Introduction
The COVID-19 global pandemic is ongoing, caused by the SARS-CoV-2 virus. The first outbreak of infection was in China in December 2019, which was followed by spread to many other countries. The first case was detected in Australia on 25th January 2020. This case was in Melbourne, the capital of the state of Victoria. In the absence of vaccines at the time, the Victorian government attempted to control the spread of the virus using lockdowns, beginning on the 30
th March 2020. Victoria, and the city of Melbourne in particular, was the region of Australia most impacted by the pandemic and by series of lockdowns, which are summarized in Table
1. As in the rest of the world, the initial concern about the pandemic was the effect on mortality and hospitalizations. However, as the pandemic progressed, the potential effects on mental health have also been recognized [
1]. Given the strong evidence for increased psychological distress and mental disorders after disasters [
2], the prevalence of mental health problems was expected to rise due to infection and fear of infection, as well as the impacts of lockdowns, including social isolation, loss of employment and income, and intensification of pressures on families.
Table 1
Chronology of COVID-19 Restrictions in Victoria Up Until the Second Survey
25 Jan 2020 | First COVID-19 case in Victoria (Melbourne) |
30 March – 12 May 2020 | Lockdown 1. This covered the whole of Victoria. People could only leave home for 4 reasons—food and supplies, exercise, medical care, work and education (if necessary). Non-essential businesses and schools were closed. People were required to work from home where possible. Gatherings of more than 2 people were banned (except for immediate household or for work or education purposes) |
13 May – 8 July 2020 | Restrictions were progressively eased across Victoria, but lockdowns were re-imposed on some areas of Melbourne that had new infections |
9 July – 27 October 2020 | Lockdown 2. This initially covered Melbourne and one shire adjacent to Melbourne, but restrictions were later imposed on regional Victoria from 4 August 2020. A night-time curfew was added for Melbourne from 2 August 2020 |
There is now an extensive international literature exploring the mental health consequences of the pandemic. Early studies involved cross-sectional comparisons of the prevalence of symptoms of distress, anxiety and depression during the early months of the pandemic (March–April 2020) compared to pre-pandemic data; they found that the prevalence of symptoms was higher than pre-pandemic [
5,
6]. Subsequent studies used stronger longitudinal designs and repeated cross-sectional surveys of representative samples. Systematic reviews and meta-analyses of these studies found that symptoms of distress, anxiety and depression increased during the early months (March–April) of the pandemic in 2020 but returned to pre-pandemic levels by the middle of that year [
7,
8].
While longitudinal studies with frequent waves have been proposed as the best methodology for investigating pandemic-related mental health impacts [
9], survivorship bias has been raised as a limitation of these studies. Czeisler et al. [
10] analysed a 4-wave longitudinal study from the USA and found that people who participated in only 1 or 2 waves had a higher prevalence of depression and anxiety symptoms. This selective pattern of retention could lead to an overly optimistic interpretation of mental health over time. Czeisler et al. [
10] therefore recommended the simultaneous use of independent cross-sectional data along with longitudinal data, as these approaches have different strengths and weaknesses, and facilitate the investigation of distinct research questions.
Most studies of the impact of the pandemic have assessed symptoms of mental disorders. However, Keyes [
11] argued that a complete assessment of mental health should take account of a dimension of positive mental health as well as symptoms. Relevant to this two-dimensional concept, there have also been some studies on the effects of the pandemic on positive mental health, including subjective well-being, life satisfaction and positive affect. There has been less research carried out on positive mental health than on symptoms, but one systematic review and meta-analysis of studies up to June 2020 found no significant effect of COVID-19 pandemic lockdowns on measures of life satisfaction, positive affect, well-being and quality of life [
12].
In addition to research examining changes in mental health measures, there have been studies of factors that may impact any changes found. A particular focus of pandemic studies has been social isolation and loneliness, as these are potentially adversely affected during lockdowns. While there was some evidence of an increase in these factors during the early phases of the pandemic [
13], other systematic reviews found a lack of effect [
7,
12]. One explanation of these findings is that there was an initial impact of restrictions and lockdowns on social relationships, but people soon adapted by finding alternative online means of social contact [
12,
13] and that the shared experience of the pandemic may have increased social cohesion [
12]. Studies from other types of disasters have also indicated that social factors such as access to social support and sense of social solidarity are associated with more positive mental health outcomes [
14]. However, it was also shown that these factors can deteriorate over time.
The above findings mainly come from studies in Europe and North America and may not be generalizable to all countries. We focus on the studies with stronger methodologies involving more systematic and representative samples.
From early in the pandemic, the Australian Bureau of Statistics (ABS) has monitored the health and social impacts [
15]. It found that symptoms of psychological distress increased in April 2020 when the nation was first affected by lockdowns, declined to pre-pandemic levels by June as restrictions eased, and then increased again in August when there was another wave of lockdowns. Another source of longitudinal data has come from the Life in Australia Panel [
16]. This study found that the prevalence of severe psychological distress rose in April 2020 compared to pre-pandemic, but then fell in May. It worsened again from May to August, mainly due to the effects of a second lockdown in the state of Victoria. This impact subsequently declined, with Victoria showing little difference from the rest of the country by November 2020. A third longitudinal panel study is the Australian National COVID-19 Mental Health and Risk Communication Survey, which involved fortnightly on-line surveys from March to June 2020 [
17]. This study found that any changes in depression and anxiety symptoms were generally transitory. The study also found that while suicidal ideation was high, it did not vary over time [
18]. Later work from this study reported greater psychological distress in parents and caregivers who were home schooling, but no effect on wellbeing [
19]. Overall, the Australian data show a worsening of mental health which was associated with lockdowns, but this effect diminished over time.
Australian studies of mental health outcomes from other types of disaster are also of relevance, including an epidemiological study showing increased risk of psychiatric disorders post disaster [
20], cross-sectional analyses demonstrating a complex relationship between social networks and individual mental health post bushfires [
21], and longitudinal analyses indicating moderate involvement in community groups can be a protective factor for mental health following bushfire exposure [
22].
In the present study, we report new repeated cross-sectional and longitudinal data from the Australian state of Victoria, which was the region of Australia most affected by pandemic lockdowns. The capital city of Melbourne was particularly affected and by 2021 was one of the most locked-down cities in the world [
23]. Consistent with Keyes’ [
11] two-factor concept of mental health, the study aimed to assess both symptoms of psychological distress and life satisfaction. We aimed to investigate cross-sectional factors associated with higher psychological distress and life satisfaction at each survey, and also social connection factors associated with longitudinal changes between the two surveys.
Method
This paper builds on the descriptive findings of the Victorian Coronavirus Wellbeing Impact Study surveys [
24], which showed significantly higher rates of low-medium life satisfaction over time and a non-significant trend towards higher rates of psychological distress over time.
Study design
The current analysis has a two-part study design: a repeated cross-sectional study of respondents who participated in Survey One and Survey Two, followed by a longitudinal nested cohort study of these same respondents. The surveys were undertaken by the Victorian Health Promotion Foundation (VicHealth), a state-based government agency in Victoria, Australia, with a remit to promote health and prevent illness.
The VicHealth Coronavirus Victorian Wellbeing Impact Surveys of Victorian residents aged 18 years and over were conducted via an opt-in ‘research only’ online panel (i.e. non-probability panel). The surveys were designed to track the impact of the pandemic and associated lockdowns on a range of behavioural and attitudinal health risk factors during the first two waves of the pandemic in Victoria, from March to June in 2020 and from July through to October 2020.
Survey One commenced on 31 May 2020 and concluded 8 June 2020. The total achieved sample size was 2,000. Survey Two, which was conducted during the second pandemic wave, commenced on 10 September 2020 and concluded on 21 September 2020. Survey Two included 1,008 respondents who were re-contacted from Survey One and 992 ‘new’ respondents (i.e. those who did not complete Survey One), to boost the total sample size to 2,000.
Participants
The opt-in online panel used for both surveys was LiveTribe, a research-only panel operated and managed by i-Link Research
https://www.i-linkresearch.com/. LiveTribe panellists are recruited via print media, online marketing initiatives, direct mail, social media platforms, affiliate partnerships, personal invitations and a range of other ad-hoc initiatives. For Survey One and Two the sample size of 2,000 was chosen as it provides a reasonable margin of error for the purposes of estimating population level results within an expedited fieldwork turnaround period.
Survey questionnaire
The 20-min survey questionnaire (see Supplementary file
1) covered a range of health and wellbeing factors, including life satisfaction, subjective wellbeing, psychological distress and social connection, as well as socio-demographics. The selection of health and wellbeing indicators for the questionnaire was guided by several key principles previously used in population surveys conducted by VicHealth [
25]. These principles included: sensitivity to change across person, place and time; strong psychometric properties; being amenable to action at a range of jurisdiction levels including local government; non-duplication of other Victorian population surveys; and brief questions that could be feasibly used in local program evaluations, thus allowing the population measure to act as a comparator for local evaluations.
Different question styles were used to minimise respondent fatigue and enhance engagement with the survey, such as Likert scales, closed-ended questions and open-ended questions. Current guidelines were followed to ensure questions were as user-friendly as possible for respondents, regardless of the device being used to access the survey (for example, mobile phones, tablets, desktops or laptops) [
26]. Respondents were also asked to provide their nearest cross-street to enable the application of geo-codes to participants’ identification codes and their question responses.
No formal pilot testing of the survey was undertaken. However, a soft launch was undertaken to confirm the integrity of the questionnaire. The soft launch involved inviting a small number of participants to complete the survey, with the aim of securing approximately 20 completed surveys. The data from these surveys were carefully checked against the Microsoft Word version of the survey to ensure the survey was error free and had been scripted as expected. No errors were detected as a result of this process and the survey was launched in full without any amendment. The average completion time of the questionnaire was 20 min.
Measures
Our primary outcomes were psychological distress and life satisfaction, and the primary exposures were feeling connected with others, staying connected with family and friends, and social solidarity.
‘Psychological distress’ was measured using the Kessler Psychological Distress Scale-6 (K6) which has excellent internal consistency reliability (Cronbach's alpha = 0.89) [
27]. The K6 is a combined score across 6 areas of psychological distress; each person can score a minimum of 6 and maximum of 30. Scores of 19 or more are classified as probable serious mental illness, while scores of 6 to 18 are classified as no probable serious mental illness. Null responses to 2 or more of the 6 statements were excluded from the mean calculation, with adjustments made for those who gave a null response to 1 statement.
‘Life satisfaction’ [
28] was derived from a rating of satisfaction with life as a whole using a scale of 0 to 10 where 0 is completely dissatisfied and 10 is completely satisfied. The measure correlates highly with the Personal Wellbeing Index measure of life satisfaction (
r = 0.79) [
29]. Low to medium life satisfaction was defined as a score between 0 and 6 out of 10. Null responses were excluded from mean calculation.
The measure of ‘Feeling connected with others’ [
30] uses a six item Likert scale to assess level of agreement with the statement ‘I feel connected with others’. The measure has a high positive correlation (+ 0.79) with the Social Connectedness Revised Scale and a high negative correlation (-0.69) with the UCLA Loneliness Scale [
30]. For analysis, responses were divided into the percentage of people who selected strongly disagree, disagree, mildly disagree, mildly agree, and those who selected agree or strongly agree.
‘Staying connected with family and friends’ was a question developed for the survey and to the best of our knowledge it is the first time that this measure has been used. It measures the extent to which people find it easy or hard to stay connected with those close to them. For analysis, the five-point Likert scale was coded as the percentage of people who reported it was 1. very easy or easy; 2. neither or 3. hard or very hard.
The
‘Social solidarity’ measure [
31] was designed to determine how close people feel with their communities using a combined score across six questions and has good reliability across samples (Cronbach’s alpha ranging from 0.78 to 0.89). These questions ask respondents whether they agree with statements regarding their connection with their local community. Responses for all six questions were assigned the following values: Strongly disagree = 1, Disagree = 2, Neither agree nor disagree = 3, Agree = 4, Strongly agree = 5. Any respondent providing a ‘don’t know’ or ‘prefer not to answer’ response to any of the six questions was excluded from the analysis. The final score was calculated by summing the values of the six categories out of a maximum of 30 and minimum of six.
In addition to the above, the variables age, gender, disability, income, the main activity in September, region, and household composition were used as covariates in the longitudinal analysis. The covariates used for each analysis are shown in the corresponding tables.
Ethics
Ethics approval for Survey One was provided by the Australian National University Human Research Ethics Committee (2020/264) on 20 May 2020. Ethics approval for Survey Two was provided by the Australian National University Human Research Ethics Committee (2020/540) on 8 September 2020. All methods were carried out in accordance with relevant guidelines and regulations. Online written informed consent was obtained from study participants as this survey was carried out as an online survey.
Statistical analysis
Data were analysed using Stata, V.17 [
32] and R software [
33]. Categorical data were summarised using frequency with percentages, continuous data using means with standard deviations, and skewed data using medians with interquartile ranges (25th –75th centile). Our primary outcomes, exposures and covariates are described in the Measures section. To assess differences in dependent means and proportions paired sample t-tests and McNemar’s tests were used respectively. Using logistic regression modelling, we explored the associations between our exposures and primary outcomes of psychological distress and life satisfaction with and without adjustment for covariates. The results from the multivariable models were summarised using adjusted Odds Ratios (aOR), 95% Confidence Intervals (CI) and p-values. To identify confounding variables for the regression analyses, we used information from previous literature and Directed Acyclic Graphs (DAG) [
34,
35]. For each outcome, we conducted three separate analyses: (i) a cross-sectional analysis at the first lockdown, (ii) a cross-sectional analysis at the second lockdown, and (iii) a longitudinal analysis exploring the change in outcome measures between the two lockdowns. We tested for exposure-outcome associations that may have been modified by other covariates using interaction terms and likelihood ratio tests, and used measures of fit for logistic regression [
36] to test model assumptions. We used complete-case analysis (i.e., analysis restricted to participants with complete data), and investigated the baseline first-lockdown characteristics of those missing and not-missing at the second lockdown using summary statistics.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.