COVID-19 pandemic profiles

After the first identification of a novel coronavirus in December 2019 in Wuhan, China, the virus spread all over the world very quickly. The trends in total COVID-19 cases by Indian states have been presented in Fig 1. The first Coronavirus positive case in India was reported approximately seven and a half months after the commencement of the fieldwork (on June 17, 2019) of the National Family Health Survey-5 (NFHS-5) 2019/21 that furnished the data for the current study [2, 24]. In India, the first Coronavirus positive case was reported in Kerala state on January 30, 2020 [2]. The virus had left its footprint in almost all of the Indian states and UTs (33 out of 36 states and UTs) by mid-April 2020 [32]. Three Indian states/UTs (Dadra and Nagar Haveli and Daman and Diu, Lakshadweep, and Sikkim) saw a coronavirus positive case later [32]. Before the lockdown was imposed on March 25, 2020, a total of 492 COVID-19 cases were identified in India, and that number rose to 18.8 million by the end of the fieldwork of the NFHS-5 on April 30, 2021. Among the selected states, Uttar Pradesh was the hardest hit by the pandemic (had a total of 1.2 million COVID-19 cases by the end of the survey), and Arunachal Pradesh had the lowest number of cases (18,256 cases) by the end of the survey [30].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Trends in total COVID-19 cases in the selected Indian states during the fieldwork of NFHS-5 2019/21.

Source: PRS Legislative Research. Number of Cases. 2022 [32].

https://doi.org/10.1371/journal.pone.0314800.g001

Socioeconomic profiles

The predicted differential effects of the COVID-19 pandemic on fertility in the countries with different levels of socioeconomic development make it important to discuss the socioeconomic development status of India and its states/UTs [31]. Currently, India is a lower middle income country [20]. According to the baseline report on the National Multidimensional Poverty Index-2021 (NMPI-2021), approximately one out of every four Indians is multidimensionally poor [33]. The NMPI-2021 report used 12 indicators from three dimensions (health, education, and standard of living) for calculating the multidimensional poverty rate (MPR) and multidimensional poverty index (MPI). The NMPI-2021 report classified the states and UTs into five groups using the MPI values [33]. Using that classification, we further classified the socioeconomic status of the states and UTs into three groups: i) rich socioeconomic status having an MPI value of 0.105 or lower; ii) moderate socioeconomic status with an MPI value between 0.106 and 0.158; and iii) poor socioeconomic status having an MPI value of 0.159 or higher. According to this classification among the 13 states selected for this study, the socioeconomic status of Punjab (MPR = 5.6%), Tamil Nadu (MPR = 4.9%), Delhi (MPR = 4.8%), Puducherry (MPR = 1.7%), Uttarakhand (MPR = 17.7%), and Haryana (MPR = 12.3%) was rich; the socioeconomic status of Chhattisgarh (MPR = 29.9%), Rajasthan (MPR = 29.5%), Odisha (MPR = 29.4%), and Arunachal Pradesh (MPR = 24.3%) was moderate; and the socioeconomic status of Jharkhand (MPR = 42.2%), Uttar Pradesh (MPR = 37.8%), and Madhya Pradesh (MPR = 36.7%) was poor.

Demographic profiles

India is the second-most populous country in the world with an 1.4 billion people [21]. The selected states and UTs comprise over 50% of the total Indian population, with Uttar Pradesh having the highest population (237.9 million) and Puducherry having the lowest population (1.4 million) among those selected states and UTs [34]. The population growth rate in India has decreased substantially during the past few decades, primarily because of its fertility decline [35].

India and all of its states have experienced noticeable fertility declines over the last three decades. The TFR in India decreased to 2.0 in 2019/21 from 3.4 in 1992/93 [23, 24]. TFR fell below two in all the selected states/UTs except Rajasthan, Uttar Pradesh, and Jharkhand. The TFRs in the two selected southern states/UTs were lower compared to other states/UTs. TFR in India decreased during the recent inter-NFHS period (2015/16–2019/21). During this inter-survey period, the TFRs in all the selected states also declined, except in Punjab and Tamil Nadu. Among the selected states and UTs, Uttar Pradesh had the highest TFR in 2019/21, while Punjab had the lowest TFR during the same period. This decline in fertility resulted from changes in the fertility behaviours of the people. The mean ideal family size in India decreased to 2.1 in 2019/21 from 2.9 in 1992/93. During this period, all the Indian states experienced substantial declines in their ideal family sizes. Among the states selected for this study, the mean ideal family sizes increased in Arunachal Pradesh, Madhya Pradesh, and Tamil Nadu, while the mean ideal family sizes stalled in Haryana, Delhi, and Puducherry during the recent inter-NFHS period (2015/16–2019/21). The remaining selected states experienced decreases in the mean ideal family sizes during that period. Contraceptive use in India increased to 66.7% in 2019/21 from 40.6% in 1992/93. Contraceptive uses in all of the selected states also increased between 1992/93 and 2019/21 [23, 24, 36].

Data

This study used secondary data from the NFHS-5 conducted in 2019/21. The NFHS-5 data are freely available on the demographic and health survey (DHS) website (https://dhsprogram.com/data/dataset/India_Standard-DHS_2020.cfm?flag=1) upon a request for accessing the data. The NFHS-5 is a cross-sectional, nationally representative survey of 636,699 households, providing information for a wide range of monitoring and impact evaluation indicators of health, reproduction, nutrition, disability, and women’s empowerment. This study used the data of ever-married women of reproductive age (15–49 years). A woman was interviewed only if she provided verbal consent to participate in the survey in response to a loud reading of an informed consent statement by the interviewer. The NFHS-5 was approved by the Institutional Review Board of ICF (project number: 190308.0.001.00). The NFHS-5 adopted a uniform stratified two-stage sample design considering urban-rural area, village population sizes, percentages of the scheduled caste and scheduled tribe populations, and literacy rates. Probability proportional to size (PPS) sampling was used to select the primary sampling units (PSUs) or clusters. Villages in rural areas and Census Enumeration Blocks in urban areas were considered clusters. The 2011 census was used as the sampling frame for selecting the clusters. Trained research investigators used computer-assisted personal interviewing (CAPI) to collect data. For ensuring data quality, NFHS strictly followed the protocol for maintaining gender-sensitive team composition with a mix of an adequate number of male and female investigators who were from the native states and proficient in local languages [37]. NFHS ensured data quality by monitoring field work at multi-layer including spot checks, backchecks, review of field check tables, continuous feedback, debriefing field workers, and continuous supportive supervision [38]. The NFHS-5 received technical support from the International Institute for Population Sciences (IIPS), Mumbai, and the United States Agency for International Development (USAID) via ICF [38]. Field staff from IIPS conducted spot checks and back checks of surveyed households in a minimum of 10% of PSUs that were randomly selected by the IIPS central office [38]. The published reports contains more detailed descriptions of the questionnaire, quality control measures, survey design, and survey management procedure [24, 37, 38].

The NFHS-5 interviewed reproductive age ever-married women during the period between June 17, 2019 and April 30, 2021, with a short pause due to the COVID-19 lockdown. The lockdown lasted from March 25, 2020, to May 31, 2020. No fieldwork was done during the lockdown period. From the 36 states and UTs (28 states and 8 UTs) in India, interviews of 724,115 ever-married reproductive-aged (15–49 years) women (cases) from 30,161 clusters were completed with a 97% response rate. Out of these total interviewed cases, our study initially selected 358,603 weighted cases only from those 13 states and UTs (11 states and two UTs) that were surveyed during both pre- and post-lockdown periods (see Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. States and UTs included in the study, India.

Source: Authors’ production using the shapefile from Runfola et al. [39].

https://doi.org/10.1371/journal.pone.0314800.g002

Out of those 358,603 cases, respectively, 42.9% and 57.1% were interviewed during pre- and post-lockdown periods. The selected 13 states and UTs constituted about half of the samples that were interviewed in India by the NFHS-5. For analysing the desire for a birth within two years, current contraceptive use, and sexual activities in the last 30 days in the pool of the 13 selected states and UTs, respectively, 56,561, 250,810, and 53,212 weighted cases (weighted using national-level sample weights) were selected following same selection criteria as the NFHS-5 published report (presented in Fig 3). The cases selected for analysing current contraceptive use excluded those who were not in a union or living with a man (including those who were never in a union) at the time of the survey, and who were not sexually active at the time of the survey. In selecting cases for analysing current contraceptive use, a woman was considered sexually inactive if currently she was not in a union or not living with a man and also did not have sex in the last 30 days preceding the survey. The selected cases for analysing desire for a birth within two years, current contraceptive use, and having sex in the last 30 days covered, respectively, 14,966, 15,645, and 4,712 clusters.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Flowchart of case selection for analyses.

Note: All the numbers presented in the diagram were weighted using national level sample weights.

https://doi.org/10.1371/journal.pone.0314800.g003

Besides the data on the ever-married women, the NFHS-5 also collected the geographic locations of the centers of the clusters. The geographic locations of the clusters that contributed to the selected samples were also analysed to examine the overall patterns of the geographic locations of the collected data.

Variables

The main objective of this study was to examine the differences between fertility related behaviours during the pre- and post-lockdown periods in India. For achieving this objective, we selected the following three response (dependent) variables based on the literature review and availability of data: i) desire for a birth within two years (before completion of two years), ii) current contraceptive use, and iii) having sex in the last 30 days. In selecting cases for each of the dependent variables, we followed the same criteria as the NFHS-5 published report. In analysing the desire for another birth, we excluded the women who did not want any more children and who gave non-numeric responses (for example, “don’t know” and other non-numeric responses). In the analysis, those who desired a birth within two years were coded 1, and those who desired a birth after two or more years were coded 0. Contraceptive use included both modern and traditional methods. For analysing contraceptive use, we included only the women who were currently in union, living with a man, or sexually active. The analysis of sex in the last 30 days was based on the women who reported the times since their last sexual intercourse.

The selection of the predictors (independent variables) for this study was guided by the literature review and availability of data. The predictors included in this study were whether a respondent was interviewed in the pre- or post-lockdown period, (respondent’s) current age, number of children (alive), religion, caste, level of schooling, wealth status, residence, and media access.

The scheduled castes and tribes are the groups of people officially designated by the Constitution of India. The Government of India uses other backward class as a collective term to classify castes that are socially or educationally underprivileged [40]. Wealth status of a respondent was determined using the wealth index that was readily provided by the NFHS-5. The wealth index was created using scores based on the types and number of consumer goods owned and the housing characteristics. Principle component analysis was used to calculate the scores. Detailed methodology of calculating the wealth score is available elsewhere [41]. After ranking the respondents using their wealth scores, their distribution was divided into five equal groups, which were classified as poorest (in the lowest quintile), poorer (in the second quintile), middle (in the third quintile), richer (in the fourth quintile), and richest (in the highest quintile). A woman was considered to have media access if she accessed at least one of the following media at least once a week: television, radio, a newspaper, and a magazine.

Statistical analysis

Both descriptive and multivariate analyses were performed to analyse the data. All these analyses were performed using the technique developed for complex sample design, which uses sampling strata, clusters, and sample weights. This technique is very important for estimating accurate confidence intervals and significance levels [42]. State-level sample weights were applied to analyse the data from individual states and UTs, and the pooled data were analysed by applying national-level sample weights. In descriptive analysis, percentage distributions and 95% confidence intervals were used to examine the differences between fertility behaviours in the pre- and post-lockdown periods.

For multivariate analysis, we used a two-level mixed effects logit model to address the hierarchical composition of the NFHS-5 data. This two-level model was used because the individuals (level 1) were nested within clusters (level 2). The mixed-effects model can be preferred because it can incorporate the variability in the estimates of the effects of the predictors over the clusters, even if they are small [43]. Estimates of F-statistic and intraclass correlation coefficient (ICC) were used to examine the goodness of the fits of the models. The ICC accounts for the extent of similarity of the outcomes in each cluster that may result from a random effect compared to the outcomes in other clusters. An ICC value of 0.01, 0.10, and 0.25 might be considered a small, medium, or large effect, respectively [44, 45]. A mixed-effects logit regression reduces to a standard logit regression in the absence of a random effect [46].

A dichotomous predictor–interview period (whether interviewed in the pre- or post-lockdown period)–was included in the model to examine the effects of the pandemic lockdown on fertility-related behaviours when controlled for the effects of other relevant predictors. A dichotomous variable was chosen instead of a categorical month variable because a very small number of women were interviewed in some months. The effects of the lockdown on each fertility variable were examined for all the selected states, UTs, and pooled data. Besides these, the geographic pattern of sample collection was examined by simply plotting the GPS locations of the clusters to which the selected cases belonged.

The descriptive analysis was performed using IBM SPSS Statistics for Windows, version 23.0 (IBM Corp., Armonk, N.Y., USA), and the multivariate analysis was performed using Stata Statistical Software for Windows, release 17.0 (StataCorp LLC, College Station, Texas, USA). The geographic locations of the clusters were plotted using ArcGIS Desktop for Windows, version 10.7.1 (ESRI, Redlands, C.A., USA).

Profiles of samples and interviews

Fig 4 shows the geographic locations of the clusters that contributed cases to the analyses in this study. All the selected states were interviewed in both pre- and post-lockdown periods, and the geographic locations of the overall clusters by interview period (pre- and post-lockdown period) did not show any systematic pattern regarding geographic direction (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Clusters of the included cases by interview period, India, 2019/21.

Source: Authors’ production using the shapefile from 2019/21 NFHS-5 and Runfola et al. [39].

https://doi.org/10.1371/journal.pone.0314800.g004

Comparatively more women were interviewed in the post-lockdown period. Among the cases selected for analysing the desire for a birth within two years, 43.1% were interviewed in the pre-lockdown period, and the remaining were interviewed in the post-lockdown period. In analysing contraceptive use, 43.3% of the selected cases were interviewed in the pre-lockdown period, and the percentage for those selected for analysing having sex in the last 30 days was 41.5%. Selected weighted cases by state and UT can be found in the supplementary S1 Table.

Among the women interviewed in the pre- and post-lockdown periods, more than 80% were Hindu. This high percentage of Hindu respondents was because of the fact that around 79% of people in India are Hindu [47]. In both pre- and post-lockdown periods, the survey interviewed around 70% of women from rural areas. The distributions of the selected cases among the subcategories of the characteristics in the pre- and post-lockdown periods showed similar trends. Detailed data on the cases with different background characteristics by interview period can be found in the supplementary S2 Table.

Results of the descriptive analysis

Table 1 shows that the percentage of the selected women desiring a birth within two years in the pooled data substantially increased to 52.4% after the lockdown from 48.6% recorded prior to the lockdown. Among the 13 selected regions, nine regions experienced an increase in the percentage of women desiring a birth within two years, while that percentage declined in the remaining regions. Among the selected regions, the percentages of women desiring a birth within two years in Rajasthan and Tamil Nadu increased substantially after the lockdown.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Fertility-related behaviours of the women by interview period in the selected Indian states, 2019/21.

https://doi.org/10.1371/journal.pone.0314800.t001

The percentage of the selected women using any contraception in the pooled data substantially declined from 68.8% in the pre-lockdown period to 67% in the post-lockdown period (Table 1). The contraceptive use rate in six of the selected regions declined after the lockdown, while that rate increased in the remaining seven regions after the lockdown. The contraceptive use rate in Jharkhand and Odisha declined substantially after the lockdown, whereas the rate in Punjab, Delhi, and Arunachal Pradesh increased substantially after the lockdown.

As it is seen in Table 1, the rate of women having sex in the last 30 days among those interviewed after the lockdown (32.7%) in the pooled data was noticeably lower than that among those interviewed before the lockdown (37.3%). This rate in the nine selected regions declined after the lockdown, while that in the remaining four states increased after the lockdown. Among the selected states, the rates of having sex in Punjab and Madhya Pradesh substantially declined after the lockdown.

Results of the mixed effects logit model

Mixed-effects logit models were fitted to estimate the net effects of the COVID-19 lockdown on fertility-related factors by controlling for relevant variables. It is seen from Table 2 that the odds of desiring another birth within two years among the pooled women interviewed in the post-lockdown period (OR = 1.20, p<0.001) was 20% higher than that among those interviewed in the pre-lockdown period. The women interviewed in the post-lockdown period in Rajasthan (OR = 1.18, p<0.05), Arunachal Pradesh (OR = 1.41, p<0.05), Chhattisgarh (OR = 1.26, p<0.05), and Tamil Nadu (OR = 1.60, p<0.001) were also significantly more likely to desire another birth within two years than those who were interviewed in the pre-lockdown period in those regions (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Adjusted odds ratios (ORs) of desiring a birth within 2 years among the women interviewed in pre- and post-lockdown periods in the selected Indian states, 2019/21.

https://doi.org/10.1371/journal.pone.0314800.t002

Table 3 shows that in the pooled data, the odds of using contraception in the post-lockdown period (OR = 0.92, p<0.001) was 8.4% lower than that in the pre-lockdown period. The odds of using contraception in the post-lockdown period in Jharkhand (OR = 0.70, p<0.001), Odisha (OR = 0.85, p<0.01), and Uttar Pradesh (OR = 0.90, p<0.001) were significantly lower than that in the pre-lockdown period. While women interviewed in the post-lockdown period in Arunachal Pradesh (OR = 1.79, p<0.001), Delhi (OR = 1.92, p<0.001), Punjab (OR = 1.50, p<0.001), and Tamil Nadu (OR = 1.14, p<0.05) were significantly more likely to use contraception than those interviewed in the pre-lockdown period (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Adjusted odds ratios (ORs) of using contraception among the women interviewed in pre- and post-lockdown periods in the selected Indian states, 2019/21.

https://doi.org/10.1371/journal.pone.0314800.t003

Among the pooled women, the odds of having sex (in the last 30 days) in the post-lockdown period (OR = 0.78, p<0.001) was 22% lower than that in the pre-lockdown period (Table 4). Women in Punjab (OR = 0.64, p<0.001), Uttar Pradesh (OR = 0.87, p<0.05), Madhya Pradesh (OR = 0.70, p<0.001), and Tamil Nadu (OR = 0.75, p<0.05) were significantly less likely to have sex in the post-lockdown period than in the pre-lockdown period. In contrast, women in Haryana (OR = 1.43, p<0.01) were significantly more likely to have sex in the post-lockdown period than in the pre-lockdown period.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Adjusted odds -ratios (ORs) of having sex in the last 30 days among the women interviewed in pre- and post-lockdown periods in the selected Indian states, 2019/21.

https://doi.org/10.1371/journal.pone.0314800.t004

The F-statistics of all the fitted models that have been presented in Tables 2–4 are statistically significant, indicating the overall significance of all of the fitted models. Table 2 shows the cluster memberships in pooled data (ICC = 0.117), Uttarakhand (ICC = 0.0.135), Arunachal Pradesh (ICC = 0.225), Jharkhand (ICC = 0.137), Chhattisgarh (ICC = 0.128), and Madhya Pradesh (ICC = 0.072) showed a medium effect on desire for another birth, and that effects in the remaining states/UTs were small.

Table 3 shows that the effects of the cluster memberships on using contraception in pooled data (ICC = 0.138), Punjab (ICC = 0.153), Rajasthan (ICC = 0.167), Uttar Pradesh (ICC = 0.120), Arunachal Pradesh (ICC = 0.190), Odisha (ICC = 0.120), Chhattisgarh (ICC = 0.176), and Madhya Pradesh (ICC = 0.174) were medium, and that effects in the remaining states/UTs were small.

It is seen from Table 4 that the cluster memberships showed a medium effect on having sex in the last 30 days in pooled data (ICC = 0.115), Arunachal Pradesh (ICC = 0.111), Jharkhand (ICC = 0.107), Chhattisgarh (ICC = 0.128), and Puducherry (ICC = 0.229), and that membership showed a small effect in the remaining states/UTs.