Data used in this study were extracted from 2 waves of the Shandong Rural Elderly Health Cohort (SREHC), an ongoing longitudinal study of community-dwelling older people in rural Shandong, China. The baseline survey was completed in June 2019, which was considered prior to the COVID-19 pandemic on the basis of the declaration from the World Health Organization on March 11, 2020, that COVID-19 can be characterized as a pandemic [32]. The detailed study design, sampling, and data collection have been described elsewhere [2,33,34]. Briefly, we used the multistage stratified random sampling method to select participants, excluding those who had a clinical diagnosis of dementia or psychiatric diseases, could not complete the interview due to severe physical conditions, or were unwilling to cooperate with the interviewers. There were 3243 respondents aged 60 years and above who participated in the baseline survey. Next, a follow-up survey was conducted during the COVID-19 pandemic from August to September 2020. The final analytic sample included 2785 (85.88%) respondents who participated in both baseline and follow-up surveys. We restricted the analysis to these 2 data points for the following reasons: First, we were specifically interested in examining the short-term consequences of combined frailty and multimorbidity on psychological distress so as to identify the most vulnerable group with the greatest risk of worse psychological outcomes and inform early interventions during the pandemic. Second, previous research has shown that social support would have an immediate effect on a scaled measure of psychological health [27]; therefore, we assessed the moderating role of CSS between combined frailty and multimorbidity and psychological distress with just a 1-year lag based on the assumption that there may be a limited time lag for CSS to mitigate one’s psychological distress.

The study purpose, significance, methods, and risks were explained to all participants, and written informed consent was obtained from each participant before the 2 surveys. Ethical approval for the study was obtained from the Ethics Committee of Shandong University (approval no. 20181228).

We used the 10-item Kessler Psychological Distress Scale (K10) [35] to assess the psychological distress in this study. K10 is a commonly global measure for screening mental health, and its reliability and validity have been confirmed among older people in China [36,37]. K10 contains 10 items using a 5-point Likert scale from “none of the time” to “all of the time” to evaluate psychological distress, including depression, anxiety, nervousness, hopelessness, restlessness, and worthlessness, in the past 4 weeks. K10 in this study showed good internal consistency reliability, with Cronbach α=.91 for both baseline and follow-up surveys. The raw score of K10 ranges from 10 to 50 points, with higher scores indicating a higher level of psychological distress. We transformed the K10 score into z-scores based on the mean and SD.

We used the 10-item Social Support Rating Scale (SSRS) [38] to assess the level of an individual’s social support, which is the most prevalent tool for measuring social support in China [39]. The SSRS contains objective support (3 items about visible, practical, and direct support), subjective support (4 items about the perceived level of support from family members, neighbors, and friends), and social support (3 items about the level of social support used). The SSRS has been shown to have good reliability and validity [40] in China, with Cronbach α>.7 for both baseline and follow-up surveys in our study. The raw SSRS score ranges from 12 to 66 points, with higher scores indicating higher levels of social support. In this study, we transformed raw SSRS scores into z-scores. A CSS score was computed by aggregating the standardized measures (z-scores) of the individual-level social support within each community, which reflected the average levels of social support for each community. This method has been proven to be valid and widely applied in previous studies related to social support [30,41,42].

In this study, frailty was assessed using the criteria of the frailty phenotype, which was proposed and validated by Fried et al [43]. It consists of 5 components: shrinking (unintentional weight loss), slowness (a walking time of 4.6 m adjusted by gender and height; individuals who met the criteria in the walking test or were unable to perform the test due to physical limitations were considered positive for slowness) [44], self-reported exhaustion, weakness (grip strength), and low physical activity. Older people who met 3-5 criteria were considered frail, while those who met 0-2 criteria were considered nonfrail.

Multimorbidity was defined as the coexistence of 2 or more chronic health conditions based on the Chinese Centers for Disease Control and Prevention (CDC) recommendations [45] and previous studies [46,47], including hypertension, diabetes, chronic lung disease, heart disease, asthma, liver disease, stroke, dyslipidemia, cancer, digestive disease, kidney disease, and arthritis. To validate the accuracy of this information, trained interviewers asked for help from village doctors to confirm the self-reported chronic condition information in the chronic disease case management system. We categorized multimorbidity into 3 groups: no chronic condition, 1 chronic condition, and multimorbidity.

The control variables included sex (male, female), age, marital status (divorced/widowed, married), educational attainment (illiteracy, primary school, junior school, high school or above), economic status (household income per capita; quartile 1 was the poorest and quartile 4 the richest), sedentary behavior (hours/day), smoking status (never/past, current smoking), drinking status (never/past, current drinking), and individual-level social support. We selected these variables as potential confounders based on existing studies [48-50].

First, we compared the baseline characteristics between respondents and nonrespondents using cross-tabulation with t tests for continuous variables and chi-square tests for categorical variables (see Table 1). To examine the combined effects of frailty and multimorbidity, we created a categorical indicator with the following 6 groups: (1) nonfrail without a chronic condition, (2) nonfrail with 1 chronic condition, (3) nonfrail with multimorbidity, (4) frail without a chronic condition, (5) frail with 1 chronic condition, and (6) frail with multimorbidity. Multilevel linear mixed effects models were used to quantify the strength of the longitudinal association between frailty and multimorbidity combinations and psychological distress using 2 waves of data for each participant (see model 1 in Table 2), and then, cross-level interactions between CSS and combined frailty and multimorbidity (see model 2 in Table 2) were included. The selected 3-level linear mixed effects modeling strategy was appropriate for repeated measures to account for the hierarchical structure of the data set that included survey time nested within individuals and individuals nested within communities. A random intercept was included in each model. Both model 1 and model 2 were adjusted for sex, age, education, economic status, marital status, sedentary time, smoking status, drinking status, individual-level social support, and survey time (to capture any individual-level idiosyncratic disturbances over time).

Next, to minimize the possible reverse causality, 2-level mixed effects models with a lagged dependent variable (LDV) were conducted to account for the cluster of participants within communities, and we also controlled the levels of psychological distress at baseline (see Table 3). Model 1 in Table 3 included follow-up psychological distress as the dependent variable and the baseline categorical indicator and the change in CSS between the 2 waves as the independent variable to examine whether baseline frailty and multimorbidity combinations predicted psychological distress during the COVID-19 pandemic. Model 2 in Table 3 included model 3 variables plus a 2-way cross-level interaction term between baseline frailty and multimorbidity combinations and the increased CSS. Both model 3 and model 4 were adjusted for baseline confounding variables, including sex, education, economic status, marital status, sedentary time, smoking status, drinking status, individual-level social support, and K10 scores, to test whether the increased CSS from baseline to follow-up mitigated the adverse impact of combined frailty and multimorbidity on psychological distress.

We ran additional analyses under the same multilevel framework to see how confounders adjusted the models. From model 0 (the unadjusted model) through model 2 (fully adjusted model), the variance partition coefficient for the community level decreased from 3% to 2% and from 6% to 4% for longitudinal association and predictive models, respectively, which implies that our confounders explained about one-third of the community-level variance in worsened psychological distress. Next, we added the separate variable of frailty status and multimorbidity in a full model to show the independent effects of frailty and multimorbidity as a comparison. All analyses were performed using Stata 14.2 (Stata Corp LP).

The baseline survey included 3243 older people from 58 communities. Of the 3243 respondents at baseline, 458 (14.12%) individuals were lost to follow-up because of the following reasons: outmigration with children (n=121, 26.4%), dropping out (n=295, 64.4%), and deaths (n=42, 9.2%). This yielded an analytic sample of 2785 (follow-up rate=85.88%) who participated in the 2 surveys. All characteristics (sex, age, education, economic status, marital status, sedentary time, smoking and drinking status, chronic condition, frailty, and social support) were similar between respondents (n=2785, 85.88%) and nonrespondents (n=458, 14.12%) at the follow-up survey (P>.05). Of the 2785 respondents, 1770 (63.55%) were female, the mean age was 69.19 (SD 6.16) years, 1164 (41.80%) had no education, and 2062 (74.04%) were married. The characteristics of the selected sample and dropouts are provided in Table 1. Psychological distress score values increased from a K10 score of 16.60 (SD 7.46) at baseline to 18.23 (SD 8.00) at the follow-up survey. The combination of frailty and multimorbidity was observed in 258 (9.26%) respondents, while 684 (24.56%) respondents were nonfrail without a chronic condition. Both individual-level social support and CSS decreased during follow-up (see Table 1).

Table 2 shows the results of multilevel mixed effects models with random intercepts at the individual and community levels. Model 1 showed that psychological distress significantly deteriorated during the COVID-19 pandemic compared with before the COVID-19 (β=.09, 95% CI 0.02-0.15, P=.008), while CSS (β=–.06, 95% CI –0.11 to –0.02, P=.008) was protectively associated with psychological distress. After adjustment for potential confounders, older adults who were nonfrail with 1 chronic condition (β=.13, 95% CI 0.07-0.19, P<.001), nonfrail with multimorbidity (β=.25, 95% CI 0.19-0.31, P<.001), frail without a chronic condition (β=.40, 95% CI 0.27-0.53, P<.001), and frail with 1 chronic condition (β=.48, 95% CI 0.38-0.57, P<.001) were at higher risk of psychological distress compared with older adults without any of the conditions, but the highest risk was observed in those with the combined presence of frailty and multimorbidity (β=.68, 95% CI 0.60-0.77, P<.001). As shown in model 2, the main effect of combined frailty and multimorbidity on psychological distress remained significant, and there was a significant cross-level interaction between CSS and frail individuals with multimorbidity (β=–.16, 95% CI –0.23 to –0.09, P<.001). To help the interpretations, the plots in Figure 1 show the estimated values of levels of psychological distress based on the estimates in model 2. Within the group of respondents who had combined frailty and multimorbidity, those with lower levels of CSS reported significantly higher psychological distress than respondents with higher levels of CSS, indicating the moderating role of CSS between frail individuals with multimorbidity and psychological distress.

Model 1 in Table 3 shows the effects of baseline frailty and multimorbidity on wave 2 psychological distress. Those with coexisting frailty and multimorbidity (β=.32, 95% CI 0.22-0.43, P<.001) showed the strongest association with greater psychological distress during the COVID-19 pandemic compared with nonfrail individuals without a chronic condition, while increased CSS (β=–.07, 95% CI –0.12 to –0.02, P=.003) from wave 1 to wave 2 was significantly associated with lower subsequent levels of psychological distress. In model 2, we examined whether increased CSS moderated the adverse impact of baseline coexisting frailty and multimorbidity on psychological distress during the COVID-19 pandemic. We found a significant interaction between increased CSS from wave 1 to wave 2 and coexisting frailty and multimorbidity (β=–.11, 95% CI –0.22 to –0.01, P=.035) for wave 2 psychological distress. As Figure 2 demonstrates, for individuals with coexisting frailty and multimorbidity at baseline, those with higher increased levels of CSS reported significantly lower psychological distress than respondents with lower increased levels of CSS, while such effect was less substantial for older adults who were nonfrail without a chronic condition.

Using data from a cohort of community-dwelling older adults in rural China, this study identified that coexisting frailty and multimorbidity are significantly associated with greater levels of psychological distress and that baseline coexisting frailty and multimorbidity predicted higher levels of psychological distress during the COVID-19 pandemic. Further, residing in communities with a higher atmosphere of social support moderates the relationship of combined frailty and multimorbidity and psychological distress. We also emphasized the role of increased CSS in mitigating the adverse effects of combined frailty and multimorbidity on psychological distress during the COVID-19 pandemic. These results highlight the community social environment, more specifically CSS, as an important buffer that may mitigate the adverse impacts of combined frailty and multimorbidity on psychological distress.

Multimorbidity and frailty are 2 major adverse health conditions among older adults [14], and the 2 conditions are also risk factors for mental health [6,7,51]. However, there is no previous research that has explicitly examined their combined effects on psychological distress in older adults. Only 1 similar study has shown that the combination of frailty and multimorbidity is associated with physical limitation and mortality [52], and our study extends the literature for psychological health. In our sample, the results of longitudinal association showed that the magnitude of the coefficient of psychological distress was strongest for coexisting frailty and multimorbidity compared to only 1 or none of the conditions. This is where this study makes the first novel contribution to previous work: we identified a vulnerable group of older adults at increased risk of psychological distress. It is plausible that frail older adults already face a higher risk of poor resolution of homeostasis [53], plus the worry that results from multimorbidity, making them more vulnerable when coping with stressful situations and increasing the risk of psychological distress. Results from this prospective study also add evidence to the literature by identifying baseline combined frailty and multimorbidity as a predictor of greater psychological distress in older adults during the COVID-19 pandemic, emphasizing the importance of identifying and preventing coexisting frailty and multimorbidity in reducing psychological distress in older adults when facing major public health issues. We should interpret our findings incorporating the context of the pandemic. Unlike normal daily circumstances, the COVID-19 pandemic would expose people to more complicated and wide-ranging stressors. In addition to the primary stressor of the fear of infection, the pandemic has brought a plethora of secondary stressors, such as social isolation and resource scarcity [51]. These sources of stress could lead rural residents to face the potential disruption of basic medical services, especially the possibly reduced access to non–COVID-19 health care, which therefore adversely affects their psychological health. For example, older persons with frailty and multimorbidity had greater needs for basic medical services; however, the lockdown and polymerase chain reaction (PCR) testing policies led to worries about access to health care services, which in turn led to greater psychological distress. One evidence is that while the rural health care system did not break down during the quarantine, most of the rural residents chose to delay health care seeking due to worry about being infected [54]. Our findings suggested that older adults with frailty and multimorbidity needed more professional psychological support during the pandemic; however, the availability of psychological therapy or support was limited, especially in rural areas. Moreover, the current mass quarantines and restrictions to public transport have inevitably become a major barrier to accessing psychological therapy or support for older adults [55], which may also cause their increased psychological distress.

Importantly, our study identified that residing in a community with higher social support mitigates the relationship between coexisting frailty and multimorbidity and psychological distress. This finding adds to the literature documenting the stress-buffering role of CSS in psychological health. The moderating role of CSS remains significant even after controlling for individual-level social support, suggesting that there are contextual effects that alleviate psychological distress regardless of individual-level social support. In other words, older adults with lower individual-level social support could also benefit from their communities with a higher social support atmosphere. One possible explanation is social contagion [56]. CSS is meaningful in shaping community cohesion and connectedness [30], which is important for individuals to obtain the resources they need. Compared with living in communities with low levels of social support, health-related information may be disseminated faster in communities with higher levels of social support [57], which may increase the likelihood of people taking up healthy standards of behavior. Older adults with coexisting frailty and multimorbidity are more likely to collect and notice health-related information and resources that are provided by social contagion (health-related information, behavioral norms, etc) and thereby benefit more from social contagion, while those without any of the conditions may not pay much attention to the information and thus may not be sensitive to the community atmosphere. We noticed that the interaction between nonfrail older individuals with multimorbidity and CSS was not significant, while that between frail older individuals with multimorbidity and CSS was significant. This finding may be due to the different severity of disease within the multimorbidity group across the frailty state (ie, frail people may have more advanced diseases than nonfrail people). Older people with multimorbidity who have not yet developed a frail state may be resilient to the negative effect of multiple comorbidities. Thus, frail individuals with multimorbidity are more sensitive and pay more attention to their own health than nonfrail individuals with multimorbidity. Concerning the dynamic role of CSS, we observed that increased CSS from baseline to follow-up mitigated the adverse impact of baseline coexisting frailty and multimorbidity on psychological distress during the COVID-19 pandemic. Practically, our findings imply that interventions targeted toward improving CSS might alleviate the effects of frailty and multimorbidity on the psychological well-being of older adults, especially when facing major public health issues.

We also observed an increase in psychological distress in the context of the COVID-19 pandemic. In addition to the fear of being infected with COVID-19, widespread secondary stressors associated with the pandemic may include decreased opportunities for interpersonal contact and excursions and potential economic loss [51]. Given the lack of control groups that were unaffected by the COVID-19 pandemic in our study, more complex statistical design is needed in future studies, and we will not discuss more at this point.

This study has several strengths and limitations. The novelties and strengths of this study include the study design and the measures we used. First, we used a large sample of community-dwelling older adults and followed up the psychological health outcomes with little missing data. Moreover, we used repeated measures of the key predictors and outcomes, including frailty, chronic conditions, social support, and psychological distress, and adjudicated K10 and CSS using standardized criteria. Furthermore, we not only used 2 waves of the data to show the longitudinal association between coexisting frailty and multimorbidity and psychological distress but also used baseline values to predicted psychological distress during the COVID-19 pandemic, which showed the robustness of the results.

Some limitations need to be considered. First, the observed relationships between combined frailty and multimorbidity and psychological distress cannot be constructed as causal relations, because this was an observational study. However, we took a set of measures to minimize the possible reverse causality, such as using lagged dependent variable models, controlling psychological distress at baseline, and adding a wave variable. Second, this study was conducted only in rural areas; whether the results are applicable to urban areas needs further study. In addition, our sample may not be representative of all rural older adults in China, as our participants were selected from 3 counties. As such, more nationally representative studies should be conducted to corroborate our findings. Third, as social support was assessed based on questionnaires, the results are subject to measurement bias: individuals with severe psychological distress might perceive themselves as having poor social support. However, we speculate that the aggregating social support at the community level could be influenced less. Fourth, some variables, such as chronic conditions and sedentary time, were self-reported, which may result in some recall bias. Finally, we cannot completely rule out residual confounding because of unmeasured variables; however, our results were robust after adjustment for multiple confounding variables.

In summary, the findings of this study indicate that community-dwelling rural older adults with both frailty and multimorbidity have higher levels of psychological distress compared to those with either condition alone or none of the conditions. Baseline coexisting frailty and multimorbidity predict the most subsequent psychological distress during the COVID-19 pandemic. Thus, frail rural older adults with multimorbidity constitute a high-risk group for psychological distress. Moreover, residing in communities with highly average levels of social support moderates the relationship between combined frailty and multimorbidity and psychological distress, and we confirmed the role of increased CSS in buffering the adverse effects of combined frailty and multimorbidity on psychological distress during the COVID-19 pandemic. Our findings suggest that more public health and clinical attention should be paid to frail older adults with multimorbidity in response to psychological distress. In addition, improving the average levels of social support within communities may be an effective approach to alleviating psychological distress in older adults who concurrently manifest frailty and multimorbidity.