Study population

This study used data from the Swedish National Study on Aging and Care in Kungsholmen (SNAC-K) COVID19 Study (May-September 2020), and the data were linked to the sixth wave of follow-up (2016-2019) in SNAC-K [15]. SNAC-K is a longitudinal study that includes a random sample of older adults from 11 age cohorts (60, 66, 72, 78, 81, 84, 87, 90, 93, 96, and ≥99 years) who live in the Kungsholmen district of Stockholm. The eligible baseline sample included 4590 individuals [16]. Of these, 3363 participated (73.3 %) in the baseline examination in year 2001-2004 and have been followed up regularly every 6 years for those aged ≤72 years and every 3 years for those aged ≥78 years. Additional cohorts were added in 2007-2010 (81-year-olds, N=194), 2010-2013 (60-year-olds, N=677), and 2013-2016 (81-year-olds, N=195).

All participants eligible for the seventh SNAC-K wave (2019-2021) were invited to participate in a telephone interview between May and September 2020 (n=1442), when COVID-19 was declared a pandemic [17] (95% of the interviews were conducted in May and June). Individuals who had previous cognitive or physical difficulties, lived in nursing homes, or who had deceased since the sixth wave of follow-up were excluded from the SNAC-K COVID19 Study (n=103), leaving 1339 individuals. Out of these, 108 persons refused to participate or could not be contacted (response rate 91.9%).

Out of the 1231 individuals who participated in the SNAC-K COVID19 Study, 951 had also participated in wave 6 (2016-2019) (age cohorts 66, 81, 84, 87, 90, 93 and ≥96 years). Wave 6 will be referred to as the baseline examination of this study. For the purpose of this study, individuals having a Mini-Mental-State-Examination (MMSE) <24 [18] (n=18) and those who had a positive COVID-19 test (n=8) were excluded from the sample. We also excluded persons with missing data regarding the questions on COVID-19 testing (n=1), changes in PA levels during the pandemic (n=40), or in pre-pandemic factors, i.e., the Montgomery Åsberg Depression Scale (MADRS, n=17) [19], social network (n=87), physical functioning (n=82), lifestyle factors (n=29), or personality traits (n=45), leaving 624 individuals for the analyses.

Informed written consent was collected directly from each participant. The Regional Ethics Review Board in Stockholm has approved all phases of the SNAC-K study, as well as the SNAC-K COVID19 study (dnr: 2020-02497).

Data collection

During the baseline examination, information was collected through clinical examinations, nurse interviews, self-administered questionnaires, and cognitive tests administered by physicians, nurses, and psychologists. Information on physical health and behavioral conditions during the COVID-19 pandemic were obtained by trained nurses through telephone interview, following standard protocols.

Outcome

During the telephone interview, a nurse asked the participants whether they had changed their levels of light- or higher-intensity PA during the pandemic, as compared to their pre-pandemic levels (before March 2020). Light PA included walking, short bike rides, light gymnastics, and golfing. Higher-intensity PA included activities such as jogging, brisk long walks or bike rides, heavy gardening, intense gymnastics, long distance skating, skiing, or swimming. The response alternatives were “more”, “the same/no change”, or “less”. Since the focus of this study was to investigate PA reductions during the pandemic, and given that the proportion of those engaging in increased PA levels was the smallest group in the total sample (increases in light PA=21% and in intense PA=10%), PA was categorized as a reduction in PA (“less”) versus no reduction (“more or the same/no change”) in light or intense PA levels, respectively.

Exposures

The exposures were assessed during the baseline examination in 2016-2019.

Social network

Both quantity and quality of social network were measured. For social connections (quantity), questions on marital status, cohabitation status, parenthood, friendships, social network size, and frequency of direct contacts with children, parents, neighbors, and friends were included. Social support (quality) was assessed by the following items: reported satisfaction with the above-mentioned contacts, perceived material and psychological support, sense of affinity with association members, relatives, and living area, and being part of a group of friends. The overall scores for each variable were transformed into z-scores [20], and categorized according to the median as poor (below median) versus good (median or above).

Chronic somatic and mental diseases

Disease diagnoses were coded based on the International Classification of Diseases Tenth Revision (ICD-10) [21], and further operationalized according to a methodology described previously [22]. For the purpose of this study, 23 chronic cardiovascular, musculoskeletal, and neuropsychiatric diseases were included given their major contribution to functional impairment and disability [23], and their tendency to cluster together into recognized multimorbidity patterns [24]. The specific conditions considered were: 1) cardiovascular disease, i.e., ischemic heart disease, heart failure, atrial fibrillation, cerebrovascular diseases, cardiac valve diseases, bradycardias or conduction diseases, peripheral vascular disease, and other cardiovascular diseases; 2) musculoskeletal disease, i.e., dorsopathies, inflammatory arthropathies, osteoarthritis and other degenerative joint diseases, osteoporosis, and other musculoskeletal and joint diseases; and 3) neuropsychiatric disease, i.e., dementia, neurotic or stress-related and somatoform diseases, migraine and facial pain syndromes, peripheral neuropathy, Parkinson or parkinsonism, epilepsy, schizophrenia and delusional diseases, multiple sclerosis, other psychiatric or behavioral diseases, and other neurological diseases. The presence of cardiovascular, musculoskeletal, or neuropsychiatric diseases were categorized as having any versus having no disease.

Global cognition was measured using the Mini-Mental State Examination (MMSE) and was classified as MMSE scores of 24-27 versus 28-30. The Montgomery-Åsberg Depression Rating Scale (MADRS) was used to assess depressive symptoms [19]. The 10-item MADRS has a maximum score of 60 and the cut-off for depressive symptoms was >6 [25].

Physical functioning

The chair stand test was performed by asking participants to raise up from a chair five times as fast as possible, without using their arms, and the results were expressed in seconds. Walking speed at a normal pace was assessed over 6 m or 2.44 m, depending on the participants’ ability and the location of the test, and presented in meters per seconds (m/s). Balance was assessed using the one-leg balance test where participants were asked to stand twice on each leg up to 60 seconds, with their eyes open. Impaired mobility was defined as a walking speed <0.8 meter/seconds [26], impaired muscle strength as a chair stand time ≥17 seconds [27], and impaired balance as a single leg standing time <5 seconds [28]. Participants who were unable to perform a test due to severe physical limitations were categorized as having impaired function in that test [29].

Lifestyle factors

Smoking was categorized as being a smoker versus non-smoker. Alcohol consumption was obtained according to the frequency and amount consumed and was classified as none/occasional, light to moderate (1-14 drinks per week for men and 1-7 drinks per week for women), or heavy (>14 drinks per week for men or >7 drinks per week for women) [30]. Body Mass Index (BMI) was classified as underweight (<20 kg/m2), normal weight (≥20-<25 kg/m2), or overweight/obesity (≥25 kg/m2) [31]. For the analyses, alcohol consumption was dichotomized as no/occasional or heavy versus light to moderate, and BMI as underweight or overweight/obesity versus normal weight since these subcategories were similarly associated with reductions in PA. This dichotomization is further supported by the literature since studies have found J-shaped associations of alcohol consumption and BMI with adverse health outcomes in older age [32,33,34].

Personality traits

The personality traits of neuroticism, extraversion, and openness to experience were obtained using a 36 item from short version of the NEO Five-Factor Inventory (NEO-FFI). The scores were transferred into z-scores and categorized as follows: neuroticism as moderate or high versus low, and extraversion or openness to experience as low versus moderate or high [35].

CovariatesDemographic factors

Demographic factors included age, sex, and education. Education refers to the highest level of education achieved and was classified as elementary, high school, or university. For the analyses, education was dichotomized as having high school or less versus more than high school education since elementary and high school education were similarly associated with reductions in PA.

Physical activity at baseline

Light or intense PA in the 12 months prior to baseline was assessed by self-report. Light and intense PA were equally defined in the pre-pandemic baseline interview and the COVID-19 telephone interview. The answers for light and intense PA were combined and categorized as follows: insufficient levels of PA (less than weekly engagement in light and intense exercise) versus sufficient levels (engagement in light or/and intense exercise several days per week or more) [36].

Statistical analyses

Differences in the proportions of the baseline characteristics by age and sex were evaluated using Chi-Square tests for categorical data. For continuous data, Mann-Whitney U tests were used because of the skewed distributions. Multiple logistic regression models were used to examine the associations between pre-pandemic levels of physical, mental, social, and lifestyle-related factors with light and intense PA reduction during the pandemic. First, each potential risk factor was entered separately, adjusting for age, sex, and education. Second, analyses were performed to examine the multiplicative interaction between each potential risk factor with age and sex. Third, all risk factors associated with a reduction in PA with P-values <0.1 were entered in multiple regression models using a backward selection procedure. There was no multicollinearity between the exposure variables as the variance inflation factors (VIFs) were <4. For all statistical tests, P-values ≤0.05 were considered statistically significant, and all statistical tests were two-tailed. The statistical analyses were performed using SPSS 26.

Sensitivity analyses

To further explore the independent effect of the pre-pandemic risk factors on PA reduction during the pandemic, baseline PA were additionally adjusted for in the fully adjusted models.