A pandemic toll in frail older adults: Higher odds of incident and persistent common mental disorders in the ELSA-Brasil COVID-19 mental health cohort

Common mental disorders (CMD), including anxiety and depression, are major contributors to global disease burden and increased in some populations during COVID-19. Mental disorders are more common in people with physical illnesses, compounding treatment complexity and contributing to adverse medical outcomes. Frailty, a geriatric syndrome reflecting increased vulnerability to stressors, is commonly defined by two models: a physical frailty phenotype (weight loss, exhaustion, low activity, slowness, weakness) and a cumulative deficit Frailty Index. Although frailty and depression are correlated yet distinct constructs, and cross-sectional links with anxiety have been reported, less is known about frailty’s relationship with CMD beyond depression, particularly under pandemic stressors. This study aimed to test whether pre-pandemic frailty status was associated with persistent and incident CMD among older adults in São Paulo during the first COVID-19 wave.

Prior research has documented bidirectional associations between frailty and depression: frail individuals have higher odds of prevalent and incident depression, and depressed individuals have higher odds of prevalent and incident frailty. Studies suggest frailty and depression are distinct but highly correlated constructs in mid to late life. Cross-sectional work has also linked anxiety symptoms with pre-frailty and frailty in older populations. During COVID-19, most frailty research has focused on mortality and adverse clinical outcomes rather than mental health impacts. One pandemic-era cohort of community-dwelling older adults did not find associations between frailty and increases in depressive/anxiety symptoms, potentially due to different assessment tools. Overall, evidence on frailty’s relationship with CMD (beyond depression) remains limited, justifying the present investigation.

Design and participants: The ELSA-Brasil is a prospective cohort of civil servants from six Brazilian states. Baseline (2008–2010) included 15,105 participants aged 35–74, with follow-up waves in 2012–2014 and 2017–2019. In 2020, an online COVID-19 mental health assessment was conducted in three waves at the São Paulo center. The present analysis included 706 participants aged ≥60 years who completed the third onsite wave (2017–2019) and the first online COVID-19 wave (May–July 2020), with complete frailty and covariate data. CMD assessment: The validated Brazilian CIS-R assessed non-psychotic psychiatric morbidity in both waves. The onsite (2017–2019) CIS-R was interviewer-administered; the COVID-19 wave used a validated electronic self-report version. CMD was defined as CIS-R total score ≥12. Persistent CMD was defined as CMD present at both onsite and COVID-19 assessments. Incident CMD was defined as CMD absent onsite and present during COVID-19. Frailty assessment: Two definitions were used. (1) Frailty Index (FI): A cumulative deficit index constructed from 36 health deficits measured in the first and third onsite waves; FI calculated as sum of deficits divided by number of available variables (range 0–1). Frailty was defined as FI ≥0.25. (2) Physical frailty phenotype: Adapted from Fried criteria—unintentional weight loss, exhaustion, low energy expenditure, slow gait speed, weak grip strength. Frailty defined as ≥3 criteria; pre-frailty as 1–2 criteria. Physical performance tests included handgrip strength (Jamar dynamometer), 4-m gait speed, and 5-repetition chair stands. Covariates: Age at third onsite wave, sex, education (college or more vs high school or lower), and self-reported race (black, brown, white, other). Statistical analysis: Descriptive statistics compared participant characteristics by frailty status using appropriate parametric/non-parametric tests. Logistic regression models estimated associations of frailty (by FI and phenotype) with persistent and incident CMD. For persistent CMD analyses, individuals with incident CMD were excluded from the reference group; for incident CMD analyses, participants with onsite CMD were excluded from the reference group. Models adjusted for age, sex, education, and race. The continuous FI was also examined. Sensitivity analyses excluded exhaustion from both frailty constructs. Effect modification by age, sex, and education was tested via interaction terms (p<0.10 threshold for stratification). Two-tailed tests; alpha 0.05. Ethics approval obtained and informed consent provided.

- Sample: Mean age 69.2 ± 5.4 years; 54.7% female. Persistent CMD: 42 participants (6.0%); incident CMD: 48 participants (6.8%). - Frailty distribution: FI mean 0.16 ± 0.08 (range 0–0.57); frail by FI: 75 (10.6%). Phenotype: 302 robust (42.8%), 373 pre-frail (52.8%), 31 frail (4.4%). - Associations with persistent CMD (adjusted): - FI frailty vs robust: OR 8.61 (95% CI 4.08–18.18), p<0.001. - Phenotype frailty vs robust: OR 23.67 (95% CI 7.08–79.15), p<0.001. - Phenotype pre-frail vs robust: OR 3.54 (95% CI 1.50–8.38), p=0.004. - Associations with incident CMD (adjusted): - FI frailty vs robust: OR 2.79 (95% CI 1.15–6.78), p=0.023. - Phenotype frailty vs robust: OR 4.37 (95% CI 1.31–14.58), p=0.017. - Phenotype pre-frail vs robust: not significant. - Continuous FI: Each 0.01 increase associated with OR 1.12 (95% CI 1.08–1.17, p<0.001) for persistent CMD and OR 1.04 (95% CI 1.00–1.08, p=0.036) for incident CMD. - Sensitivity (excluding exhaustion): Associations of frailty with persistent CMD remained; associations with incident CMD were no longer significant; pre-frailty association with persistent CMD was no longer significant. - No effect modification by age, sex, or education.

Pre-pandemic frailty predicted both persistent and incident CMD during the first COVID-19 wave among older adults, regardless of whether frailty was defined by a cumulative deficit index or the physical phenotype. The findings reinforce the vulnerability of frail older adults to mental health burdens under severe societal stressors. The observed association of pre-frailty with persistent CMD suggests that even intermediate frailty states confer risk for ongoing psychiatric morbidity. The robustness of associations with persistent CMD after removing exhaustion from frailty measures indicates that results are not driven solely by overlapping symptom constructs. Results align with pre-pandemic literature linking frailty with depression and, to a lesser extent, anxiety, while offering novel evidence for broader CMD outcomes during a global crisis. Discrepancies with some pandemic-era cohorts may reflect differences in psychiatric assessment tools and timing. Clinically, the findings support incorporating frailty screening into mental health risk stratification for older adults, particularly during public health emergencies.

Frailty status assessed before the COVID-19 outbreak was associated with substantially higher odds of both persistent and incident CMD in older adults during the first pandemic wave in São Paulo. These results highlight frailty as an important marker to identify older individuals at elevated mental health risk, informing targeted allocation of resources and preventive strategies. Future research should examine temporal dynamics of CMD more granularly, evaluate specific psychiatric diagnoses in relation to frailty, investigate mechanisms linking frailty to CMD, and test interventions that integrate frailty assessment into mental health screening and care pathways for older adults.

- Psychiatric assessments were 1–3 years apart; CIS-R captures symptoms in the prior 30 days (and past week if present), potentially missing fluctuations between waves and misclassifying intermittent cases. - Different administration modes: onsite interviewer-administered vs online self-applied CIS-R during quarantine, though the latter is validated and shows comparable performance. - Low numbers of specific ICD-10 diagnoses precluded analyses by individual diagnoses. - Frailty Index included some variables from baseline (2008–2010); lack of up-to-date data may have led to misclassification, assuming some factors remained stable. - No data on timing of social isolation exposure or SARS-CoV-2 infection relative to CMD evaluations. - Participation rate <50% of the São Paulo sample for the online assessment, raising potential selection bias toward cognitively/functional fitter participants and underrepresentation of frailer individuals.