Mental Health Impacts of COVID-19: Does Prepandemic Cognition and Dementia Status Matter?

Abstract

Background

The coronavirus disease 2019 (COVID-19) pandemic disrupted daily life and led to sharp shocks in trends for various health outcomes. Although substantial evidence exists linking the pandemic and mental health outcomes and linking dementia and mental health outcomes, little evidence exists on how cognitive status may alter the impact of COVID-19 on mental health.

Methods

We used prepandemic data from the Longitudinal Aging Study in India-Diagnostic Assessment of Dementia study and 9 waves of data from the Real-Time Insights of COVID-19 in India study (N = 1 182). We estimated associations between measures of prepandemic cognition (continuous cognition based on 22 cognitive tests, dementia status) and mental health measures during the pandemic (Patient Health Questionnaire [PHQ]-4 [9 time points], PHQ-9 [2 time points], Beck Anxiety Inventory [3 time points]), adjusting for age, gender, rural/urban residence, state, education, and prepandemic mental health.

Results

Summarizing across time points, PHQ-9 score was marginally or significantly associated with prepandemic cognition (PHQ-9 difference: −0.38 [−0.78 to 0.14] points per SD higher cognition; p = .06), and prepandemic dementia (PHQ-9 difference: 0.61 [0.11–1.13] points for those with dementia compared to no dementia; p = .02). Associations with BAI were null, whereas associations with PHQ-4 varied over time (p value for interaction = .02) and were strongest during the delta wave, when pandemic burden was highest.

Conclusions

We present initial evidence that mental health impacts of COVID-19 or other acute stressors may be unequally distributed across strata of cognitive outcomes. In dynamically changing environments, those with cognitive impairment or dementia may be more vulnerable to adverse mental health outcomes.

The coronavirus disease 2019 (COVID-19) pandemic altered the lives of people globally, with unprecedented spread and accompanying waves of mortality. According to the World Health Organization, there have been over 769 million confirmed cases and over 6.9 million deaths reported worldwide, through August 2023 (1). Although public health measures, such as lockdowns and business closures, were enacted in early stages of the pandemic to slow the spread of the disease, evidence quickly grew on the impacts of these stringent policies on mental health (2).

Some studies highlighting high levels of depression and anxiety in early stages of the lockdown periods found that there was a recovery shortly after, suggesting resilience and adaptation to the circumstances (3–6). In contrast, others have found symptoms of anxiety, depression, and stress increased over time (7–9). Heterogeneity across studies could be attributed to differences between time periods and the lack of research conducted during the later stages of the pandemic, highlighting the need for evaluation of the long-term mental health impacts of COVID-19. Although the COVID-19 pandemic will likely have broad, long-lasting effects on trends in mental health and other health outcomes, more research is needed to better characterize and predict these potential consequences.

Current evidence suggests differential shocks of the pandemic on mental health in various populations. Older adults have been found to be particularly vulnerable, given their risk of mortality, illness severity, and other comorbidities such as cognitive impairment (10). Prior research found that older age was associated with greater COVID-19 worry (11), possibly due to the greater risk of negative outcomes. Although there has been mixed evidence on whether older adults are at higher or lower risk of developing symptoms of depression and stress than younger or middle-aged adults (12–14), overall findings still suggest a deterioration of mental health and inequalities in experiences for vulnerable older adult populations including socially isolated individuals and those with cognitive impairment (15). With population aging, the likelihood of persons living with cognitive impairment and dementia increases. Evaluation of mental health outcomes in this group is important, as symptoms of anxiety and depression frequently occur in individuals with mild cognitive impairment (MCI) and dementia. There is some evidence of a bidirectional association, with prior research suggesting both that depression may be a risk factor for dementia (16,17), and that persons with dementia are at a higher risk of having depression as compared to persons without dementia (18–20). Regardless of the causal direction of the link, given the higher prevalence of depression among those with MCI and dementia, cognitive status should be considered when investigating the mental health experiences of older adults.

Despite overwhelming literature linking the pandemic to mental health outcomes as well as the documented relationship between cognitive status and mental health, limited evidence exists on how cognitive status may alter the impact of COVID-19 on mental health. Studies that have investigated this association found worsening symptoms of anxiety and depression during the COVID-19 pandemic in individuals with dementia (21,22). Further, lockdowns and unpredictable public health measures appeared to negatively affect the mental health and physical well-being of persons with dementia over time (23–25). However, the evidence base is sparse, highlighting a need for expanded evaluation of this vulnerable population to mitigate their risk of poor mental health outcomes and better understand how the COVID-19 pandemic may alter long-term trends in health outcomes and comorbidities among this population.

We aim to contribute to this literature by assessing whether cognitive status alters the impact of the COVID-19 pandemic on mental health symptoms among older adults in India, a setting that experienced a nationwide lockdown, followed by extremely high levels of infections and mortality during the delta and omicron waves. Using a nationwide longitudinal cohort, we evaluate how prepandemic cognitive functioning and dementia affected mental health burden during the COVID-19 pandemic, both overall, and during different time periods.

Method

Sample

We used data from the Longitudinal Aging Study in India (LASI; N = 73 408), the LASI-Diagnostic Assessment of Dementia (LASI-DAD) Study (N = 4 096), and the Real-Time Insights-COVID-India Study (RTI-COVID; N = 3 797). LASI is a nationally representative sample of adults 45 years of age and older and their spouses and serves as the sampling frame for LASI-DAD, which includes adults 60 years of age and older (26,27). The RTI-COVID study used LASI-DAD households with valid phone numbers as a sampling frame, but recruited new participants (1 woman and 1 man over age 18 from each household) in initial waves. In subsequent waves efforts were made to include all LASI-DAD participants as well (28). The RTI-COVID survey includes 9 waves of data spanning May 2020 to May 2022. Surveys were administered approximately every 2 months with slightly longer breaks during the delta and omicron waves due to challenges with data collection during these time periods (Figure 1). Mobile phone credits were offered to respondents as incentives to increase participation and reduce sample selection. Protocols were approved by institution-specific Institutional Review Boards for all participating institutions, and all participants provided informed consent.

In this study, we used data from participants in all 3 surveys (N = 1 191). We excluded individuals with missing data on all mental health outcomes (N = 5) and missing data on caste (N = 4). To maximize sample size, individuals with data available on one or more of the included pandemic and prepandemic mental health measures were included in analyses for which complete data were available. Therefore, the final analytic sample size varies somewhat across analyses (N = 886–1 178; Supplementary Figure 1).

Assessment of Mental Health Outcomes

We used data from the Patient Health Questionnaire (PHQ)-4 and PHQ-9 (29,30). The PHQ-9 was designed to measure depression and is strongly related to clinical measures of dementia, with prior work reporting an Area Under the Curve (AUC) of 0.95 (30). The PHQ-4 combines 2 items from the PHQ-9 with 2 items from the Generalized Anxiety Disorder (GAD)-7 to comprise an ultrabrief measure of depression and anxiety (29). Prior work has found that the PHQ-4 has acceptable AUCs for both anxiety (0.84) and depression (0.79) (31). The PHQ-4 was administered at all 9 waves of the RTI-COVID survey, whereas the PHQ-9 was administered only at waves 3 and 6. The PHQ asks participants about how often they are bothered by symptoms over the previous 2 weeks; response options include not at all, several days, more than half the days, and nearly every day. Additionally, we used data from the 5-item version of the Beck Anxiety Inventory (BAI), which was administered at waves 5, 7, and 9 (32). The BAI asks respondents about the frequency with which participants experienced feeling specific symptoms over the prior week with response options including never, hardly ever, some of the time, and most of the time. The BAI has been shown to have high internal consistency and good convergent and discriminant validity (32).

Assessment of Prepandemic Cognitive Functioning and Dementia

We used previously derived scores of general cognitive functioning based on a large battery of neuropsychological tests administered in the LASI-DAD study, covering domains including orientation, memory, language, executive functioning, and visuospatial functioning (full list included in Supplementary Materials) (33). Cognitive tests were adapted from tests used in the Harmonized Cognitive Assessment Protocol battery, with some modifications necessary to ensure adequate performance in populations with low education and literacy. Cognitive scores were estimated from confirmatory factor analysis models following Cattell-Horn-Carroll theory, which provides a comprehensive framework for organizing cognitive abilities into broad domains, narrow domains, and general sources. Scores are scaled such that the distribution of scores in the full LASI-DAD sample has a mean of 0 and standard deviation of 1.

We used classifications of dementia from an algorithm designed to replicate the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5) classifications of minor and major neurocognitive disorder in the LASI-DAD study (34,35). Briefly, the algorithm defined domain-specific cognitive impairment using a robust neuropsychological norms approach. A normative sample without functional limitations or other exclusionary criteria was used to estimate cognitive functioning in the absence of disease; we defined objective cognitive impairment using comparisons between demographically matched individuals from within and outside the normative sample. Domain-specific functioning in memory, language, executive functioning, and visuospatial ability was considered. Informant-rated cognitive decline was ascertained using the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) (36) and informant-rated functional decline in activities of daily living and instrumental activities of daily living were measured using the Blessed Dementia Rating Scale (37). Additional details on the implementation of the algorithm are available in Supplementary Materials. Due to small numbers of individuals with major neurocognitive disorder (dementia) in the current study sample, we collapsed the minor and major neurocognitive disorder categories into a single group, which we refer to as DSM-5 neurocognitive disorder.

As an alternative classification of dementia, we also used data from an online clinical consensus evaluation of dementia using the Clinical Dementia Rating Scale (CDR), which was conducted among a subset of the LASI-DAD sample (38). We combined all CDR ratings greater than 0 into a single category representing MCI, questionable dementia, and dementia (referred to as MCI/dementia). Because the online clinical consensus process was conducted for a subset of the LASI-DAD sample, available sample sizes for these analyses were slightly smaller (Supplementary Figure 1).

Covariates

We used data on self-reported age, gender, educational attainment (none/less than secondary/secondary/some graduate), caste (no caste or other caste/scheduled caste/scheduled tribe/other backwards class), and marital status (married or partnered/other). Additionally, we considered prepandemic mental health using data from the Composite International Diagnostic Interview (CIDI) (39), which was administered as part of the main LASI study. We used the total symptom score as a continuous measure of mental health symptomatology.

We used self-reported measures of COVID-19 infection collected in waves 6–8. Given the lack of available testing during the pandemic, we classified those who either reported testing positive for COVID-19 or suspected having a COVID-19 infection if they were not tested, as having had COVID-19. We carried forward information on COVID-19 status, so that the derived indicator represented the history of ever reported testing positive or suspecting COVID-19 infection.

Statistical Methods

We used a timeline to visualize the timing of data collection relative to the evolving COVID-19 pandemic, which we characterized using estimates of the number of cases in India from models derived by the Institute for Health Metrics and Evaluation (40). We described the data using medians and interquartile ranges for continuous data given that some measures of interest were skewed. We used proportions for binary variables and provided statistics both overall and by tertiles of prepandemic cognitive functioning. For the purposes of initial data description, when individuals were seen at more than one visit during the COVID period, we summarized health outcomes during this time period by taking the mean across all nonmissing observations. In subsequent visualizations and analyses, each visit during the RTI-COVID survey was treated as a distinct observation with appropriate corrections for standard errors to increase power given the limitations of the available sample size. To aid in the visualization of the data on the mental health outcomes considered, we categorized the mental health measures considered based on the distribution of the data, as the data were not normally distributed, and a large proportion of responses were 0 (no mental health symptoms reported). Conclusions were consistent when substituting wave-level data for mean mental health outcomes over the entire COVID period for each participant (Supplementary Figure 2).

We used linear regression models to assess the association between prepandemic cognitive functioning and mental health outcomes. We used generalized estimating equations with an exchangeable correlation structure to account for the inclusion of multiple observations per participant. We estimated 3 models for each mental health outcome considered: the first was unadjusted, the second adjusted for age, gender, rural/urban residence, educational attainment, marital status, caste, and state, and the third additionally adjusted for prepandemic mental health symptoms. We then repeated the same set of models using prepandemic DSM-5 neurocognitive disorder or CDR MCI/dementia as the exposure. Although the PHQ-9 and BAI were only available for a small number of specific visits in the RTI-COVID survey, precluding the assessment of effect heterogeneity across pandemic periods, the PHQ-4 was assessed at every wave of the RTI-COVID survey. Therefore, we evaluated effect heterogeneity across pandemic period using data on the PHQ-4 from wave 1 (lockdown), wave 7 (delta wave), and wave 9 (omicron wave). We assessed whether effect estimates were significantly different for different pandemic periods using an interaction term between prepandemic cognition and wave. We also assessed interaction between COVID-19 infection and prepandemic cognition on observed associations with mental health outcomes in additional linear regression models. Initial descriptive statistics in Table 1 and all regression analyses accounted for survey weights to correct for sampling processes and selection bias. All analyses were done using R version 4.2.2.

Results

The 1 182 participants included in the study were observed on average 5.71 times over the course of the pandemic, resulting in a total of 6 749 observations. Participants had a median age of 69.0 (interquartile range [IQR] 66.0–74.0) years, and there were fewer women than men (43.3%; Table 1). The majority of participants were from rural areas (66.7%) and had no education (79.6%). Those with higher levels of cognitive functioning before the pandemic were younger, more likely to be married, and more highly educated, but were less likely to be women or to live in rural settings. There were also differences in prepandemic cognitive functioning by caste; those in less advantaged caste groups were more likely to have lower levels of prepandemic cognitive functioning.

Median PHQ-9 score was higher among the lowest tertile of prepandemic cognitive functioning (2.00 IQR: 0–4 vs 1.00 [0–3] for Tertile 2 and 1.50 [0–3] for Tertile 3); a similar pattern was apparent for PHQ-4, but not BAI (Table 1). However, patterns by cognitive tertile were apparent when examining the full distributions of mental health outcomes (Figure 2). Those with the highest levels of prepandemic cognitive functioning were the most likely to report 0 symptoms across all 3 mental health scales, and there was a dose–response pattern for PHQ-9 and BAI. The pattern was inverted for the endorsement of the highest levels of symptoms across all 3 scales—those with the lowest levels of prepandemic cognitive functioning were most likely to be in the highest category of mental health symptom burden when examining crude patterns.

Weighted regression models indicated that those with higher prepandemic cognitive functioning had lower mental health symptom burden (Figure 3). Effect estimates were similar in unadjusted and adjusted models, although estimates were less precise in models adjusting for potential confounders. After confounder adjustment, associations between prepandemic cognition and both PHQ-4 score and BAI were not statistically significant, but the association between prepandemic cognition and PHQ-9 score was marginally significant (p = .06). The estimated coefficient for PHQ-9 score, adjusting for potential confounders including prepandemic mental health, indicated that every SD unit increase in prepandemic cognitive functioning was associated with a difference in mean PHQ-9 score of −0.38 (95% confidence interval [CI] −0.77 to 0.01) units.

The associations between prepandemic DSM-5 neurocognitive disorder and both PHQ-4 score and BAI score were null and coefficient estimates were small in magnitude (Figure 3). However, the association between DSM-5 neurocognitive disorder and PHQ-9 score was strong; individuals with DSM-5 neurocognitive disorder had PHQ-9 scores that were 0.62 (95% CI 0.11–1.13) units higher than individuals without DSM-5 neurocognitive disorder, adjusting for all potential confounders and prepandemic mental health symptoms. Although associations between prepandemic CDR MCI/dementia and both PHQ-4 and PHQ-9 scores were null, coefficient estimates were in the expected direction, and the magnitude of the coefficient was somewhat large (0.41). The association between CDR MCI/dementia and BAI score was statistically significant, indicating that those with CDR MCI/dementia before the pandemic had BAI scores that were 0.60 (95% CI 0.03–1.17) units higher than individuals without CDR MCI/dementia, after adjusting for potential confounders and prepandemic mental health symptoms.

Further examination of effect modification of the association between prepandemic cognition and PHQ-4 score by timing (lockdown vs delta wave vs omicron wave) during the COVID-19 pandemic provided evidence of meaningful effect modification (p value for interaction = .02; Figure 4). Although power was lower for stratified estimates, coefficient estimates were positive for the lockdown period, indicating that those with higher cognitive functioning had higher PHQ-4 scores, whereas for both the delta wave and omicron wave, coefficient estimates were negative, consistent with findings in the main analyses and across the other mental health outcomes considered. In main analyses for the association between prepandemic cognition and PHQ-4 score, after excluding data from waves 1 and 2 of the RTI-COVID survey, coefficients remained imprecise but point estimates were more than doubled in magnitude in models controlling for all confounders including prepandemic mental health symptoms (before exclusion: −0.10 95% CI −0.34 to 0.15; after excluding waves 1 and 2: −0.21 95% CI −0.47 to 0.04).

There was no significant effect modification by COVID-19 infection status based on data in waves 6–8; estimates were imprecise, likely due to measurement error and low reporting of confirmed or suspected COVID-19 infections (19% of the weighted sample at wave 8; Supplementary Figure 3). Although not significant, examination of point estimates indicated that the association between prepandemic cognition and PHQ-9 may have been larger among those who reported confirmed or suspected COVID-19 infection.

Discussion

In a nationwide longitudinal cohort study in India, the burden of mental health symptoms varied depending on prepandemic cognitive function and stage of the pandemic. Overall, results suggested that lower prepandemic cognitive functioning and dementia were associated with higher mental health symptom burden, particularly for the PHQ-9 scale. However, investigation of pandemic timing showed heterogeneity in mental health outcomes by cognitive status and pandemic period. Those with higher cognitive functioning exhibited greater mental health burden during the initial lockdown period, with lower burden during the delta and omicron waves.

Findings show an increased burden of mental health symptoms among those with lower prepandemic cognitive functioning. This aligns with some studies investigating the effect of the pandemic on neuropsychiatric symptoms of people with dementia (22,24,41). However, the literature is mixed, with other studies from the UK indicating little significant change in mental health status among different cognitive groups (42,43). These conflicting findings suggest potential complexities surrounding the interactions between study context and the underlying associations between mental health trends and cognitive functioning; differences may be due to heterogeneity in the populations under study or the evaluation timeframe. Heterogeneity highlights the need to consider cultural and socioeconomic characteristics while investigating the mental health burden of different populations.

The prior literature has largely focused on the effects of lockdown measures on the psychosocial and neuropsychiatric symptoms of individuals with MCI or dementia (10,44,45), but little evidence exists on the long-term implications of these acute events on cognitively vulnerable populations. With data collected at multiple time points, we were able to investigate the mental health burden during various stages of the pandemic. We found suggestive evidence of heterogeneity across time, indicating that older adults with lower cognitive functioning exhibited lower mental health burden (using PHQ-4) during the lockdown period, with higher burden during the delta and omicron waves. One explanation for such heterogeneity across time could be that the more structured response and lockdown policies were more helpful, or led to less harm, for those with cognitive impairment or dementia. In comparison, the fear of illness and the less successful policy measures employed during the delta and omicron waves may have had a more negative impact on those with cognitive impairment. As time progresses, we may see a worsening of physical and mental health in the older adult population due to decreases in physical activity, prolonged social isolation, and uncertainty in the outcome of the pandemic (41). We did not observe effect modification by reported or suspected COVID-19 infection, which may indicate that the pandemic environment had a larger impact on mental health than individual experiences with COVID-19 infection. However, underreporting and measurement error in self-reported infections likely affected results and may have also led to reduced power for these analyses. Future follow-up and better ascertainment of COVID-19 status are needed to improve estimates of associations and characterize lingering effects of policies and pandemic experiences on the mental health of cognitively impaired older adults.

There are several plausible mechanisms that may help explain why those with low cognitive functioning or dementia may be at increased risk of poor mental health outcomes due to the COVID-19 pandemic. Studies have shown that the COVID-19 pandemic led to functional decline in daily activities, interruptions in functional rehabilitation, and disruptions in medical care services among people experiencing cognitive decline and dementia (10,21,22,46). These factors may contribute to psychological distress leading to poor mental health outcomes. Repercussions of the pandemic, such as decreased physical activity and isolation, should therefore be considered when studying the long-term implications of the COVID-19 pandemic on mental health. Such pandemic-related repercussions are not unique to the COVID-19 pandemic but are likely to accompany any future emerging pandemic or emergency environment. Advanced planning for such events will be critical to prevent the reoccurrence of the anticipated effects of such environments on mental health among older adults with cognitive impairment.

Prior research has found differences in the prevalence of anxiety and dementia across countries during the COVID-19 pandemic (47). Locations hit hardest by the pandemic, in terms of infection rates and reductions in human mobility, had the greatest increases in prevalence of major depressive disorder and anxiety disorders (48). Consideration of this patterning is imperative for informing policies aimed at reducing mental health burden in low- and middle-income countries (LMIC), such as India, which suffered some of the highest rates of infections and mortality globally (1). Older populations and those with cognitive impairment and dementia in LMICs are particularly vulnerable, given their risk of mortality and severe illness (49,50). Moreover, older adults are more likely to suffer psychological impacts due to social isolation, physical health problems, and reduced access to care (51,52). Our results, combined with this prior literature, underscore the need for targeted approaches to the development of policies and responses to future pandemics or other public health emergencies.

This is the first study to assess whether cognitive status alters the impact of the pandemic on mental health symptoms among older adults in a nationwide study in India. We were able to evaluate this association not only during the initial lockdown, but throughout later phases of the pandemic as well. Though our findings make a meaningful contribution to the existing literature, limitations should be noted. Potential selection bias may arise from telephone-based survey administration due to challenges with participation for individuals with the highest levels of cognitive impairment. Therefore, our results may not fully capture the experiences of those who may be most vulnerable to the effects of isolation and decreased access to care, among other repercussions of the pandemic. Results presented here may be conservative in light of the exclusion of those with the lowest levels of cognitive functioning. We also recognize there is difficulty in teasing apart differences due to the mental health measure considered (PHQ-4 vs PHQ-9 vs BAI) from differences in timing due to the changes in survey content over time. Although the PHQ-4 was collected throughout every round of the phone survey, the PHQ-9 and BAI scales were only collected at certain time points to minimize burden on the respondents. The PHQ-9 was asked in wave 3 (September 2020–October 2020) and wave 6 (March 2021–May 2021). We do not have PHQ-9 information after the delta wave. Further, BAI was only collected in wave 5 (January 2021–February 2021), wave 7 (July 2021–September 2021 [immediately after the delta wave]), and wave 9 (March 2022–May 2022 [immediately after the omicron wave]). Given this timeline, it is difficult to differentiate whether observed differences are due to differences in the timing of measurement, or the scale administered. In light of this challenge, we refer to all outcomes as indicators of poor mental health throughout the article and do not overinterpret differences in the content of the scales. Although we differentiated PHQ-4 data collection during the lockdown phase, delta wave, and omicron wave to explore effect heterogeneity, data collection efforts corresponding to the omicron and delta waves largely occurred after the conclusion of the infection peak due to challenges with data collection. Therefore, our assessment of mental health outcomes during these time periods may underestimate the true burden.

Overall, this study investigated how cognitive status alters the impact of COVID-19 on mental health outcomes. Lower prepandemic cognitive functioning was associated with higher mental health symptom burden. However, estimated associations differed depending on the pandemic period, with the greatest burden on those with cognitive impairment observed during later stages of the pandemic. Results highlight the potential for long-term negative consequences related to the COVID-19 pandemic among older adults with cognitive decline or dementia. These findings underscore the need for more research investigating the mental health implications and long-term effects of emerging pandemics and other public health emergencies. Increased attention to the needs of vulnerable populations is needed to inform policies aimed at enhancing resilience and preserving the mental health of these groups.

Funding

This work was supported by the National Institute on Aging, National Institutes of Health (U01AG065958 and 3R01AG030153).

Conflict of Interest

None.

Acknowledgments

The authors acknowledge the participants and families who participated in the RTI-COVID India study, the staff at the study sites, as well as the personnel involved in the data collection and data release.

Author Contributions

E.N.: Contributed to study conceptualization, formal analysis, methodology, visualization, and writing (original draft, review, and editing). S.P.: Contributed to conceptualization, investigation, project administration, and writing (original draft, review, and editing). J.L.: Contributed to conceptualization, funding acquisition, investigation, project administration, supervision, and writing (review and editing).

References

Author notes