In this study, pandemic refers to COVID-19, which is defined as an acute infection, that is, the diagnosis of COVID-19 infection. Based on the COVID-19 study, an add-on study to the Dalina Health Management Cohort (DHMC) (ChiCTR2300073363) observed the health status after infection of the health examination population in the northeast of China. DHMC is an open cohort study established by personal physical examination of the Second Hospital of Dalian Medical University in China, which is a longitudinal survey launched in 2014. DHMC was designed to collect the results of personal physical examination included laboratory test results (biochemical examination), auxiliary examinations (ultrasonic testing, etc.), health questionnaire (demographic variables, lifestyles, disease history, medication history, nutritional status, psychological situation), and evaluate population health status, and then conduct health management.

We included all individuals (aged ≥18 years) who underwent two health assessments. The first assessment was conducted at some point during the year 2022. The second assessment took place between 1 January and 30 April 2023. Therefore, the approximate interval between the first and second assessments for any given participant ranged from approximately 9 to 16 months, with a typical interval of approximately 1 year. We only included individuals who completed both assessments. We excluded 1,321 individuals with incomplete information and ultimately included 2,471 participants (43.10% female) in a self-control study on the changes in physical health before and after COVID-19 infection. Analyses were performed using the entire data set (N = 2,471) and one subset (N1 = 2,264, who were infected after 1 December 2022, and had undergone physical examination). The flowchart is shown in Supplementary Figure S1. This study was conducted in accordance with the principles of the Declaration of Helsinki, and the health management cohort study and its add-on study were approved by the Medical Ethical Committee of the Second Hospital of Dalian Medical University. This study has informed consent obtained from all participants.

·

This study updated the components of frailty by combining Fried, Rockwood, and Mitnitski (9–11). We used a combination of self-reported health and routine examination results to measure the frailty index (FI) in order to estimate frailty. FI is a comprehensive indicator that integrates information from biomarkers, medical history, living habits, etc. In our study, FI was measured for each participant separately using 54 items, 14 categories (Supplementary Table S1) in each medical examination year. Each item in each category is defined as a binary classification variable according to the diagnostic criteria (12). The score of the sum of the items in each category≥1 is defined as a new binary classification variable indicating having a health deficit; otherwise, there is no health defect. For each participant, FI is calculated as the proportion of health attributes that a person has in a deficit state, ranging from 0 (no health deficit) to 1 (health deficit). Finally, we divided the degree of frailty according to the tri-sectional quantiles of FI, namely mild frailty, moderate frailty, and severe frailty. We defined acute infection as the outcome based on the questionnaire “whether you have confirmed COVID-19 infection through antigen self-test or nucleic acid test recently” reported by the participants during physical examination.

All variables were defined as binary classification variables. Smoking and alcohol consumption were categorized as yes (having related experience) or no (not having related experience). Metabolic equivalent (MET) was calculated from the relevant physical activity types and time reported by participants, and MET ≤ 3 was defined as low physical activity, MET > 3 and ≤ 6 was defined as moderate physical activity, MET > 6 was defined as vigorous physical activity. (13). We defined individuals under the age of 60 who sleep for more than 7 h and those over 60 years who sleep for between 5 and 8 h as normal sleep (14). Having good tooth brushing habits means brushing teeth twice or more a day. The dietary nutrition standards are evaluated according to the provisions of the Chinese Dietary Guidelines 2022 (15), which include a comprehensive evaluation of 24 h of grains, meat, eggs, aquatic products, tubers, vegetables, fruits, nuts, soy products, white water, and milk. The evaluation criteria are the same as the rules for calculating FI, and ultimately, the population that meets dietary standards is considered to have a dietary score of 0. The drinking amount of tea is defined as 2–3cups per day as mild intake, 4–5 cups as moderate intake, 6–8 cups as excessive intake, and more than 8 cups as severe intake, based on the reference of less than one cup (250 mL) per day. The soda water is defined as no intake per week, 1–5 cups per week, 6–9 cups per week, and more than 10 cups per week as reference, mild intake, moderate intake, and excessive intake, respectively. Compared to not drinking coffee, we defined 1–2 cups per week as mild intake, 3–5cups per week as moderate intake, and more than 5 cups as excessive intake. Mild intake of sugar-sweetened beverages was defined as 1–2 bottles per week, 3–4 bottles of sugar-sweetened beverages as moderate intake, and more than 4 bottles per week as excessive intake. There were 11 items about psychological stress for the subjects, and those with a total score of 0 were defined as optimistic about psychological stress. We defined nodules of BI-RADS 3 and above in thyroid ultrasound and C-TI-RADS 3 and above in breast ultrasound as having nodules. Ultrasound diagnosis of any degree of fatty liver was defined as having fatty liver. Pulmonary nodules and lung inflammation were defined through the diagnosis of low-dose computed tomography (LDCT). Cervical vascular abnormalities are defined as the phenomenon of plaques, narrowing, and thickening of the intima media in cervical blood vessels. The body mass index (BMI) was calculated by the ratio of weight (kg) and the square of height (m²), and a BMI between 18.5 kg/m2 and 24 kg/m2 is considered normal. The definition of having a family history, personal history, and medication history was that the participant had any related historical event.

We preliminarily described the characteristics of the data, which included continuous normal variables summarized by means and SDs, and medians and IQRs for continuous non-normal variables. Differences in baseline characteristics between groups were compared using independent samples t-tests or Mann–Whitney U-tests for continuous variables (as appropriate based on distribution), and chi-square tests or Fisher’s exact tests for categorical variables.

We used the Kaplan–Meier method to estimate the uninfected rate of COVID-19 and used the log-rank test to compare the rates by the degree of frailty. The Cox proportional hazards model (Cox PH) was tested using the cox.zph function in the R Survival package. We adjusted three models of the Cox PH to assess the association between frailty status and infection. Analyses were conducted stratified by frailty category. Within each stratum, regression was used to assess the association of health behaviors with the likelihood of frailty, using the favorable level of each behavior as the reference. Frailty status was considered to have a divergent association with outcomes if they were significantly associated with COVID-19 infection, but the log2 hazard ratios had opposite signs. For the variables in the model that failed the assumption for a Cox PH model (p < 0.05), a time-dependent COX model was considered (as detailed in Supplementary Figure S2). We use the tt parameter and the log conversion to construct the interaction term of the variables that time transform in order to solve the problem of coefficients changing over time. The model is specified as: tt = function ( x , t , … ) x ∗ log ( t + 20 ) where x is the covariate with time-varying effects, and t is the analysis time.

Subsequently, we still established three traditional Cox regression models to analyze the association between frailty and health behaviors. The first model was a crude model with unadjusted. The minimal adjusted model was adjusted sex and age. The last model was adjusted for vaccination based on the minimal model.

We performed an interrupted time series analysis (ITSA) to estimate the impact of COVID-19 infection on the change in frailty. ITSA is a common method in public health, which tests the effectiveness of an intervention by controlling the upward or downward trend of outcome variables before intervention, including changes in levels and trends before and after interruption. First, we used least squares estimation for the Breusch–Godfrey test of independent variables and hysteresis residuals. Second, we used segmented regression to establish a model to fit the impact of a sudden COVID-19 incident on frailty. We compare outcomes before and after a policy intervention using two time points. The model is specified as: Y t = β 0 + β 1 time + β 2 int ervention + β 3 post

Y t is the value of FI in time t; time is a continuous variable indicating the time span, where the first observation (2022.1.1) was defined 0, and the last observation (2023.4.31) was defined as 2,471; intervention is a binary variable indicating whether the participant was infected; post is a continuous number beginning with 0 after the infection on 1 January 2023. β 0 estimates the FI at the beginning of the time series. β 1 estimates the preinfection trend of FI. β 2 estimates the immediate change in the level of FI after infection. β 3 estimates the trend of FI after infection.

A sensitivity analysis was conducted using a segment time-dependent Cox model. We use a hierarchical function for time and divide it into two periods based on the proportional risk assumption graph: 0–14 days and after 14 days. Then we fit and explore the impact of frailty on the occurrence of COVID-19 infection based on the inflection point.

All statistics and data analysis were performed in Stata SE15 and R 4.3.0. Estimated effects with 95% CIs were calculated, and a p-value of <0.05 was considered statistically significant.

The baseline characteristics are presented in Table 1. A total of 793 female participants and 743 male participants were acute positive. Compared to an uninfected female, an infected female tends to be younger, more alcohol consumption, have a worse psychological status, have a family medical history, have higher FI, and be of moderate or severe frailty status (p < 0.05). Compared to an uninfected male, an infected male tends to be younger, non-smoking, non-alcohol consumption, poor sleeping, and have a poor psychological status, having a personal and family medical history, higher FI, and being of moderate or severe frailty status (p < 0.05).

A significant increasing trend in the frailty index (FI) was observed in the overall population following acute infection, with a mean change of +0.029 (95% CI: 0.016, 0.041; p < 0.001). At baseline, males exhibited a higher FI (0.369) than females (0.320). However, following infection, the immediate increase in FI was more pronounced in females (+0.053; 95% CI: 0.035, 0.070; p < 0.001) than in males (+0.032; 95% CI: 0.015, 0.049; p < 0.001). The most marked instantaneous change in FI was observed among individuals aged 45 to 65 years. The details about the change in the frailty index of the population before and after the pandemic are shown in Figure 1. Subgroup analysis based on infection status revealed that individuals who experienced an acute infection had a slower immediate change ( β 2 ) in FI (0.028; 95% CI: 0.013, 0.043; p < 0.05) compared to those who were not infected. This difference was particularly evident among adults aged over 65 years. A detailed comparison of FI changes between infected and uninfected individuals is provided in Supplementary Table S2. Furthermore, analysis of the compositional changes in the frailty index indicated that 20 biomarkers exhibited significant immediate alterations after infection in the overall population. Additional details are available in the Supplementary Figures S3–S8; Supplementary Tables S3–S6.

Table 2 shows that moderate frailty participants and severe frailty participants had a significantly higher hazard of infection than the mild frailty participants in the adjusted model; the HR was 1.19 (95% CI: 1.04–1.35) and 1.34(95%CI: 1.15–1.56). The results did not change after using the segment time-dependent Cox model for sensitivity analysis.

The improvement of different health behaviors varied in different frailty statuses after adjusting for age, gender, and vaccination (Figure 2). In mild frailty, the HR of 1.49 (95% CI = 1.11, 2.01) and 1.47 (1.14, 1.89) for participants with poor sleep quality and higher psychological scores. Among moderate frailty, grains and vegetables can reduce the risk of disease by approximately 70% when they meet the standard. The higher psychological score increases the risk of infection by 38% (HR = 1.38, 95% CI = 1.02, 1.86). People in severe frailty engaging in excessive physical activities actually increase infection risk by 38% (HR = 1.38, 95%CI: 1.02, 1.86). It is worth noting that regardless of the degree of frailty, smoking can reduce the risk of infection by more than 30%. Meanwhile, sugary drinks increase the risk of infection by 20–90%.