Impact of pandemic-related worries on mental health in India from 2020 to 2022

Descriptive statistics

Table 1 presents a summary of both unweighted and weighted statistics for the variables within the full sample (N = 2,576,174 respondents aged ≥18 years in India). Regarding mental health outcomes, the weighted proportions of respondents experiencing feelings of depression and anxiety were 8% and 6%, respectively, in both periods. In terms of subjective worries, in Period 1, 21% of the respondents post-weighted reported financial stress, 8% reported food insecurity, and 24% reported COVID-19-related illness concerns. Period 2 showed similar weighted statistics (17% for financial stress; 8% for food insecurity), except that COVID-19-related illness concerns were discontinued from the survey at this time. Notably, the demographics of the CTIS respondents, even after applying weights (considering age and gender), did not align accurately with the general Indian adult population. Specifically, the weighted sample had a larger proportion of males (66% in Period 1 and 57% in Period 2, as opposed to 52% in the census), people aged between 25 and 34 years (37% in Period 1 and 31% in Period 2, compared to 27% in the census), people having education levels of at least high school (83% in Period 1 and 93% in Period 2, as opposed to 27% in the census), and people residing in urban areas (75% in Period 1 and 75% in Period 2, compared to 31% in the census).

We analyzed the unadjusted correlations between the outcomes and primary exposures using the full sample (N = 2,576,174) and the complete Likert scales (5-point scale for depression and anxiety, and 4-point scale for worry variables), without applying weights. For the outcomes, the responses for depression and anxiety were moderately correlated, with Kendall rank correlation coefficients of 0.59 in Period 1 and 0.58 in Period 2 (Supplementary Table 2). For the exposures, financial stress and food insecurity were moderately correlated, while both were weakly correlated with COVID-19-related illness concerns. For instance, the Kendall rank correlation between financial stress and food insecurity was 0.55 in Period 1, compared to 0.25 between financial stress and COVID-19-related illness concerns (Supplementary Table 2). Additionally, depression and anxiety showed weak correlations with pandemic-related worries. For example, the Kendall rank correlation between depression and financial stress was 0.20 in Period 1 and 0.24 in Period 2 (Supplementary Table 2). Similar patterns were observed with Pearson correlation coefficients (Supplementary Table 2). Pearson correlation coefficients showed upward inflation compared to Kendall rank correlation coefficients as expected.

Figure 1 illustrates the weekly weighted prevalence of depression and anxiety from June 27, 2020, to June 25, 2022, using the full sample (N = 2,576,174). The vertical dashed line marks the time of the major change in the survey structure. Notably, there were three peaks in both the weekly new cases and deaths during the study period, occurring approximately in September 2020, May 2021, and January 2022. During the second wave in May 2021, there was a noticeable increase in the prevalence of self-reported depression (10.11%) and to some extent for anxiety (7.61%), compared to January 2021 (7.58% for depression and 5.80% for anxiety). This shows a concurrent worsening in self-reported mental health outcomes during the peak of the second COVID-19 wave. However, this pattern was not as prominent during the first and third waves. When comparing the weighted proportions of individuals experiencing depression and anxiety at the start and end of the study period, the values in 2022 (8.29% for depression and 6.17% for anxiety) were higher than those observed in 2020 (7.24% for depression and 4.77% for anxiety). This underscores the enduring impact of the pandemic-related concerns on mental health within the Indian population.

The full sample was used to calculate the proportions of mental health outcomes. Facebook-provided weights were applied to account for non-response and coverage bias. A 7-day smoothing method was used to calculate the weighted averages. COVID-19 case and death data were sourced from the Johns Hopkins Coronavirus Resource Center. A vertical dashed line marks the major survey structure change on May 20, 2021.

Our gender-stratified analysis revealed notable distinctions in the weighted proportions. Regarding self-reported mental health problems, females showed a higher prevalence of depression than males in both Period 1 (10% vs. 6%) and Period 2 (9% vs. 8%), as well as increased anxiety in Period 1 (7% vs. 5%). Regarding the worry variables, males reported greater concerns about their financial situation (23% vs. 18% in Period 1; 19% vs. 14% in Period 2) and food status (9% vs. 6% in Period 1; 9% vs. 5% in Period 2) compared to females.

Likewise, we examined urban–rural differences in weighted proportions. In Period 1, the incidence of self-reported depression and anxiety among rural residents mirrored that of urban dwellers (8% vs. 8% for depression and 5% vs. 5% for anxiety), whereas Period 2 saw a higher occurrence in rural areas (9% vs. 8% for depression and 7% vs. 5% for anxiety). Concerning pandemic-related worries, rural individuals consistently reported greater levels of financial stress (28% vs. 19% in Period 1; 25% vs. 15% in Period 2) and food insecurity (12% vs. 7% in Period 1; 13% vs. 6% in Period 2) compared to urban residents.

Associations between worry variables and mental health outcomes

Table 2 displays the odds ratios for worry variables (considered as exposures) obtained from weighted logistic regression analyses. The analysis was based on complete cases from two distinct periods: Period 1 (N = 595,229) and Period 2 (N = 152,767). Here, we first report results from the fully adjusted models that concurrently incorporated all worry variables and other covariates, namely demographic factors and calendar time. The results from the fully adjusted models in Period 1 indicate significant associations between all three subjective worries (financial stress, food insecurity, and COVID-19-related illness worries) and mental health problems (both depression and anxiety). Notably, financial stress emerged as the most influential factor among the three worries. In particular, the odds of experiencing depression and anxiety were 2.36 (95% confidence interval, CI: [2.27, 2.46]) and 1.91 (95% CI: [1.81, 2.01]) times higher for those concerned about their financial situation compared to those without such worries, after adjusting for other variables, respectively. Individuals reporting food insecurity had higher adjusted odds for depression (odds ratio: 1.45, 95% CI: [1.38, 1.53]) and anxiety (odds ratio: 1.56, 95% CI: [1.47, 1.66]) compared to those without this concern. Similarly, the adjusted odds ratios regarding COVID-19-related illness worries were 1.65 (95% CI: [1.59, 1.71]) for depression and 1.81 (95% CI: [1.73, 1.90]) for anxiety.

As depicted in the second part of Table 2, both financial stress and food insecurity had significant associations with mental health problems in the fully adjusted models in Period 2, mirroring findings from Period 1. Particularly, after accounting for all other variables, individuals experiencing financial stress had higher odds of feeling depressed (odds ratio: 2.52, 95% CI: [2.30, 2.77]) and anxious (odds ratio: 1.91, 95% CI: [1.70, 2.15]) compared to those without financial stress. Likewise, individuals reporting concerns about food insecurity exhibited significantly elevated adjusted odds of experiencing depression (odds ratio: 2.32, 95% CI: [2.07, 2.60]) and anxiety (odds ratio: 2.71, 95% CI: [2.31, 3.11]) compared to those without this concern.

In addition to the fully adjusted models, Table 2 presents outcomes from two other model types for each worry variable: unadjusted models, which only include a single exposure, and partially adjusted models, which include one exposure at a time together with demographic factors and calendar time. The comparisons across the three models for both periods highlight differences in the results. Notably, the odds ratios for all worry variables, consistently significant across models, were higher in the unadjusted models for both depression and anxiety. These values diminished in the partially adjusted models and were further reduced in the fully adjusted models. Taking financial stress as an example, in Period 1, individuals reporting financial stress had a 3.45 (95% CI: [3.34, 3.57]) times higher odds of experiencing depression compared to those without such worries, as indicated by the unadjusted model. This odds ratio decreased to 3.03 (95% CI: [2.93, 3.14]) with adjustments for demographics and calendar time, and further dropped to 2.36 (95% CI: [2.27, 2.46]) with additional adjustments for food insecurity and concerns about COVID-19-related illness. The ranking of the worry variables by their odds ratios also varied depending on whether adjustments were made, even within the same period and for the same outcome. For instance, in the first period, food insecurity showed the highest odds ratios for experiencing anxiety in both unadjusted (odds ratio: 3.29, 95% CI: [3.14, 3.45]) and partially adjusted models (odds ratio: 2.98, 95% CI: [2.83, 3.13]). Yet, in the fully adjusted models, the odds ratio for food insecurity (odds ratio: 1.56, 95% CI: [1.47, 1.66]) was lower than that for financial stress (odds ratio: 1.91, 95% CI: [1.81, 2.01]).

Associations between other demographics covariates and mental health outcomes

Supplementary Table 3 and Supplementary Table 4 present odds ratios of other covariates derived from weighted logistic regression models for both unadjusted models, featuring a single covariate, and fully adjusted versions that incorporate a complete set of relevant covariates, including the collective inclusion of all three worry variables. In Period 1, the analysis of fully adjusted models, as presented in Supplementary Table 3, revealed significant relationships between demographic factors—such as gender, age, urban–rural residential status, geographical region, and occupation—and the prevalence of depression or anxiety. Key findings indicate that factors associated with poorer mental health outcomes include being female, being in the young adult age group, residing in urban areas, hailing from the southern regions, and engaging in occupations outside the agricultural sector. Notably, after controlling for other variables, individuals employed in the education sector had 1.23 (95% CI: [1.14, 1.34]) and 1.30 (95% CI: [1.18, 1.43]) times higher odds of reporting depression and anxiety, respectively, compared to those in agriculture. Similarly, those working in the tourism industry faced 1.55 (95% CI: [1.36, 1.77]) and 1.46 (95% CI: [1.24, 1.71]) times higher odds of experiencing depression and anxiety, respectively, in comparison to agricultural workers.

In Period 2, the influence of demographic variables on depression and anxiety mirrored the results observed in Period 1, with significant associations identified for gender, age, residential status, geographic region, and occupation after adjusting for other factors (Supplementary Table 4). This period also highlighted the significance of education, a factor not previously noted in Period 1. Notably, the fully adjusted models showed that individuals without a high school diploma had a 1.24 (95% CI: [1.11, 1.39]) and 1.17 (95% CI: [1.04, 1.32]) times higher odds of experiencing depression and anxiety, respectively, compared to individuals with at least a high school education.

Effect modification by gender on the associations between worry variables and mental health outcomes

Our analysis explored potential modification of the associations between three pandemic-related worries on depression and anxiety by self-reported gender, by adding an interaction term between gender and each of the worry variables separately into our fully adjusted survey-weighted model. Figure 2 and Supplementary Table 5 illustrate the odds ratios for depression and anxiety across different worry variables within each gender, after adjusting for other worries, demographic factors, and calendar time. Throughout both periods, every pandemic-related worry variable demonstrated a significant association with mental health problems in both genders from the fully adjusted survey-weighted models. In Period 1, females had a significantly lower odds ratio of reporting depression and anxiety compared to males regarding the same concern (P_int <0.01, where P_int denotes the P value corresponding to the gender by worry interaction term in the fully adjusted model). Specifically, in terms of concerns about financial status, the odds ratios for feeling depressed and anxious were 2.03 (95% CI: [1.88, 2.19]) and 1.74 (95% CI: [1.58, 1.91]) for females, respectively, lower than the corresponding odds ratios for males, which were 2.59 (95% CI: [2.49, 2.70]) and 2.02 (95% CI: [1.92, 2.12]). Similarly, the odds ratios regarding food insecurity for depression and anxiety were 1.21 (95% CI: [1.07, 1.36]) and 1.32 (95% CI: [1.16, 1.51]) for females, respectively, again lower than those for males, who showed odds ratios of 1.56 (95% CI: [1.49, 1.64]) for depression and 1.67 (95% CI: [1.57, 1.77]) for anxiety. In Period 2, the smaller sample size resulted in wider confidence intervals, leading to no statistically significant gender differences in the odds ratios for the worry variables (for depression, P_int = 0.17 in finance stress and P_int = 0.43 in food insecurity; for anxiety, P_int = 0.34 in finance stress and P_int = 0.70 in food insecurity), despite varying point estimates.

Due to the significant survey structure change on May 20, 2021, the analysis was split into two periods: Period 1 (June 27, 2020, to May 19, 2021) and Period 2 (May 20, 2021, to June 25, 2022). Complete cases from each period were analyzed using separate fully adjusted models that included pandemic-related worries (financial stress, food insecurity, and COVID-19-related illness concerns), demographics (gender, age, education, urban–rural residential status, region, and occupation), and calendar time (categorized by month and year), with each model also adding an additional interaction term between gender and one of worry variables. Presented are odds ratios and their corresponding 95% Wald confidence intervals. Significant interaction terms (P_int) from Wald tests (significance level: 0.05) are highlighted in red. All models integrate survey weights within the logistic regression. Robust sandwich estimators were used for variance calculation. The vertical dashed line denotes the odds ratio of 1.

Effect modification by urban–rural residential status on the associations between worry variables and mental health outcomes

We also examined how urban and rural residents were differently affected by three specific pandemic-related worries in terms of self-reported depression and anxiety. We added an interaction term for residential status and each worry variable separately into the fully adjusted model, accounting for survey weights. Figure 3 and Supplementary Table 6 display the odds ratios for depression and anxiety associated with each worry variable within urban and rural settings, after controlling for other worries, demographic factors, and calendar time. Consistently, across both urban and rural contexts, each worry variable was significantly linked to mental health challenges. In Period 1, urban dwellers exhibited greater odds ratios of experiencing depression and anxiety regarding all of the three COVID-19-related worries than rural residents, with statistically significant interaction terms (P_int <0.01) observed in all models. For instance, in terms of financial stress, urban individuals showed odds ratios of 2.44 (95% CI: [2.33, 2.56]) for depression and 1.98 (95% CI: [1.87, 2.09]) for anxiety, surpassing the odds ratios for rural residents, which stood at 2.12 (95% CI: [1.98, 2.26]) and 1.71 (95% CI: [1.58, 1.86]), respectively. In Period 2, urban residents continued to report significantly higher odds of depression and anxiety regarding these worries when compared to rural residents (for depression, P_int <0.01 in finance stress; for anxiety, P_int <0.01 in finance stress and P_int = 0.03 in food insecurity) expect the case with food insecurity for depression (P_int = 0.31). For example, in terms of worries about the financial situation, the odds ratios for urban individuals were 2.80 (95% CI: [2.52, 3.12]) for depression and 2.19 (95% CI: [1.92, 2.50]) for anxiety, which were higher than those for rural individuals, with odds ratios of 1.98 (95% CI: [1.73, 2.26]) for depression and 1.41 (95% CI: [1.20, 1.66]) for anxiety).

Due to the significant survey structure change on May 20, 2021, the analysis was split into two periods: Period 1 (June 27, 2020, to May 19, 2021) and Period 2 (May 20, 2021, to June 25, 2022). Complete cases from each period were analyzed using separate fully adjusted models that included pandemic-related worries (financial stress, food insecurity, and COVID-19-related illness concerns), demographics (gender, age, education, urban–rural residential status, region, and occupation), and calendar time (categorized by month and year), with each model also adding an additional interaction term between residential status and one of worry variables. Presented are odds ratios and their corresponding 95% Wald confidence intervals. Significant interaction terms (P_int) from Wald tests (significance level: 0.05) are highlighted in red. All models integrate survey weights within the logistic regression. Robust sandwich estimators were used for variance calculation. The vertical dashed line denotes the odds ratio of 1.

Temporal variation in the associations between worry variables and mental health outcomes

We further delved into how the impact of the three subjective worries on depression and anxiety evolved over time by introducing interaction terms for time (categorized by month and year) with each worry variable individually into our fully adjusted survey-weighted models. Figure 4 presents the adjusted odds ratios for financial stress, food insecurity, and COVID-19-related concerns in relation to feeling depressed and anxious across time, in both periods. Vertical dashed lines represent the timing of the three peaks in newly confirmed COVID-19 cases and COVID-19-related deaths (September 2020, May 2021, and January 2022). During Period 1, financial stress emerged as the most influential factor affecting both depression (the range of odds ratios: [2.02, 2.71]) and anxiety (the range of odds ratios: [1.59, 2.19]), after adjusting for other variables. Although no discernible temporal patterns were evident for financial stress and food insecurity, there was a notable upswing in the odds ratios regarding COVID-19-related health concerns of reporting depression (odds ratio: 2.13, 95% CI: [1.90, 2.39]) and anxiety (odds ratio: 2.25, 95% CI: [1.98, 2.56]) during the second wave in May 2021 compared to the previous month (odds ratio 1.62, 95% CI: [1.46, 1.80] for depression and odds ratio: 1.73, 95% CI: [1.54, 1.94] for anxiety). During Period 2, the interpretation of time trends is constrained due to the smaller sample size, resulting in wider confidence intervals compared to Period 1. For anxiety, the point estimates for the effects of food insecurity (the range of adjusted odds ratios: [2.34, 3.75]) consistently surpassed those for financial stress (the range of adjusted odds ratios: [1.60, 2.43]) over time, while no significant differences were observed between these two sources of worry concerning depression.

Due to the significant survey structure change on May 20, 2021, the analysis was split into two periods: Period 1 (June 27, 2020, to May 19, 2021) and Period 2 (May 20, 2021, to June 25, 2022). Complete cases from each period were analyzed using separate fully adjusted models that included pandemic-related worries (financial stress, food insecurity, and COVID-19-related illness concerns), demographics (gender, age, education, urban–rural residential status, region, and occupation), and calendar time (categorized by month and year), with each model also adding an additional interaction term between time and one of worry variables. Presented are odds ratios and their corresponding 95% Wald confidence intervals. All models integrate survey weights within the logistic regression. Robust sandwich estimators were used for variance calculation. Vertical dashed lines denote the three peaks in new COVID-19 confirmed cases and related deaths in September 2020, May 2021, and January 2022, respectively. Due to limitations in sample size, data from June 2020 were merged with the June 2021 category.