Study population

India is a large country with six broad regions—North, North-East, South, West, East and Central India. To ensure representation of the regional diversity of the country, the study population was determined by grouping the 29 Indian states into six categories based on a set of economic and health indicators. One state from each category was chosen for our sample.1 Specifically, we have selected six states considering per capita net state domestic product,34 child mortality (under 5 years old), life expectancy at birth, schooling and the social subclass of scheduled caste and scheduled tribe (SC/ST) population. The sampled population is drawn from six states: Uttar Pradesh, Odisha, Tamil Nadu, Haryana, Rajasthan and Maharashtra (figure 1). These six states together have around half (48%) of the total population of India.35

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

The six states included in the sample design.

Table 1 shows the variation of our selection parameters of human development across the selected states and India including population growth, disability-adjusted life-years (DALYs), literacy rates and proportion of SC/ST, that is, socioeconomically disadvantaged groups. The six study states are at different stages of demographic and epidemiological transition. For instance, as per the Registrar General of India,36 the natural growth rate of the population varies from 18.7% (crude birth rate, CBR=25.1, crude death rate, CDR=6.1) (CBR: the number of births in a year per thousand population; CDR: the number of deaths in a year per thousand population.) in Uttar Pradesh to 7.7% (CBR=13.8, CDR=6.1) in Tamil Nadu. Additionally, the value for the ratio of DALYs due to communicable, maternal, neonatal and nutritional diseases over DALYs due to non-communicable disease and injuries varies across these states: 0.26 in Tamil Nadu, 0.33 in Maharashtra, 0.40 in Haryana, 0.66 in Rajasthan, 0.58 in Odisha and 0.68 in Uttar Pradesh.37

Household as the sampling unit

An important feature of this study is using the household as a sampling unit and interviewing all eligible (aged 18 years and older) members of the household. The motivation for considering a household rather than an eligible individual from the household lies in the contagious nature of the COVID-19 disease, post-COVID symptoms and the potential impact of changes in socioeconomic circumstances due to the pandemic. We categorise a household as a COVID-19 positive household (henceforth, COVID-positive household) if at least one member of the household reports having a history of COVID-19.

There are several advantages of the general population-based sampling approach. First, this approach allows us to assess the socioeconomic gradient in post-COVID conditions for both the overall population and the subgroup of COVID-positive households. Second, since we analyse the association between multidimensional deprivation and experiencing COVID-19 and post-COVID, having non-COVID households in the sample allows us to benchmark our analysis against the COVID households. Third, given the highly contagious nature and rapid transmissibility of this disease, there is a high likelihood of intrahousehold transmission of COVID-19. With households losing family members, jobs and income due to COVID-19, experiencing post-COVID by an individual and/or other members of households could have vicarious effects on the health and well-being of all members of the household. It is critical to understand the variation in experiencing post-COVID symptoms and HRQoL of the members within a COVID-19 household (ie, intrahousehold HRQoL). Our approach to studying HRQoL for each member allows us to explore the vicarious mental and physical health effects of post-COVID within the household. Fourth, the impact of the pandemic on household’s socioeconomic status would vary by the severity of disease incidence, that is, the number of household members experiencing post-COVID symptoms. For instance, the socioeconomic circumstances of a COVID-positive household where only one member lost their job due to COVID-19 and is now experiencing post-COVID would be different from that of a household where only one household member lost their job and multiple members of the household are experiencing post-COVID. Since the possibility of more than one family member of a COVID-positive household experiencing post-COVID is expected to be high, the impact on the HRQoL and well-being of the entire household will also be high. Finally, while most existing data on COVID-19 pandemic has collected data for COVID-19 survivors only, this population-based study will collect data on socioeconomic circumstances, COVID-19 history, post-COVID and HRQoL for the general population in India.

Sample size

The sample size is determined based on the level of confidence and statistical power analysis following Cohen’s38 guidelines, along with the effect sizes reported in recent literature.29 39–42 Our sample size accounts for the effect size (ie, the difference between the highest and the lowest level needed to detect a meaningful difference) between the HRQoL reported by deprived and not deprived groups with COVID-19 and post-COVID. The power calculation assumed a small effect size, 95% CI and a statistical power of 90%. Based on F-test analysis of variance (fixed effect, main and interaction effects, α err prob=0.05, power=0.90, effect size=0.1 (small)), a sample size of 2063 was suggested (The power analysis was carried out using the G*Power V.3.1.9.7 software). This provides sufficient power to detect statistical effects in multiple regression models.43–45 The sample size is also validated using the effect sizes reported in recent literature for Middle Eastern and North Africa (MENA) countries,39 Hong Kong,40 India,29 Asia41 and China.42 The project will recruit a sample of 3000 randomly selected households, representative across age, gender and education (≥12 000 individuals based on an average household size of 4.7 in India), through Morsel India, a field survey company in India. All members of the household (>18 years old) will be interviewed. Following the norm for survey data collection in household sample surveys in India,46 we aim for a 70% response rate across waves. Assuming a 30% attrition rate, which aims to collect a 70% response rate for household sample surveys in India, an initial sample of 3000 households will suffice for a final sample of at least 2063 households.

Sampling strategy

The sample design will follow the statistical approaches recommended for designing household surveys. Henderson and Sundaresan46 demonstrated that 30 clusters, that us, primary sampling units (PSUs), selected based on probability proportional to size (PPS) from strata (Henderson and Sundaresan46 created 12 hypothetical stratums by considering different levels of immunisation coverage in communities (ie, PSUs)) of a geographical unit (ie, state), provide an estimate of immunisation coverage with 95% confidence limits. Following Henderson and Sundaresan,46 we will adopt a two-stage sampling design strategy to select PSUs in both rural and urban areas of a state. A total of 30–32 PSUs in a state will be allocated and selected from rural and urban areas according to population proportions. Before selection, all PSUs with fewer than 40 households will be linked to the nearest PSU. We will form separate strata in rural and urban areas for household selection in a state.

Second stage selection

Sample households will come from each selected PSU. Selected PSUs with more than 150 households will be divided into segments (We will follow geographical boundaries such as waterways, roads and streets) containing 60–100 households. One segment will be randomly selected for mapping and listing operation in the PSU. A complete mapping of listing of each household in the PSU will be done to select households for interview. From each of the selected PSUs or segment of a PSU, 20 households will be randomly selected with systematic sampling.

Survey design

The survey will be conducted four times over 2023 and 2024, at equally spaced intervals, starting with winter of 2023 and ending in winter of 2024. The survey data collection team at Morsel India will conduct the survey in consultation with the project team. The survey questionnaire will be developed and conducted using computer-assisted personal interview software SurveyCTO (https://www.surveycto.com/). All interviewers will be trained by researchers who directly work with us. Each survey will take approximately 40–50 min using the official EQ-5D-5L (EuroQol health-related quality of life questionnaire) local language translations approved by the EQ-5D legal team. To maximise participation across all four surveys, phone numbers of respondents will be collected for the purpose of scheduling follow-up interviews. Respondents will be given a gift voucher (worth €2), at the end of follow-up surveys.47 Attrition and non-response in the first follow-up will be appropriately tested following the best practices in the literature.48

The objective of collecting data over four waves (including one prepandemic recall) is twofold: (a) to monitor and evaluate the socioeconomic circumstances experienced by the population amid high rates of COVID-19 and post-COVID conditions and (b) to apply our theoretical framework for measuring of multidimensional deprivation,24 which necessitates multiple waves to accurately measure the persistence and severity of multidimensional deprivation. Furthermore, since most post-COVID symptoms last at least 3 months, collecting data at equally spaced intervals will allow us to closely track health patterns and disease symptoms.

Survey questionnaire

Each survey wave will collect data on the same set of individual and household-level socioeconomic circumstances, HRQoL, COVID-19 history and vaccination information. Detailed information on the data to be collected is presented in table 2. All members of the household (≥18 years old), including the elderly, disabled and low literacy or illiterate members, will be interviewed. The core dataset will include respondents’ age, gender, educational attainment, employment status, occupation, social caste, household size, income and material well-being. Material well-being encompasses housing conditions, access to electricity, fuel, clean water, sanitation, vehicles, land and house ownership. We will also inquire about schooling, financial hardship during COVID-19 and social engagement. Respondents will be asked to report their HRQoL using the EQ-5D-5L instrument.49 The resulting health profiles will be translated into index scores using the Indian EQ-5D-5L value set.50 Additionally, respondents will answer questions on chronic pre-existing health conditions, COVID history, smoking status (following the Fagerstrom test51), alcohol consumption (using the Alcohol Use Disorder Identification Test52), disability status (using the WHO disability assessment tool53) and anxiety (using the COVID-19 anxiety scale54). For respondents reporting a COVID-19 history, we will inquire about their post-COVID symptoms as established by the WHO.52 Focusing on the overall population, rather than only COVID-positive households, aligns with the sampling approach used in ongoing studies in the USA and the UK; specifically, studies such as the RECOVER project (USA)55 and the EuroQol-funded POPCORN.56 57 The POPCORN study investigated the direct and indirect effects of the COVID-19 pandemic on HRQoL in Greece, Italy, the Netherlands, the UK and the USA.

The first wave of the IndiQoL survey will include questions requiring respondents to recall their health and socioeconomic circumstances prior to the onset of the COVID-19 pandemic. To establish a common reference point, we will use Diwali 2019, a festival celebrated across the entire country, as the time point in the year before the COVID-19 pandemic began. To address potential bias associated with retrospective measurement, we will conduct a test–retest analysis using both retrospective and prospective data, following recent literature. Specifically, we will use the test for concordance and Bland-Altman test.

Ensuring data quality is crucial. The survey company and our research team will independently monitor data quality after every survey wave. We will record the number of households contacted, the acceptance or rejection of participation invitations, and, where known, the reason for non-participation. We will also quantify the rate of non-response for each of the items of the questionnaire.

Patient and public involvement

There is no patient and public involvement in the study.

Empirical analysis

The primary objective of estimating the average impact of COVID-19 and post-COVID conditions on HRQoL, as measured by EQ-5D-5L, will be achieved by comparing individuals with post-COVID symptoms to two reference groups: (a) those who have never had COVID-19 and (b) those who have recovered from an acute infection. Since individuals can experience multiple symptoms that may persist, fluctuate or relapse over time,52 longitudinal EQ-5D-5L data will be particularly valuable for characterising and capturing the longitudinal pattern of HRQoL effects. Our framework will analyse the individual EQ-5D-5L domain responses using ordered logistic regression. To account for observed respondent characteristics (such as age and gender) and prepandemic HRQoL (recall), we will employ either a matching approach or regression adjustments. The EQ-5D-5L index scores will use the value sets recently developed for India.33 The longitudinal nature of the data also allows us to control for some unobserved confounding effects through fixed effects models. To address potential missing values in multiple responses or covariates in longitudinal clustered data, we will use imputation methods.58 Additionally, to capture the heterogeneity of the impact of COVID-19,59–61 we will conduct stratified analysis by age, sex and across six states to provide further insights into the domain wise impact of the number of post-COVID symptoms on HRQoL.

The second objective of this study is to evaluate how disadvantage in multiple domains of well-being affects the HRQoL impact of COVID-19 and post-COVID. We consider several domains based on the global multidimensional deprivation indicators.6 Specifically, the domains include income, schooling, sanitation, electricity, housing conditions, unemployment, social interaction, social capital (trust, security), access to transport, access to healthcare, domestic violence and abuse, food insecurity, social isolation, indoor air pollution and clean drinking water. This objective will use an algorithm to measure persistence augmented multidimensional deprivation developed by Nicholas et al26 and Sinha et al.24 Specifically, we adopt a regression framework to estimate HRQoL values (index scores; domain responses) as a function of variables (including multidimensional deprivation, age, gender, household size, location, health behaviours, post-COVID status). We use generalised linear models (where the dependent variable is related to covariates through a specific link function) for the EQ-5D-5L value scores and ordered logit regressions for domains. We will not only evaluate how overall disadvantage impacts the HRQoL but also explore the specific role of severity and persistence of deprivation in each domain. Using the Shapley decomposition method,62 we will compute the proportional contribution of the components of multidimensional deprivation (ie, severity and persistence) on the HRQoL effect associated with post-COVID. Separately accounting for these components in our regression analyses will allow us to identify how severity and persistence in specific dimensions of deprivation have affected HRQoL for post-COVID. Further ranking of subgroups (gender, rural/urban, COVID-19 status) based on this decomposition technique will allow clear identification of COVID-sensitive dimensions of deprivation specific to each subgroup that policy could target.

Limitations

The survey will be lengthy as it will collect household-level and individual-level self-reported data, which can result in measurement error due to self-reporting bias. The longitudinal nature of this survey would result in some attrition in later waves of the survey, which is anticipated to affect the analysis and results of the study. The first wave of survey includes a module which requires respondents to recall their pre-COVID-19 HRQoL and socioeconomic circumstances. As these events are almost 3–4 years old, there is a possibility of recall bias.