Study design
Corona Immunitas Ticino (CIT) is a population-based prospective seroprevalence study that started in July 2020, after the first COVID-19 epidemic wave and lockdown in Southern Switzerland. The main aim of CIT was to assess the spread and impact of the COVID-19 pandemic. CIT is part of Corona Immunitas (CI), a nationwide research program led by the Swiss School of Public Health (SSPH+) [14]. Detailed information about the study design is reported elsewhere [14, 15].
Study setting and participants
CIT was conducted in Southern Switzerland—Canton Ticino, which borders Lombardy, Italy, the epicentre of the epidemic in Europe. We recruited participants using age- and sex- stratified random sampling based on regional registries of the Federal Office of Statistics. For this study on frailty, we considered 874 older adults aged 65 years and older, who agreed to participate in September and November 2020, provided informed consent and responded to the baseline questionnaire. Of these we excluded 114 participants due to missing values for the computation of the FI (described below). This reduced the analytical sample size (n1) for the FI construction to 660. The included and excluded participants had similar age distributions. For the further assessment of the relationship between frailty and seroprevalence, we analysed data of the subsample of older adults who took part in the serosurvey and who had a valid serological test (n2 = 481).
Data collection
The Research Electronic Data Capture (REDCap) [16, 17], a secure, web-based platform hosted at the Università della Svizzera italiana (USI) was used for questionnaire development, data collection, storage, and management. Older participants were also offered the possibility to participate over the phone with a dedicated interviewer using computer assisted telephone interviewing (CATI).
We asked participants to complete a baseline questionnaire following registration with the study. One week after the completion of the baseline assessment, we administered monthly and weekly questionnaires for repeated measures. We enquired about socio-demographic characteristics, physical and psychological health status, social relationships, and lifestyles of participants and their household environments.
To measure seropositivity to SARS-CoV-2, we invited participants to two rounds of blood testing. Professional nurses collected peripheral venous blood samples at a chosen healthcare facility or in their homes. The samples were analysed with the Luminex binding assay SenASTrIS (Sensitive Anti-SARS-CoV-2 Spike Trimer Immunoglobulin Serological) to detect SARS-CoV-2 antibodies. Previous validations in population-based samples showed high sensitivity and specificity [18].
For the current paper, we used data from the baseline questionnaire, the third monthly questionnaire and the laboratory results from the serological test performed between November 2020 and January 2021 (before the introduction of vaccines).
Frailty Index Construction
We derived the Rockwood FI using the method described by Searle et al. [4]. We included 30 variables that covered domains including chronic diseases, activities of daily living, lifestyle, physical measurements, self-reported health status, and psychological symptoms (Supplementary Table 1). The rationale of the FI is that the deficits/impairments should be related to ill-health status, should cover a variety of health domains, should progressively increase with age, and should not saturate at relatively younger age. The 6 deficits related to psychological symptoms and signs were selected from the third monthly questionnaire and the remaining 24 deficits were extracted from the baseline questionnaire. The FI is calculated as the ratio of the sum of deficits reported to the total number of deficits considered, and we applied a previously validated cut-off (i.e., 0.21) for frailty caseness [19]. We considered 0.1 ≤ FI < 0.21 as pre-frail, FI < 0.1 as robust [19]. For example, if a participant reported 6 out of 30 deficits considered, his/her FI would be 0.2, and would be classified as pre-frail. Before computing the FI, we recoded binary variables according to their possible answers as either ‘0’ or ‘1’, with ‘1’ representing the presence of a health deficit. For ordinal variables we used a Likert-like scale. We assigned values of ‘0’, ‘0.5’, ‘1’ to variables with 3 possible answers; ‘0’, ‘0.33’, ‘0.67’, ‘1’ to variables with 4 possible answers; and ‘0’, ‘0.25’, ‘0.5’, ‘0.75’, ‘1’ to variables with 5 possible answers. For the continuous variables, such as BMI (Body Mass Index), we used universally agreed cut-offs [20] to categorize the responses (see Supplementary Table 1), then we assigned values accordingly using the abovementioned method.
Covariates
All the covariates were collected in the baseline questionnaire, and included: age (modelled as continuous and categorical variable for analytic purposes, namely 65–69 years old, 70–74 years old, 75–79 years old and 80 years old and plus); sex (male/female), income satisfaction (i.e. ‘not enough’, ‘enough’ and ‘more than enough’), education (‘none’, ‘compulsory education’, ‘higher secondary education’, and ‘university’), smoking (‘non-smoker’, ‘past smoker’ and ‘current smoker’), and multimorbidity (yes/no) for participants who reported two or more chronic conditions.
Statistical analysis
We used complete case analysis in our study because those participants with missing values on the items used to construct the FI were not significantly different from those with no missing values across socio-demographic characteristics. We first describe the socio-demographic characteristics of the sample. Means and standard deviation (SD) were used for normally distributed data, median and interquartile range (IQR) for skewed data, and Spearman correlations and Wilcoxon rank sum tests to compare groups.
We used Poisson regression to estimate the prevalence of frailty by sex and age group. We used log-transformed linear regressions to explore the association between age and FI. In addition, we used univariate and multivariable logistic regressions to assess associations between age group, sex, income satisfaction, education level and smoking status and frailty (yes/no).
Potential risk factors for SARS-CoV-2 seropositivity including age group, sex, smoking status, frailty status, multimorbidity and self-reported health status, were tested separately using logistic regressions. A p value less than 0.05 was considered statistically significant. We performed all analyses using Stata Version 17.0 [21].