Perceived physical activity during stay-at-home COVID-19 pandemic lockdown March–April 2020 in Polish adults

Study design

The study was approved by the Ethics Committee of the University School of Physical Education in Wrocław, Poland (no. 10.2020). All procedures were performed in accordance with the Declaration of Helsinki (World Medical Association, 2013). All participants volunteered to participate in the study (by ticking the online box). Although the study was anonymous, i.e., we did not collect any personal data, we asked participants to consent by clicking an appropriate checkbox at the beginning of the online survey. Given the nature of the study, we did not predetermine the sample size, i.e., we tried to recruit as many participants as possible.

Pandemic-related national health restrictions

The data was collected during the epidemic state between 31^st March and 20^th April 2020. The following restrictions affecting PA during the lockdown were in place:

• Free movement of people was restricted;

• People were allowed to leave home in precisely specified cases, such as for a trip to one’s work or business, for essential trips to family and relatives for care or caregiving, shopping trips to purchase food and essential supplies, medicines or pet supplies, visits to the doctor or to deal with urgent official matters, volunteering related to coronavirus help, dog walk (only one person at a time);

• All public events were cancelled;

• All schools, universities, theaters, cinemas, cultural centers, restaurants, clubs, hairdressers, tattoo, beauty salons, gyms, sport clubs, libraries, zoo, public gardens, etc. were closed;

• Obligatory 14-day quarantine for all returning to the country;

• Foreigners were not allowed to enter the country;

• All flights from and to Poland, as well as other transport means (buses, trains, tourist border crossings), were canceled;

• Non-essential shopping centers and shops were closed;

• Children under 18 were not allowed to walk outside without a legal guide’s assistance;

• Maximum five people during funerals and in churches were allowed;

• Access to parks, forests, boulevards, etc. was prohibited;

• Cars, bicycles, and other rent centers were closed;

• Limited personal contacts with the public or clients and between employees and reduction of employees in the workplace to the necessary minimum were allowed.

Penalties for not adhering to the restrictions policy were between 5 and 30 thousand PLN (about 1106–6640 EUR according to the National Polish Bank exchange course on 21st April 2020).

On 21st April, the restrictions were loosened, including the permission to walk outside for other than listed purposes, e.g., for running, walking, cycling, etc. We decided to analyze the data gathered up to 21st April since IPAQ-LF is a 7-day recall questionnaire, which means the questionnaire completed on the 21st was referring to the lockdown period.

Participants

Due to the total lockdown, the ways we could recruit participants were quite limited. We decided to attract potential participants via social media, including Twitter, Facebook (personal, university alumni, physical activity societies, media, and other profiles), private emails, and via radio (interviews in Meloradio and Radio Wrocław) and TV (interview in a regional public TVP 3). Interviews and a short description, and an invitation to participate in the study, were published on the media web pages.

We included all adults who were not quarantined due to the COVID-19 during the surveyed period (the last 7 days). Data of participants who did not tick the consent was not saved.

Survey

The survey was anonymous and was done using Google Forms. The link: https://docs.google.com/forms/d/e/1FAIpQLSfeErFVyUGRav-05JeG_iE7hJ5ITVfceyjfxuQ7FlOE3RBf1Q/viewform?vc=0&c=0&w=1 was used for further dissemination. For convenience, we also used shorter links: https://tiny.pl/7qd7j and https://bit.ly/aktywnoscpolakow.

The survey consisted of three parts. In the first part of the survey, we asked to thick the informed consent box and collected demographic data about sex, age (date of birth), height, body mass, living place size (m²), a number of household inhabitants, accessibility to own garden, and social status (schoolchild, student, white-collar worker, worker, pensioner, unemployed). Participants were able to add their comments too (indicating, e.g., maternity leave).

In the second and third parts of the survey, we used IPAQ-LF. Since there is no more appropriate questionnaire than others (Terwee et al., 2010), we decided to apply IPAQ (Craig et al., 2003) as it is one of the most widely used questionnaires worldwide. It shows good reliability (r = 0.81) and is the most reliable (ICC = 0.97) physical activity questionnaire (Helmerhorst et al., 2012).

We chose the long-form (Hagströmer, Oja & Sjöström, 2006) as it assesses the PA level related to four domains: (1) during self-powered transportation, (2) at work, (3) during household and gardening work, and (4) during leisure time, including exercise and sport participation. VPA refers to activities that take hard physical effort and make you breathe much harder than normal. MPA refers to activities that take moderate physical effort and make you breathe somewhat harder than normal. It also allows collecting detailed information about sedentary activity.

We used the reliable Polish version of the questionnaire (Biernat, Stupnicki & Gajewski, 2007; Biernat, 2013). In the second part of the survey, we used IPAQ-LF to assess the PA level before the pandemic. We asked about seven typical days in January 2020. Although the IPAQ was not ensured to be so much retrospective and the reliability and validity of 3–4 months recall questionnaire could be doubted, we decided to collect the data nevertheless. The third part of the survey consisted of IPAQ-LF dedicated to assess PA level during lockdown (7 days recall).

Data processing and truncation

The time of PA was calculated according to the “Guidelines for data processing analysis of the International Physical Activity Questionnaire (IPAQ)—Short and long forms” (Sjöström et al., 2005). We processed the data following the rules of data cleaning, unreasonably data exclusion, minimum values inclusion, and truncation of data described by Sjöström et al. (2005).

Based on the assumption that an average of 8 h per day is spent sleeping (Sjöström et al., 2005), five participants were excluded from further analysis. Their sums of all Walking, MPA, and VPA time variables were greater than 960 min (16 h) (Sjöström et al., 2005).

A total of 152 records of 111 participants were truncated, i.e., the variables total Walking, total MPA, and total VPA were calculated, and then, if any of them exceeded 180 min (3 h), they were truncated to 180 min.

Given the VPA, MPA, and Walking are multiplied by different coefficients in each PA domain, i.e., occupational, transport, household, and leisure-related PA, it was impossible to use the truncated time to calculate metabolic equivalents-minutes (MET). Instead of using METs as initially proposed by Sjöström et al. (2005) we report the time (minutes per day) for each activity (VPA, MPA, and Walking). Time spent in VPA, MPA, and Walking are widely used measures of PA, e.g., they are used by World Health Organization (2010).

Data analysis

We report descriptive statistics on VPA, MPA, Walking, and Sitting time (min/day).

Bayesian one-sided t-test was used to compare PA before and during the lockdown. We assumed that VPA, MPA, Walking before lockdown > VPA, MPA, Walking during lockdown. The opposite hypothesis was advanced concerning the sitting time, i.e., Sitting before lockdown < Sitting during the lockdown.

The null hypothesis (H₀) assumed an absence of differences between the pre-lockdown and lockdown measurements, whereas the alternative hypothesis (H₁) assumed a directional difference.

Bayes Factor (BF) was used to provide the alternative hypothesis’s quantified evidence relative to the null hypothesis. The evidence categories for Bayes factor (BF) were set as proposed by Jeffreys (1998), Wetzels et al. (2011).

A Cauchy scale parameter of 0.707 (prior) was assigned for the effect size (Rouder et al., 2009).

Bayes factor robustness checks were performed. We used JASP specified wide and ultra-wide priors to examine the extent to which our conclusions depend on our prior specification (van Doorn et al., 2020). They display the Bayes factor as a function of the width of the Cauchy prior on effect size using the wide (scale = 1) and ultrawide prior (scale = √2).

Additionally, we performed pairwise Bayesian Pearson correlation tests.

The Bayesian statistics were performed using JASP (version 0.14.1).

Study population

There were 326 records registered in Google Forms between 8–21 April 2020. Data from one participant was removed as she was 14 years old, whereas the IPAQ-LF is dedicated to adults. Records of 320 participants were analyzed. Three participants were quarantined. Out of 320 participants, 10 participants (six females, four males) indicated they had contraindications (as advised by a medical doctor) to exercise. 45.62% of the participants (n = 146, 93 women, 53 men) declared they had gardens for their own use.

Some of the records were incomplete; however, it did not disqualify them from further analysis, i.e., the missing data referred to the body height or size of the living place.

Descriptive statistics are presented in Table 1. Frequencies of the participants’ social status are presented in Table 2.

Physical activity before and during the lockdown

Results of the Bayesian t-test for VPA, MPA, and Walking are presented in Table 3.

Our hypothesis that VPA time before the lockdown was longer than during lockdown was anecdotally (BF = 1.48) supported by our data (Fig. 1). Although the time spent on VPA before the lockdown was longer, the difference was slightly above 9 min/day. Bayes factor robustness check with wide prior also anecdotally supported H₁ (BF = 1.065). On the other hand, our data anecdotally supported H₀ while the prior distribution was set to ultrawide (BF = 0.76).

In contrast, our data negatively supported the hypothesis that MPA before the lockdown was longer than during the lockdown (see Table 3 and Fig. 2). The evidence against the alternative hypothesis H₁ was strong (BF = 0.046). It is not surprising since the average time spent on MPA was longer during than before lockdown by about 2 min/day (Table 4). The Bayes factor robustness check with wide and ultrawide priors also supported H₀ (BF = 0.032; BF = 0.023; respectively).

Our data strongly supported H₁ related to Walking time (BF = 26.804). We assumed that Walking time before the lockdown was shorter than during lockdown. As shown in Table 3 and Fig. 3, the mean Walking time decreased from 72.431 min before lockdown to 56.528 min/day during the lockdown. Our data still strongly supported H₁ when the prior distribution was set to wide and ultrawide (BF = 19.62; BF = 14.12; respectively).

Sedentary behavior

We tested our H₁ related to sitting time, assuming that Sitting time before lockdown will be shorter than during lockdown. The H₁ was strongly (decisively) supported by our data (Fig. 4). The BF = 866.38 (see Table 4). The Sitting time during lockdown increased by an average of 40%, i.e., by almost 100 min/day compared to the pre-lockdown period. Our data extremely strongly supported H₁ also with wide (BF = 646.5) and ultrawide prior (BF = 470.4) (Fig. 5).

VPA, MPA, Walking, and Sitting time correlates

One of the most bothersome restrictions during the lockdown was to stay-at-home order. People could have been physically active only in their houses and gardens unless still working professionally. Therefore, we decided to see if there were any correlations between the size of their living places (m²) and the number of inhabitants sharing a living place. We divided the living place size by the number of inhabitants, and we correlated it with VPA, MPA, Walking, and Sitting time. We performed pairwise Bayesian Pearson correlation tests. We tested H₁ that the available living space (m²/person) is positively correlated with VPA, MPA, Walking against H_0, assuming there is no such correlation. The H₁ was anecdotally supported by our data (r = −0.052, BF = 0.107; r = −0.007, BF = 0.071; r = −0.016, BF = 0.073; for VPA, MPA, Walking accordingly).

Regarding the Sitting time, we advanced the opposite hypothesis (H₁) that that available living space (m²/person) is negatively correlated with Sitting time. Likewise, this hypothesis was anecdotally supported by our data (r = −0.077, BF = 0.031).

To assess whether participants with their garden were more active than those without, we applied the Bayesian t-test for VPA, MPA, Walking, and Sitting time. We tested H₁ that garden owners will be more physically active than the participants without gardens. All three tests did not yield supportive evidence toward these hypotheses (Table 5).

On the other hand, we hypothesized that participants without their garden would be sitting longer than garden owners. Conversely, our data strongly supported H₀, i.e., the groups were not different in this regard (BF = 0.077). Garden owners spent on average 405.589 ± 552.521 min/day sitting during lockdown while those without gardens 371.803 ± 262.102.