Chronic non-communicable disease burden among reproductive ... - BioMed Central

Data

The present study utilized data collected under the fifth round of the National Family Health Survey (NFHS-5), 2019–2021, available in the public domain for legitimate research purposes. It can be obtained through https://dhsprogram.com/data/available-datasets.cfm. The survey covered a range of health-related issues, including non-communication diseases.

NFHS-5 is the second nationwide community-based survey after NFHS-4 in India to provide estimates of blood glucose levels and blood pressure in the general population. Specifically, among women aged 15–49 years and men aged 15–54 years for all the Indian States and Union Territories (UTs), and districts. Survey data consists of 724,115 women samples. After dropping 28,408 pregnant women observations, the final analysis used information on the remaining 695,707 women samples. Women generally modify their dietary and lifestyle behaviours during pregnancy; therefore, including pregnant women in the study might affect the study estimates. For the same reason, they were dropped off from the final analysis.

Ethics statement

The present study utilizes a secondary data set from the recent NFHS-5 survey with no identifiable information on the survey participants. This dataset is available in the public domain for legitimate research purposes. Hence, there is no requirement for any additional ethical approval. The study utilizes data from a national survey conducted under the stewardship of the Ministry of Health & Family Welfare, Government of India, with the help of the International Institute for Population Sciences, Mumbai. The survey received ethical clearance from the Institutions Review Board (IRB) of the International Institute for Population Sciences, India. Additionally, the NFHS survey has taken consent from all the eligible participants age 18 & above. However, participants in the age 15–17 years required consent was taken from their parents.

Outcome variables

The outcome of interest was the chronic disease score (CDS), computed using the information on eight non-communicable diseases available in NFHS-5. Out of these eight, four were self-reported; these included asthma, cancer, chronic heart disease, and thyroid disorders. Whereas diabetes, hypertension, obesity, and anemia were measured as amalgamating self-reported and measured diagnosis of chronic conditions. A woman was categorized as diabetic if their random blood glucose level ≥ 140 mg/dl. Women with average systolic blood pressure > 140 mmHg or average diastolic blood pressure > 89 mmHg were considered hypertensive. Obesity was measured using Quetelet Index, also known as Body Mass Index (BMI), calculated as:

$${\varvec{Body\,Mass\,Index}}=\frac{Weight\,(in\,Kgs)}{{Height}^{2} (in\,{m}^{2})}$$

A woman was considered obese if her BMI ≥ 25 (kg/m²) [29].

All the eight diseases were coded into binary categories of absent—'0' and present—'1'. Finally, the outcome variable, i.e., chronic disease score (CDS), was generated and was further classified into three, no morbidity (women with zero chronic disease), single morbidity (women with exactly one chronic illness), and multimorbidity (women who are suffering from two or more chronic conditions simultaneously).

Explanatory variables

The present study included three sets of explanatory variables: (1) socio-demographic and economic factors, including age (categorised into 5 years age group between 15 to 49), place of residence (categorised as urban and rural), religion (categorised as "Hindu", "Muslim", and "Other"), marital status (categorised as "Ever married" and "Never married"), parity (categorised as "no children", "one child", and "two or more"), and menopause (categorised as "yes' and "no), working status (categorised as "yes" and "no"), and wealth index (categorised into "poorest", "poor", middle", "rich", and "richest") (2) health behaviours; including tobacco use (categorised as "yes' and "no"), alcohol consumption (categorised as "yes' and "no"), dietary habits (categorised as "normal/healthy' and "unhealthy), and (3) anthropometric indicator: waist-hip ratio (WHR) (categorised as "high risk WHR' and "low risk WHR").

Dietary Index

In NFHS, nine questions pertaining to dietary practices were asked. The frequency (frequently, occasionally, and never) of consuming nine food items in a week, namely milk/curd, pulses/beans, dark green leafy vegetables, fruits, eggs, fish, chicken/meat, fried food, and, aerated drinks were available. However, the use of MCA facilitated in making an index that combines good and bad eating habits after re-coding nine items in a unidirectional manner, such that each item measures the same concept, where "0" corresponds to those who "frequently consume, say, cereal," which can be considered as a good eating habit. Whereas, in the case of junk/sweet/fried foods, "0" were those who "never consume junk food," which is again a good thing. Similarly, re-coding was done for the remaining items. Prior to index computation, Cronbach's alpha 'α' was used to verify internal consistency between the nine features. Cronbach's alpha measures internal consistency, that is, how closely related a set of items are as a group. Finally, using MCA, an index was generated and divided into two categories after sorting in ascending order [30]. Thus, the "index value" in the second half would be higher values which were considered as an unhealthy diet index coded as "1" and the other half as "0". Here unhealthy diet practitioners were identified as those who consume milk/curd, pulses/beans, dark green leafy vegetables, fruits, eggs, fish, chicken/meat occasionally/never and consume fried food, and, aerated drinks occasionally/daily.

Waist-Hip Ratio (WHR), was measured using:

$$WHR=\frac{Waist\,Circumferenc\,(in\,cm)}{Hip\,Circumference\,(in\,cm)}$$

A woman is considered at a high risk of developing long-term health conditions if her WHR ≥ 0.85.

Statistical analysis

Firstly, descriptive statistics were conducted to study the sample distribution. Further, women's disease profile was explored using prevalence measured using:

$$\begin{aligned}&\mathbf{Prevalence}\,(\mathbf{per}1000\,\mathbf{women})\\ & =\frac{\mathrm{All\,new\,and\,existing\,cases\,during\,a\,given\,time\,period}}{\mathrm{Surveyed\,women\,during\,the\,same\,time\,period}}\,*\,1000\end{aligned}$$

In addition, bivariate analysis was used to understand the chronic disease burden among reproductive-aged women by socio-economic and demographic variables across India in 2019–2021.

In epidemiological and biomedical studies, the proportional odds model (POM) has often been used [31]. The proportionality assumption was checked using the brant's test before further analysis. However, if the proportionality assumption does not hold, the partial proportional odds model may have been a better choice (which was not the present case) [32]. If the log odds ratio across the cut points is identical, i.e., the proportional odds assumption is satisfied, the proportional odds model is used.

Observations on the chronic condition related to multimorbidity (Y) for each woman are classified into three categories. Likewise, covariates (x_i) denote the p-dimensional vector of covariates (i = 1, 2, …, p), containing the observation on the complete set of p explanatory variables. Accordingly, the dependency of Y on x_i can be expressed as:

$$\mathrm{Pr}(Y\ge {y}_{j}|{x}_{i})=1/(1+exp\left(-{\alpha }_{j}-{x}_{i}^{^{\prime}}\beta \right), j=\mathrm{0,1},2$$

Or

$$\mathrm{log}\left[\frac{Pr(Y\ge y\_j |x)}{1-Pr(Y\ge y\_j |x)}\right]=-{\alpha }_{j}-{x}_{i}^{^{\prime}}\beta , j=\mathrm{0,1},2$$

where \(\mathrm{Pr}(Y\ge {y}_{j})\) is the cumulative probability of the event \(Y\ge {y}_{j}\); \({\alpha }_{j}\) are the respective intercept parameters; β is a (p by 1) vector of regression coefficients corresponding to \({x}_{i}\) covariates. Results are then presented as an odds ratio (OR) with a 95% confidence interval (CI).

Statistical analysis and data visualization was performed with STATA version 15.0 (StataCorp™, Texas) and MS Excel. A p-value < 0.05 was considered as statistically significant for all calculations. All estimates were reported by applying appropriate sampling weights. As the data used in the study was taken from Women's file, national women's weights were employed in the analysis. Additional information on survey weight can be seen from national report of NFHS (Add reference of the report here).