in the 2011 National Nutrition Survey of Pakistan (NNSP) and 13% in the 2012–13 Pakistan Demographic and Health Survey (PDHS). Concurrently, the proportion of overweight women (BMI ≥ 25) has risen from 22.5% in the 1993–94 NHSP to 34% in the 2011 NNSP and 39% in the 2012–13 PDHS [7, 8]. This increasing trend in overweight and obesity among women has also been observed in various regional studies across Pakistan [9,10,11,12]. Despite these observations, a comprehensive assessment of the factors driving this shift is still lacking. It has been suggested that economic development, leading to higher incomes and subsequent changes in diet and physical activity, may be contributing to this trend, similar to patterns observed in other regions.

Demographic transitions in the last few decades have contributed to a sharp increase in the prevalence of malnutrition, particularly overweight and obesity, in Low-Middle-Income Countries [13]. Most countries in South Asia, including Pakistan, experience a double burden of malnutrition, which is the coexistence of both under- and over-nutrition, alongside an exponential upsurge of non-communicable diseases [14]. The National Nutrition Survey (NNS, 2018) in Pakistan reported that nearly 65% of women of reproductive age suffer from malnutrition. When women in this age group lack essential nutrients, the children born to them are at a significantly increased risk of long-term diseases and early death [15]. The same report revealed that more than 41% of Pakistan’s women were anemic, primarily due to iron deficiencies, leading to maternal mortality from anemia during pregnancy. These alarming figures have led the government to take several initiatives targeting malnutrition among women. However, progress on improving nutrition has been impeded by factors such as food insecurity, limited dietary diversity, high vulnerability to natural disasters, population growth, and rapid urbanization [16,17,18,19,20].

A provincial study suggests that the highest rates of undernourishment are found among women who participate in the labor force, those who breastfeed, consume nicotine, or live in rural areas [16]. Similarly, a cross-sectional study found that nearly 61.3% of women were anemic in rural regions [21]. Moreover, iron deficiency among reproductive-age women is considered a national concern, especially among those with a history of four or more pregnancies, birth intervals of less than 24 months, and obesity [5].

Furthermore, malnutrition poses significant threats to women's health, encompassing a range of consequences that impact both maternal well-being and the health of their offspring. Undernutrition in women can result in adverse pregnancy outcomes, including low birth weight, preterm birth, and maternal complications during childbirth. Additionally, undernourished women may experience impaired physical and cognitive function, increased susceptibility to infections, and higher rates of morbidity and mortality. Conversely, over-nutrition, characterized by overweight and obesity, contributes to an elevated risk of non-communicable diseases such as heart disease, stroke, diabetes, and certain cancers, compromising long-term health outcomes for women. These consequences highlight the urgent need for comprehensive interventions addressing malnutrition among women to safeguard their health and that of future generations.

Despite the severity of this issue, in Pakistan only a few studies have accounted for spatial dependence while estimating the prevalence and risk factors of body mass index (BMI) among women [22]. This study addresses this gap by assessing the spatial distribution and exploring the risk factors associated with the four forms of malnutrition—severely thin, underweight, overweight, and obesity—in Pakistan by using a Bayesian geo-additive quantile regression. This model is utilized for its flexibility to incorporate spatial and non-linear effects, providing a more comprehensive analysis compared to traditional linear regression models. Unlike previous studies, which included all women in the analysis without accounting for the nutritional requirements of pregnant and lactating women [23, 24], our study is based on a sample (n = 5,255) of non-pregnant and non-lactating women extracted from the Pakistan Demographic and Health Survey 2017–18. This approach helps to remove the bias in estimating malnutrition among women. The research findings may help policymakers to devise more effective programs for addressing the risk factors of malnutrition. We set up the following research question:

What are the spatial and socio-demographic risk factors associated with the four forms of malnutrition (severely thin, underweight, overweight, and obesity) among non-pregnant and non-lactating women in Pakistan?

Theoretical Framework

The specific analysis method of Bayesian geo-additive quantile regression was chosen for this study because of several reasons. Firstly, malnutrition prevalence and its risk factors can vary significantly across different regions. Bayesian geo-additive models allow for the incorporation of spatial effects, capturing regional variations and dependencies that traditional regression models might overlook. This aligns with the research question that seeks to understand spatial risk factors. Secondly, the relationship between socio-demographic factors and malnutrition is often complex and non-linear. Geo-additive models offer the flexibility to model these non-linearity, providing a more accurate representation of how various factors influence different forms of malnutrition. Thirdly, quantile regression enables the analysis of the entire distribution of the response variable, rather than just the mean. This is particularly useful for understanding the impact of risk factors at different points in the malnutrition distribution (e.g., severely thin versus overweight). This comprehensive analysis is crucial for formulating targeted interventions for different segments of the population. Fourthly, The Bayesian approach allows for the incorporation of prior information and the quantification of uncertainty in the estimates. This is beneficial in contexts with limited data or where prior studies provide valuable insights that can be integrated into the analysis. It also provides credible intervals for predictions, which enhances the interpretability and reliability of the findings. Finally, by focusing on non-pregnant and non-lactating women, the study aims to remove bias that could arise from the unique nutritional requirements of pregnant and lactating women. The Bayesian geo-additive quantile regression method, with its flexibility and robustness, is well-suited to handle the nuances and complexities of this specific sample, ensuring that the analysis remains accurate and relevant.

The outcome variable used in this study was the women’s nutritional status, measured in body mass index (weight in kilograms/height in meters squared), ages 15–49. Following WHO guidelines, we used four indicators of BMI, as follows, severe thinness (BMI < 16 kg/m2), underweight (BMI < 18.5 kg/m2), Overweight (BMI > 24 kg/m2), and obese (BMI >  = 30 kg/m2) [25]. Previous studies have used similar cutoffs to explain the prevalence of malnutrition in developing countries [24].

The theoretical framework for analyzing malnutrition among women in Pakistan incorporates individual, household, and environmental determinants within the context of the Social Determinants of Health (SDH), and Sustainable Development Goals (SDGs). The study used some individual and household characteristics as explanatory variables in the model to identify the patterns of different forms of malnutrition. Women’s education was measured as a categorical variable comprising no education, primary, secondary, and higher. Women's education serves as a key determinant, influencing health literacy, dietary knowledge, and access to resources, with higher educational attainment associated with improved health outcomes. Women’s exposure to media was measured using three separate indicators, as follows, reading newspapers, watching television, and listening to the radio. PDHS 2017–18 provides information on the frequency of using various mass media channels. We have combined all women’s responses for each mass-media channel into the following two options; responses indicating the use of a particular channel at least once a week were coded as “yes” or “no” otherwise. Exposure to media, including reading newspapers, watching television, and listening to the radio, impacts health behaviors and awareness about nutrition, affecting lifestyle choices. Women’s working status was a categorical variable, comprising “currently working women” or “not working”. The women’s use of bank services was measured in terms of their holding of bank accounts, consisting of categorical responses as “yes” or “no”. The proportion of the adult population with a bank account or access to other financial institutions is an indicator of decent work and economic growth, which is one of the critical targets of the Sustainable Development Goal 8 [26]. Employment status influences economic independence and resource access, while women's use of bank services reflects financial inclusion, which plays a crucial role in dietary patterns and healthcare utilization. These factors collectively shape women's nutritional status, underscoring the importance of examining them within the theoretical framework provided by SDH, and SDGs to comprehensively understand the determinants of malnutrition among women. The relevance of the selected variables to the Sustainable Development Goals (SDGs) is significant. Women's education (SDG 4: Quality Education) is crucial for enhancing health literacy, dietary knowledge, and resource access, thereby improving health outcomes. Exposure to media impacts health behaviors and awareness, aligning with SDG 3 (Good Health and Well-being) by promoting informed lifestyle choices. Employment status and financial inclusion, measured by women's working status and use of bank services, are linked to SDG 8 (Decent Work and Economic Growth), as they influence economic independence and resource accessibility, which are critical for better dietary patterns and healthcare utilization.

PDHS provides assets scores, based on household assets, possessions, and housing characteristics. The survey used Principal Component Analysis or PCA for creating a weighted wealth index. The ranking of each household member occurs by their asset scores. Based on household asset ratings, the distribution of national wealth is divided into five equal categories: poorest, poorest, middle, richer, and the richest. The household wealth status was used in this study as given by PDHS 2017–18.

This paper used the household’s access to electricity as a categorical variable: yes or no. It represents a household’s access to affordable, reliable, and sustainable clean energy, which is also an indicator of the Sustainable Development Goal 7-SDG 7 [27]. This study used two explanatory variables related to Water Sanitation and Hygiene or WASH, access to protected water and improved toilets. PDHS collects data on households’ access to various sources of water, including rivers, streams, tap water, wells, and others at the household. We combined the information on all responses into two groups: access to tap water was labelled as “protected water” and access to all other water sources was coded as “unprotected water”.

Moreover, the PDHS collects information on access to various toilet types, including pit latrines, covered latrines, open defecation, flush latrines, etc. This study pooled responses indicating the household’s access to a covered latrine or flush latrine under the “improved toilet” category and all other rejoinders as access to an “unimproved toilet”. The study used the place of residence (rural or urban) of women as a spatial unit of analysis.

Methods

Sample and Data

The study analyzed a subsample of 5,252 ever-married women aged 15–49 from the Pakistan Demographic and Health Survey (PDHS) 2017–18. Inclusion criteria comprised non-pregnant and non-lactating women who had given birth in the three years preceding the survey, ensuring homogeneity in nutritional requirements. Exclusion criteria removed pregnant women and those within three months postpartum to mitigate bias from transient physiological changes (e.g., gestational weight gain/loss). While the Bayesian geo-additive quantile regression accounted for spatial heterogeneity, the study had limitations. First, the cross-sectional design precludes causal inferences. Second, self-reported data (e.g., media exposure, toilet access) may introduce recall bias. Third, unmeasured confounders (e.g., cultural dietary practices, local food availability) could influence malnutrition outcomes. These limitations highlight the need for longitudinal studies to validate findings and explore contextual drivers of malnutrition disparities.

The PDHS 2017–18 sampling frame comprised the list of all enumeration blocks (EBs) used in the sixth Population and Housing Census 2017. The survey followed a two-stage stratified random sampling design. In the first stage, EBs were selected using a probability proportional to size rule, representing households living in the block during the census. A total of 580 primary sampling units (PSUs) were selected at this stage. In the second stage, 28 households per cluster (PSU) were chosen using an equal probability systematic selection procedure, making up a sample of 16,240 households. The estimates drawn from PDHS are representative at the national and regional/provincial levels. The National Institute of Population Studies implemented PDHS 2017–18 in collaboration with ICF and the United States Agency for International Development (USAID).

In reporting descriptive frequencies, DHS weights were used to ensure the results are representative of the national population. The complex sampling design was also accounted for in the analysis by incorporating sampling weights, stratification, and clustering. The weights provided by the PDHS 2017–18 were applied to adjust for non-response and the unequal probability of selection. This approach ensured accurate and reliable estimates, maintaining the validity and generalizability of the findings at both national and regional/provincial levels.

Statistical analysis

This paper used structured Bayesian Quantile regression [28] to examine the association of the explanatory variables with the entire conditional distribution of the response variable [29, 30]. Given the set of variables $(y_{i} ,x_{i} ,\upsilon_{i} ,s_{i} )$ and a fixed quantile $\theta \in (0,1)$, the general form of additive conditional quantile regression is given below,

$$Q_{{y_{i} }} \left\langle {\theta } \mathrel{\left | {\vphantom {\theta {x_{i} ,\upsilon_{i} ,s_{i} }}} \right. \kern-0pt} {{x_{i} ,\upsilon_{i} ,s_{i} }} \right\rangle = \eta_{\theta i} = x_{i}^{T} \beta_{\varphi } + \sum\limits_{k = 1}^{n} {f_{\theta k} } (\upsilon_{ki} ) + f(s_{\theta i} ) + b_{\theta i}$$

(1)

where $\eta_{\theta i}$ is the conditional $\theta th$ quantile outcome for the given values $x_{i} ,\upsilon_{i}$$s_{i}$, where $x_{i}$ a vector of categorical variables, and $\upsilon_{i}$ is a vector of continuous variables, and $s_{i}$ is the strata of the residence of the ith women; $\beta_{\theta i}$ is the vector of fixed coefficient categorical variables; $f_{\theta k}$ refers to kth smoothing functions for nonlinear effects; $f(s_{\theta i} )$ is the spatial effects of the strata of residence; and $b_{\theta i}$ accounts for unstructured random components. For the estimation of the parameters of Eq. 1, the check or loss function is as follows,

$$\begin{gathered} \arg \min \sum\limits_{i = 1}^{n} {\rho_{\theta } } (y_{i} - \eta_{\theta i} ); \hfill \\ \rho_{\theta } (u) = \left\{ \begin{gathered} u_{\theta } ,ifu \ge 0 \hfill \\ u(1 - \theta ),ifu \le 0 \hfill \\ \end{gathered} \right. \hfill \\ \end{gathered}$$

(2)

Equation 2 can be solved using a linearization problem for linear quantile regression, whereas a nonlinear case such as additive structured quantile regression requires a Bayesian inferential approach. This study considered fully Bayesian inferential for quantile regression. It requires relegating the suitable prior distribution to every continuous or metric covariates, coefficients of categorical variables, and spatial effects. The study modeled spatial effects using Gaussian Markov random fields (GMRF) [28]. The empirical form is given below:

Empirical model

$$\eta_{\theta i} = x_{i}^{T} \beta + f_{d\theta } (strata_{J} ) + b_{\theta } ,J = 1,2.$$

(3)

$x_{i}$ Refers to categorical variables, i.e., women’s education, exposure to mass media, holding a bank account, access to electricity, access to protected water, access to an improved toilet, and wealth status, and their effect is quantified in $\beta_{\theta }$; $f_{\theta } (.)$ represents spatial effects or place of residence (rural/urban) in the present case, and $b_{\theta }$ is the unknown random term. The estimation was carried out using BayesX, R software package developed and freely distributed by [31] to execute the structured additive Bayesian regression model.

Results

Table 1 presents descriptive statistics. The analysis shows that out of 5,252 women, 51.5% lived in rural areas, whereas 48.5% of women were from urban municipalities. Most women (53.4%) had no formal education, and only 13% had higher secondary or above-level education. The percentages of women who had listened to the radio, watched TV, and held bank accounts were 4.3%, 51%, and 9.5%, respectively. About 23% belonged to the wealthiest quantile, whereas 16% of women belonged to the poorest. Amongst women, only 16% were working women, and nearly 13% had improved toilet in their houses.

Table 1 Descriptive Statistics

The results in Table 2 indicate the likelihood of women being in different BMI categories (underweight, thin, overweight, obese) compared to normal weight. Specifically, women living in urban areas were more likely to be overweight (mean: 0.3; 95% CI: 0.06, 0.58) than those in rural areas. Women with primary education were more likely to be thin (mean: −0.93; 95% CI: −1.20, −0.65) and less prone to obesity (mean: −0.72; 95% CI: −1.11, −0.36), while those with secondary education were more likely to be overweight (mean: 0.53; 95% CI: 0.15, 0.91) and obese (mean: 0.77; 95% CI: 0.29, 1.24). Higher education levels were associated with a decreased likelihood of being thin (mean: 0.87; 95% CI: 0.58, 1.17) and underweight (mean: 0.95; 95% CI: 0.37, 1.47). Women with access to electricity were more prone to being thin (mean: −0.28; 95% CI: −0.53, −0.02). Wealthier women were less likely to be thin (mean: 0.50; 95% CI: 0.16, 0.85) and underweight (mean: 0.50; 95% CI: 0.122, 0.87) but more likely to be overweight (mean: 0.75; 95% CI: 0.33, 1.17) and obese (mean: 0.98; 95% CI: 0.49, 1.46). Similar results were observed for the richest women. Working women were less likely to be obese (mean: −0.51; 95% CI: −0.79, −0.23). Women with access to unimproved toilet facilities were more prone to being overweight (mean: 0.7; 95% CI: 0.34, 1.04) and obese (mean: 0.90; 95% CI: 0.48, 1.33). Lastly, women who watched television were less likely to be thin (mean: 0.34; 95% CI: 0.19, 0.49) but more likely to be obese (mean: 0.53; 95% CI: 0.27, 0.76). These associations are all referenced to normal weight, ensuring clarity in understanding the results.

Table 2 Posterior mean estimates for the linear effects of severely thin, underweight, overweight and obesity among women

Women’s financial inclusion indicator, bank account holding, indicates that women who had bank accounts were more likely to be thin (mean: −0.27; 95% CI: −0.55, −0.03). The household wealth index showed that women who fall under the poorer category are more likely to be underweight (mean: −0.81; 95% CI: −0.121, −0.42). In contrast, women who belonged to more affluent families were more likely to be overweight (mean: −0.86; 95% CI: −1.30, −0.41) and obese (mean: −0.92; 95% CI: −1.37, −0.44).

Figure 1 represents the estimates of spatial effects for the four forms of malnutrition. The maps on the left side (a) severely thin, c) underweight, e) overweight, g) obese demonstrate the estimates of the posterior means, whereas the maps on the right-side b) 95% CI of Severely thin, d) 95% CI of underweight, f) 95% CI of overweight, and h) 95% CI of obese indicate the statistical significance. The black-shaded areas (at 95% CI) reveal that districts in southern Pakistan, including districts from Punjab, Sindh, and Baluchistan are more likely to be severely thin or overweight (a negative spatial association with BMI means a positive association between BMI and each severely thin and underweight), and less prone to overweight or obese in the model focused on upper tails of distribution. The reverse explanation holds for districts represented by unshaded areas (districts in northern Pakistan including KP and GB) in maps. In contrast, the grey-shaded region implies no significant output from the spatial effects.

Fig. 1

Province maps of Pakistan showing the spatial effects of a) severe thinness, b) its 95% CI; c) underweight and d) its 95% CI; e) overweight and f) its 95% CI; g) obesity and h) its 95% CI

The spatial effects estimates indicate that the likelihood of being severely thin, underweight, overweight, and obese varies across geographical locations in Pakistan. The maps indicate (Fig. 1) a higher prevalence of severe thinness among women in the southern regions, such as districts of Punjab and Sindh and Baluchistan. A high prevalence of overweight women could be observed in the districts in Khyber Pakhtunkhwa. Many districts in Baluchistan depict a high prevalence of undernutrition, however the results are not statistically significant, as indicated by grey-shaded areas.

Figure 2 shows the nonlinear effects of the woman’s age for thin (a), underweight (b), overweight (c), and obesity (d), along with posterior means-solid black lines at 80% and 95% confidence intervals-dashed lines. It shows that women in the age group 20–30 are more likely to be underweight, whereas women tend to become overweight and obese over 30 years of age.

Fig. 2

The nonlinear effects of the woman’s age for severely thin(a), underweight (b), overweight (c), and obesity; posterior means-solid black lines at 80% and 95% confidence intervals

Discussion

This paper examined the spatial distribution of severe thinness, underweight, overweight, and obesity among women in Pakistan by following the cutoff criteria of WHO. Using Bayesian quantile regression technique, this study quantifies the effect of various covariates on different malnutrition categories (thin, underweight, overweight, and obesity). Women’s undernutrition has always been a daunting challenge despite concerted efforts to improve the nutritional status of women of reproductive age [14]. Previous studies have reported that women have a higher prevalence of poor nutrition than men in South Asia regions [32,33,34]. The situation is severe in the rural regions because of the massive poverty and socioeconomic inequalities that directly affect the dietary patterns of women together with the food insecurity and low dietary diversity in [14, 35,36,37].

Urbanization and Lifestyle Factors

This study reveals that women of reproductive age living in urban parts of Pakistan are more prone to become overweight, inferring that women have greater consumption of junk food or more high-calorie food. This result is in line with previous studies [3, 13]. Further, the sedentary lifestyle has caused a surge in the prevalence of overweight in developing countries [4]. The rising trend in the prevalence of overweight and subsequently chronic diseases, such as hypertension, heart disease, diabetes, and stroke can jeopardize the entire health system in developing countries where the readiness of health services is commonplace [38, 39]. More than half of the global chronic diseases are accounted for by developing countries, which can somehow be controlled through the provision of adequate nutrition [40]. However, managing both over and under-nutrition calls for integrated healthcare planning, political commitment, and radical reforms in nutrition and diet diversity [35]. The higher prevalence of overweight and obesity among women in urban areas of Pakistan can be attributed to several urbanization-related factors, including changes in dietary patterns, increased availability and consumption of processed foods, and sedentary lifestyles. Urban women are more likely to consume high-calorie, nutrient-poor foods due to the convenience and affordability of fast food and ready-to-eat meals, replacing traditional diets rich in vegetables and whole grains. The availability and marketing of processed foods are more prevalent in cities, contributing to poor dietary choices. Urban environments often lead to reduced physical activity levels, with more sedentary jobs, increased use of motorized transport, long working hours, and limited recreational spaces. To address these challenges, policymakers should implement public health campaigns promoting balanced diets, regulate the marketing and sale of processed foods, and create urban infrastructure that encourages physical activity. Initiatives such as urban farming, farmers'markets, nutritional labeling, and taxes on sugary drinks can improve dietary habits, while developing parks, walking paths, and recreational centers can promote active lifestyles. Workplace wellness programs and integrating health considerations into urban planning are also crucial strategies to mitigate the rising prevalence of overweight and obesity among urban women in Pakistan.

Socioeconomic Factors: Education, Wealth, and Employment

Our study showed that women with low education levels are more prone to become thin and underweight. It is generally believed that education level and eating attitude go hand in hand. Higher education is a proxy indicator of understanding a nutritious diet, eventually leading to better health [4]. Women with higher education can have better access to knowledge and understanding about the kind of food they eat, and thus, they can be more conscious about their food patterns [3, 4]. Findings indicate that women from the wealthiest wealth quintile are less likely to be thin and underweight, but they are more prone to overweight and obesity than women from the poorest wealth quintile. This is in line with the results reported by [32, 34]. Women belonging to a wealthy family can afford to buy multiple food items. They can have access to and purchase sugary and energy-rich junk food, which eventually results in over-nutrition, as indicated by [41]. The results of this study revealed that working women are less likely to become obese than non-working women. Working women can walk to the office or field, positively impacting their health [37, 42]. In South Asian culture, male spouse or male family members do not take part in household chores due to pre-defined gendered roles, which force women to solely perform all household responsibilities irrespective of their job status. Moreover, a large chunk of the population in Pakistan lives in rural areas where traditional agriculture is the main livelihood source, and women often work in agriculture with their partners and other family members, which could be a reason behind a reportedly healthy BMI among rural women [43].

Our study showed that women who had access to unimproved toilet facilities were more likely to be overweight and obese than their counterparts who had access to the improved toilet. In low-income countries, unimproved water and toilet facilities are the leading causes of infectious diseases, contributing 7% of diseases and aggravating the malnutrition problem [44].

Media Exposure

Access to media among women is common in Pakistan. Television is readily available in all regions of the country; thus, watching television might significantly impact malnutrition. This study shows that women who watched television were less likely to be thin and more prone to obesity. This result aligns with the prior evidence on the positive association of watching television (a sedentary activity) with the likelihood of obesity in women [45].

Spatial patterns and policy implications

The results of nonlinear models indicated that the tendency to become thin or underweight was higher in women of younger age. A similar study established that young women in developing countries have lower BMI than their counterparts in developed countries primarily because of their greater involvement in physical activities [46]. In contrast, the likelihood of women becoming overweight or obese tended to increase with the age. Our study results align with similar evidence in Bangladesh and Ethiopia [6, 32]. As women age, they tend to have a sedentary lifestyle, increasing the body mass composition [36].

The spatial patterns of malnutrition among women in Pakistan, as revealed by the study, have significant implications for public health policies and programs. Targeted interventions are crucial, particularly since urban areas are prone to overweight and obesity due to high-calorie diets and sedentary lifestyles, while rural regions face severe thinness and underweight issues due to poverty, socioeconomic inequalities, food insecurity, and low dietary diversity. Public health strategies should therefore include urban-specific health campaigns promoting physical activity and healthy eating, alongside rural initiatives that improve access to nutritious foods through subsidized food programs, agricultural support, and socioeconomic development efforts. Equitable allocation of resources based on the identified hotspots ensures that interventions are focused where they are most needed, preventing resource dilution and guaranteeing adequate attention to the most affected regions, such as Sindh and Baluchistan. Enhanced collaboration among governmental and non-governmental agencies can ensure a comprehensive approach, with community-based interventions fostering local ownership of health programs. Potential interventions include nutrition surveillance systems, school feeding programs, mobile health clinics, agricultural support, and public health campaigns. Leveraging spatial patterns for strategic, targeted, and effective public health policies not only improves health outcomes but also ensures a more equitable distribution of resources, ultimately reducing malnutrition rates across the country.

Limitations and future research

While this study provides robust insights into the factors influencing malnutrition among women in Pakistan, it has some limitations. The cross-sectional nature of the data limits causal inference. Furthermore, the study relied on available covariates, which may not capture all relevant behavioral or cultural factors. Future research could incorporate longitudinal data and qualitative insights to deepen understanding. Moreover, expanding the analysis to include adolescent girls or older women may provide a more comprehensive picture of gendered nutritional dynamics across the life course.

Conclusion

This study extended the literature by examining the spatial distribution of various forms of malnutrition among reproductive age women. Structured Additive Bayesian Quantile Regression was employed to analyze critical factors determining malnutrition. This approach quantifies the effect of various covariates on different malnutrition categories (thin, underweight, overweight, and obese). Findings reveal that place of household residence, wealth status, women’s education, women’s job status, and exposure to media were critical determinants of malnutrition among non-pregnant Pakistani women. The nutritional status of women has always been a challenge for poor and rural women. Several efforts have been made in this context by the government in coordination with international and local donor agencies to improve the access to nutrition and health diet across regions, but more concerted efforts are needed to address regional disparities in the nutritional status of women. National and provincial policies need to include more women in the planning and implementation of programs relating to maternal nutrition, women education and health, and sanitation and hygiene. Targeted interventions in nutrition and health are a matter of social well-being and a crucial step towards achieving zero hunger and ending regional inequalities in women’s malnutrition.