Changing Scenario of Anemia Among Women in India: Understanding the Maternal Health Concern and Its Associated Predictors Through National Family Health Survey

Objectives: To study the changes in Anemia among women and to study the association between anemia and socio-demographic and economic predictors, also estimating the percentage contribution of selected predictors to Anemia among women belonging to poor and non-poor households. Methods: The State-wise percentages were taken from the extracted factsheets of National Family Health Survey 2015-2016 and 2019-2020. Absolute and relative percent changes to understand the changes in prevalence have been calculated. Multiple logistic regression was done to understand the associated varying predictors of Anemia among women. Percentage contribution of selected varying predictors through multivariate decomposition analysis have also been estimated. Results: Anemia prevalence was found the highest among the women in Goa and least in Lakshwadeep. The relative change in prevalence of Anemia was highest in Assam and lowest in Lakshwadeep. Socioeconomic factors like marriage at a young age, lack of education, exposure to media, malnutrition, and poor economic status contribute signicantly to the prevalence of Anemia among women of age 15-49 years. Conclusion: It is seen that the socio-demographic and economic burden on Anemia still continues to be higher than other medical predictors. The focus should be made more on education, malnutrition, and economic status to reduce the prevalence of Anemia.


Introduction
The disease of anaemia affects approximately 2.5 billion individuals globally, and its effect is mostly on women and children whose epidemiology varies according to socio-economic, cultural, and geographical context. [1] Although there is a noticeable progress in socio-economic status and health welfare in most low-income countries in South East Asia, countries, especially India, still have a high malnutrition burden. [2]. The prevalence of under nutrition among less than 5 years aged children is high among Empowered Action Group (EAG) states of India, and the most important predictors were socioeconomic factors, exposure to media, and biomedical factors. [3] One of the leading causes of maternal deaths in India is Anemia and is a major factor of poor health, economic loss, and social burden. [4] According to the National Family Health Survey 4 (NFHS-4), 58.4 % of children aged 6-59 months and 53 % of all women aged 15-49 years, were anaemic in India. [5] The World Health Organization (WHO) de nes Anemia as "a condition in which the number of red blood cells or the haemoglobin (Hb) concentration within them is lower than normal". [6,7] The disease is associated with increased morbidity and mortality in women, especially pregnant and lactating women. Child growth faltering, impairment of cognitive function, increased chances of various kinds of infection, loss of productivity from impaired work capacity eventually also results in substantial economic burden to the family and entire population. [8][9][10] The prevalence of Anemia has declined from an estimated 55.4 % in 1990 to an estimated 51.4 % in 2016 with a decrease of roughly four percentage points over this time. [7,10] Additionally, it is to be noted that from 2014, the prevalence had increased from 51.19 % to 51.43 % in 2016, which is an increase of 0.24 percent points which is a serious concern [11].
To strategies reduction of Anemia, the program named Anemia Mukt Bharat (AMB) was launched by the Government of India under the Prime Minister's scheme for Holistic Nourishment (POSHAN) Abhiyaan. [12,13] One of the goals of the program is to reduce the prevalence of the disease by 3 % per year to support the end of goal of attaining a malnutrition free India by 2022. [12][13][14][15] However, the progress made is far less than the expected with socioeconomic burden still being a major concern. [16-19] The Sustainable Development Goals (SDGs) also targeted under SDG-2 to reduce the malnutrition burden among under-5 children, pregnant women, lactating mothers, adolescents girls, and older people [20][21][22][23].
To achieve these targets, it is essential that enough evidence is produced on the predictors of Anemia to eventually lead to optimum contribution to timely interventions in anemia reduction and prevention.
Additionally, exploring the commonalities and differences across the states in the country can inform and upgrade the state policies. Several previous studies have attempted to estimate the prevalence of Anemia, [24,25] but few studies have attempted to utilized nationally representative data to investigate the prevalence and determinants of Anemia among the states of India. Furthermore, the updated evidence using NFHS 5 data on prevalence and factors associated with Anemia among at-risk populations is lacking for the country.
In this context, we aimed to study the Anemia among women with the following objectives, a) To study the changes in Anemia among women across selected Indian states, b) To study the association between anemia and socio-demographic and economic predictors, and estimating percentage contribution of selected predictors to Anemia among women belonging to poor and non-poor households.

Methods
Our study utilized data from the National Family Health Survey of India, round 4 (NFHS-4) and round 5 (NFHS-5) fact sheets. NFHS is the Indian version of the global demographic and health survey (DHS). NFHS is a large scale periodic national survey conducted by the Government of India to obtain nationwide data on health and family welfare. Twenty-nine Indian states and 7 union territories were covered for data collection in NFHS-4. Separate questionnaires were used for men, women and households. Biomarker details for different disease conditions were also recorded. The data collected for the variables was self-reported. The NFHS-4 sample used a strati ed two stage sampling with a crosssectional study design. The details of the study, methods, sampling frame and questionnaire have been mentioned in the published report of NFHS-4. 699,686 women between the age of 15-49 years and 112,122 men between the ages of 15-54 years and 259,627 children were included in the survey and the overall response rate was 98 per cent. Our study included data of women in the age group 15-49 years from NFHS-4 (n=684,911). Blood samples from eligible and consenting women in the 15-49 years age group were collected by trained health investigators during NFHS surveys. The Hb analysis was done onsite with a battery-operated, portable HemoCue Hb 201+ analyzer. Participants with severe Anemia (Hb level less than 9 g/dL for women) were sent to a health facility for further evaluation and treatment. The details of the survey methods, sampling frame and questionnaire are detailed elsewhere.

Outcome and predictor variables
The outcome variable used in the study is 'anemia level'. The variables age, education, media, wealth index, source of drinking water and body mass index were used as predictors. The variable 'respondent current age' was categorized into '15-29', '30-39'and '40-49'. The variable 'wealth index' which was categorized as 'poorest', 'poorer', 'middle', 'richer' and 'richest' in the existing data (NFHS-4), was recategorized as 'poor', middle' and 'rich. We have further re-categorised into dichotomous categories as 'poor 'and 'non-poor' to understand the differentials of Anemia among poor and non-poor women. The variable 'education in single years' was categorised into 'less than 10 years of schooling' and '10 or more years of schooling. The variable media was generated on clubbing and then dichotomising the responses of three pre-existing variables in NFHS-4 data, i.e., 'frequency of reading newspaper', 'frequency of listening radio' and 'frequency of watching television. We have categorized the maternal body mass index into 'underweight (<18.5)', 'normal (18.5-24.9) and overweight/obese (>25.0) as per the guidelines of the world health organization (WHO).We have also dichotomised the pre-existing variable 'anemia level' in the existing NFHS-4 data, into 'anemic=1' and 'not anemic =0'for logistic regression and decomposition analyses. The dummy variables having dichotomous categories for all selected predictor variables were used in multiple logistic regression as well as in multivariate decomposition analysis.

Data Analyses
The analyses were done after removing list wise missing, agged and no information cases. The Statewise percentages were taken from the extracted factsheets of NFHS 4 and NFHS-5. We have calculated absolute and relative percent changes to understand the changes in prevalence of anaemia among women. The multiple logistic regression was done to understand the association predictors and Anemia among women on NFHS-4 data. We have also estimated the percentage contribution of selected predictors (education, media, BMI, source of drinking water and smoking) through multivariate decomposition analysis. The package 'mvdcmp' was used to estimate the contributing factors of anaemia among women in the age group 15-49.
The 'mvdcmp' package was primarily made for use in non-linear decomposition and was based on recent contributions, which include convenient method to handle path dependency, calculating asymptotic standard errors, and overcoming the identi cation of problem associated with the choice of a reference category when dummy variables are included among the predictors [26].
The relative change percentages were calculated to understand the variation in prevalence of Anemia among women across 22 states of India. The growth rate was calculated from 2015 to 2020 for projecting the prevalence numbers and rates till 2030. We used the mathematical growth modeling (geometric growth model) approach for estimating projected values of prevalence rate as well as numbers. In particular, mathematical projections are nothing but tting of the curve to the observed data. Where Z= n* 1 dependent variable vector anaemia (dependent variable), Y= n*k matrix of independent variables and σ is k*1 vector of coe cients. G(.) is any once-differential function mapping a linear combination of Y(Yσ) to Z.
The average difference in Z between two groups H (poor) and I (non-poor) can be decomposed as -Here component 'E' refers to the part of the differential attributable to difference in endowments or characteristics while, 'C' refers to the part of the differential attributable to differences in coe cients.
Here, H has been selected as the comparison group and I as the reference group. All the statistical analyses have been performed using STATA (version 13.0).

Status of Anemia among women in age-group 15-49 years across Indian States
It is clear that the prevalence of Anemia among women has increased over a period of 5 years in most of the Indian states except Lakshadweep, Dadra & Nagar Haveli, Andaman & Nicobar, Nagaland, Himachal Pradesh, and Meghalaya. Anemia prevalence was found the highest among the women living in Goa

Factors associated with Anemia among women in India
The results of multiple logistic regression revealed that socio-economic factors were signi cantly associated with Anemia. The women belonging to age group 40-49 were less likely [OR(CI),0.96 (0.95-0.98)] to developing Anemia as compared to women of age group 15-29, followed by age group 30-39 [OR(CI),0.95 (0.94-0.97)] respectively. The chances of Anemia is lesser [OR(CI), 0.76 (0.74-0.77)] among highly educated women than uneducated.The odds ratio was signi cantly lower in women who had exposed to media [OR(CI), 0.97 (0.96, 0.98)]. The overweight women [OR (CI), 0.71(0.64-0.79)] were less likely to developing Anemia as compared to underweight women followed by women having normal weight. The women consuming water from well water and spring [OR (CI), 1.22(1.20-1.24)], [OR (CI), 1.10(1.08-1.13)] were more likely to developing Anemia as compared to those having tap water respectively. The women belonging to richer sections of the society are less likely to become anemic than poor section of the society (Table.2). ®-Reference Category, p**<0.001, p*** <0.0001, LL -Lower limit, UL -Upper limit, Main contributing factors affecting Anemia among poor and non-poor women The multivariate decomposition results revealed that both endowment and coe cient are signi cant among high outcome group (poor). Differences in effects account for about 44.8 % of the observed poornon-poor differential in the prevalence of Anemia among women, with differences in intercepts (baseline logits). As per the convention, positive E(characteristic) coe cient indicates the expected reduction in poor to non-poor anemia gap in pregnant women if poor were equal to non-poor on the distribution of independent variables.
Equalizing the level of education would be expected to reduce the poor to non-poor anemia gap by around 11 %. Percentage contribution of exposure to media was around 12 %. This suggests us that by increasing awareness through mass media like newspaper/radio/television may be effective for reducing risk of developing Anemia among women. The chances of Anemia may be reduced by around 16 % among women if we equalised the distribution of safe drinking water among households in all regions of India. Similarly, probability of Anemia may be reduced by around 15 % if the women have normal BMI (Table 3). The main limitation of the projective mathematical model: The projection modeling helps us to understand the futuristic trend so that we can get the idea on the seriousness of the diseases in coming days, months, or years. In geometric growth model, there is no problem like arithmetic growth for the selection of the base year population. The mathematical growth modeling curve may t the observed data with greater accuracy and, yet fail to produce the prevalence number/rate for future with same accuracy if the population trends undergo drastic change.
Other limitations: The study's one of the other main limitation was that it was not feasible to compare entire national estimates due to unavailability of NFHS-5 data. Also, we cannot establish any causal relationship because of a cross-sectional study.

Conclusion And Recommendation
It is clearly seen that the socio-demographic and economic burden on Anemia still continues to be higher than other medical predictors. To accelerate the reduction of the prevalence of Anemia, it is important to equalize the level of education among women and spread awareness on the development of risk of Anemia and its effect on children. This can be done by increasing awareness through mass media like newspaper/radio/television, may be effective for reducing the risk of developing anaemia among women.