Methodology

To augment existing literature with more diverse methodologies, the Exemplars study took a more holistic approach than prior reviews of anemia reduction among women of reproductive age (WRA) in Pakistan, triangulating across literature reviews, quantitative analyses, and qualitative inquiry.

Literature review

EGH performed a systematic search of published peer-reviewed literature to gather information on contextual factors, interventions, policies, strategies, programs, and initiatives that may have affected anemia among WRA in Pakistan over time.

The epidemiology and etiology of anemia are multifactorial and involve a complex interplay of distal, intermediate, and proximal causes.^1,2,3 As part of the larger Exemplars in Anemia Reduction Among WRA project, EGH carried out a systematic review of peer-reviewed and gray literature on the determinants and drivers of anemia reduction among WRA in low- and middle-income countries. Our search identified several review articles, which we then used to create a conceptual framework to guide our analytical approach and assist in identifying and interpreting determinants of anemia among WRA.

Initial indexed literature database searches returned 4,535 records, which were reduced to 2,701 after de-duplication. Applying screening criteria to titles and abstracts left 302 records, which were then reduced to 60 upon full-text review.

Quantitative analyses

Quantitative methods involved (1) descriptive analyses to provide contextual understanding of the stunting decline across geographic, socioeconomic, gender, and age segments, and (2) hierarchical multivariable regression and regression-based decomposition analyses to understand the major predictors of the stunting decline, as well as their relative importance to Pakistan's progress.

Pakistan's series of National Nutrition Surveys (NNSs) (2011, 2018) served as the primary quantitative data sets in this study.^4,5 NNSs are nationally representative household surveys. During each survey round, data for a wide range of indicators in the areas of population, health, and nutrition are collected, using a comparable standardized method.⁶ Anemia testing for WRA is included in each Pakistan NNS.

Determinants of anemia among WRA can be grouped in different hierarchical levels, in terms of their causal proximity to the impact of interest-specifically, as "distal causes," "intermediate causes," and "proximal causes." Each of these lies on a causal path toward nutritional outcomes, with more-proximal causes functioning as mediators of the distal determinants.

Descriptive analyses

EGH created panel data sets from NNS 2011 and NNS 2018 to assess factors associated with the increase of hemoglobin (Hb) and the decrease of anemia prevalence during the study period. Given that Hb levels change throughout pregnancy due to hemodilution and an increase in blood volume, EGH restricted all multivariable analysis to non-pregnant women only.

The conceptual framework was used to select possible predictors of Hb and anemia (defined as Hb <12 g/dL in non-pregnant women).⁷ Univariate statistics were estimated using means and standard deviations for continuous variables and frequencies/proportions for categorical variables, as appropriate. All analyses accounted for survey design and weighting, as appropriate.

Equity analyses were designed to illustrate subnational variations in anemia prevalence among WRA across household wealth, maternal education, area of residence, and geographic region. The slope index of inequality and concentration index were also calculated to measure absolute and relative socioeconomic inequalities, respectively.

Step 1 was a series of bivariate regressions to determine crude associations between indicators in our conceptual framework and Hb/anemia outcome. Step 2 was to use all candidate variables for multivariable model building (i.e., with p value ? 0.20) irrespective of their direction to move forward for multivariable modeling. Bivariate models estimated the absolute crude associations between the covariable and the outcome, and they highlighted the total (unadjusted) effect of the factor on Hb/anemia.

To examine the association between Hb and various indicators, EGH conducted a series of hierarchical models using distal-, intermediate-, and proximal-level variables to generate the final multivariable models. Variables within each level were selected from our general conceptual framework. The final multivariable regression coefficient was adjusted for child age, sex, and region (control variables) and all confounders in preceding levels.

Oaxaca-Blinder decomposition

The Exemplars study only included the survey years of 2011 and 2018 by design because the Oaxaca-Blinder decomposition only uses two survey time points in a given analysis and thus "ignores" in-between survey rounds and any intermittent fluctuations in the predictors. As has been suggested in previous decomposition analyses, EGH operationalized Hb as the linear outcome due to its greater statistical efficiency relative to the dichotomous anemia outcome.

The Exemplars study undertook the commonly used Oaxaca-Blinder decomposition methods to assess determinants of change in anemia among WRA in Pakistan over time.⁸ These methods are based on individual-level data; as such, they have increased statistical power and have been widely used to assess nutrition determinants over time in low- and middle-income settings.^9,10,11,12

Qualitative analysis

The qualitative element of the study was conducted at the national, district, and community levels to investigate a diverse range of stakeholder perspectives. The study was conducted in Pakistan, which comprises four provinces (Punjab, Sindh, Balochistan, and Khyber Pakhtunkhwa) and two territories. Two districts were selected from each province based on their performance on anemia reduction; one best-performing district and one worst-performing district were selected from each province. For Punjab, Muzzaffargarh was selected as high performing, Bhakkar as low performing; for Sindh, Tando Muhammad Khan was selected as high performing, Tando Allahyar as low performing; for Balochistan, Harnai was selected as high performing, Pishin as low performing; in Khyber Pakhtunkhwa, Batagram was selected as high performing, Torghar as low performing.

At the national level, a total of 12 informants were interviewed for the study, selected on the basis of their experience in areas or institutions that were instrumental in the design and implementation of interventions to reduce anemia among WRA in Pakistan. The national interviews involved key government ministries, departments, and agencies; development partners (World Food Program, UNICEF, and international nongovernmental organizations); and industries engaged in food fortification, such as the Pakistan Flour Mills Association.

At the district level, interviews were conducted with 21 stakeholders. Eleven were conducted with provincial/ regional government representatives-for example, the director general of health, nutrition program specialist, director of nutrition of the specific province, among others. These representatives were purposively selected for having extensive experience in designing and planning nutrition-related programs and policies or maternal health and nutrition services, or experience in implementing nutrition-related programs and policies and evaluation activities. Ten interviews were conducted with representatives of provincial / regional non-governmental organizations-for example, World Health Organization, World Food Program, and Nutrition International, among others. These participants had experience working on nutrition-related projects and direct food fortification.

At the community level, 48 interviews were conducted across district health officers, provincial/regional nutrition focal persons, health care providers, and community members.

Program, policy, and financing review

To understand implementation of nutrition-sensitive and nutrition-specific policies, programs, and strategies, EGH conducted additional research and corroborated our findings with country experts. EGH undertook a similar multipronged data collection and corroboration exercise to track financial data linked to the nutrition policy and program timeline.

Limitations

The main strength of the Exemplars study was the mixed-methods approach, which contextualized the quantitative findings and situated them within a framework that clearly demonstrated how Pakistan was able to achieve a reduction in anemia burden between 2011 and 2018. However, Exemplars analyses had limitations as well. First, the main data source for our analysis was the Pakistan NNS, which is a cross-sectional survey conducted approximately every five years. Therefore, we could not infer causality. However, EGH used data from two survey years to create a panel data set, which enabled assessment of the effect of changes in anemia determinants on anemia among WRA over time. Additionally, multivariate statistical analysis also controlled for time-invariant region of residence, which accounted for any trend effects. Second, as previously established, the etiology of anemia is complex and multifactorial, and includes nutrient deficiencies, infection, inflammation, and hemoglobinopathies. Quantitative data on these-specifically, dietary intake and iron deficiency-were not available. Finally, although qualitative inquiry was used to address some of the gaps in our quantitative data, analyses was only carried out subnational and community-level data collection in two regions and districts, respectively. Therefore, findings were not generalizable nationally.