This study utilized data obtained from the National Family Health Survey (NFHS) conducted in 2015-16 (NFHS-4) and 2019-21 (NFHS-5), representing a nationally comprehensive cross-sectional survey managed by the Ministry of Health and Family Welfare, Government of India coordinated by the International Institute for Population Sciences (IIPS). These surveys offer insights into household and individual-level data regarding population, health, nutrition, and diverse demographic indicators.
The NFHS implemented a standardized sample design that covers national, state/union territory, and district levels. The methodology employed a two-stage cluster sampling technique.
In the initial phase, primary sampling units (PSUs) were selected using the probability proportional to size (PPS) method. In rural areas, villages serve as PSUs, while in urban areas, census enumeration blocks (CEBs) are designated PSUs. The process included the generation of a household list through mapping and household listing within each selected PSU. This method ensured a representative sampling approach in both rural and urban settings. Subsequently, in the second stage, researchers employed an equal probability systematic selection method to pick a set number of 22 households per cluster (i.e., PSUs). The comprehensive approach utilized by these surveys is outlined in the NFHS India Report11,13.
Out of all rounds of NFHS, NFHS-4, for the first time, provided information at the district level by increasing the sample size nearly fivefold compared with the previous survey. Like NFHS-4, NFHS-5 also provides district-level estimates for many important indicators. The information recorded during the NFHS-5 is similar to NFHS-4 to allow comparisons over time. However, NFHS-5 includes some new topics, such as preschool education, disability, access to a toilet facility, death registration, bathing practices during menstruation, and methods and reasons for abortion. The scope of clinical, anthropometric, and biochemical testing has also expanded11,13.
For this study, individual-level data were derived from both datasets to obtain information about 6 to 59-month-old children in Maharashtra. These datasets were derived after obtaining official permission to download the data from the Demographic and Health Surveys (DHS) website14. Ethical permission was obtained from respondents before collecting information during both surveys. The respondents' information was kept confidential, and their identities were not revealed at any step.
Information about a cumulative count of 249,967 and 230,870 children aged 6 to 59 months were obtained from the kids' files of NFHS-4 and NFHS-5 datasets, respectively. Of these total, 242,852 and 223,938 children were excluded from the NFHS-4 and NFHS-5 data files, respectively, for the study. This exclusion involved various criteria: 11,022 (NFHS-4) and 8,637 (NFHS-5) children who had passed away, 1,715 (NFHS-4) and 1,815 (NFHS-5) children who were not living with their mother or residing elsewhere, 22,876 (NFHS-4) and 22,559 (NFHS-5) children who were younger than six months or older than 59 months and, 1,607 (NFHS-4) and 9,071 (NFHS-5) children for whom height/weight data were unavailable. Moreover, children who did not reside in Maharashtra [204,756 (NFHS-4) and 180,911 (NFHS-5)] in addition to cases of any missing information with implausible height measurements and flagged cases [876 (NFHS-4) and 945 (NFHS-5)] among children were also excluded. Finally, a cohort of 7,115 and 6,932 children aged between 6 and 59 months from the fourth and fifth rounds of the NFHS, respectively, were chosen for inclusion in this study (Figure 1).
Outcome indicators:
I. Severe acute Malnutrition (SAM):
For this study, SAM was calculated as weight-for-height/weight-for-length, with a Z score of -3 Standard Deviations (SD) from the median of the reference population based on the WHO child growth standards. Children affected with SAM were coded as “1” or “0” otherwise.
II. Other Covariates:
The selection of covariates for this study was informed by a review of previous literature, considering their relevance to the research objectives, and assessing their availability in the NFHS dataset. This approach ensured that the chosen covariates were grounded in existing knowledge, aligned with the study's goals, and could be effectively analyzed using the available dataset, enhancing the study's validity and comprehensiveness.
The variables included three levels: Household, Maternal, and Child. Household-level variables included - place of residence (Urban, Rural), religion (Hindu, Muslim and Others), caste [Scheduled caste (SC) or Scheduled tribe (ST), Other backward class (OBC) and Others], household members (Less than five and five and more), wealth index (Poorest, Poorer, Middle, Richer, Richest). The maternal level characteristics included age (15-19, 20-34, 35-49), highest level of education (no education, primary, secondary and higher), total number of children (1-2, 2-3 and more than 3), media exposure (no, yes), Birth interval (first birth, less than 24 months, 24 months and more), birth order (1, 2-3, more than 3). The child level variables included in the study were - age (6-23 months, 24-35 months, 36-59 months), gender (male and female), place of birth (home, health facility/institute, others), birth weight (less than 2500 grams, 2500 grams and more, not reported), presence of diarrhea in the past two weeks (no, yes) and presence of fever in the past two weeks (no, yes).
Statistical analysis:
For this study, a descriptive analysis was carried out to estimate the frequency and percentages of children affected with SAM. The chi-square test was used for the comparison of the distribution of SAM by socio-demographic characteristics of the children. A multivariable binary logistic regression model was used for the NFHS surveys to estimate the significant associations between independent factors and SAM. The adjusted odds ratio (AOR) and confidence intervals (CIs) of 95% were calculated for both surveys to control for the effect of potential confounding factors, and results were considered significant if the p value was less than 0.05 (p < 0.05). The data analysis was conducted after assigning appropriate survey weights to the NFHS-4 and NFHS-5 datasets. The data were analyzed using the statistical analysis package STATA (version 14.1 Stata Corporation, TX, USA).