How the mental health of working parents in Australia changed during COVID-19: A pre- to during pandemic longitudinal comparison

Abstract

Background

Cross-sectional studies indicate that mental health has deteriorated in Australia during the COVID-19 pandemic, including for parents. However, robust longitudinal studies interrogating change from before to during the pandemic remain rare. The current study analysed data from Australian parents surveyed in 2016 and August 2020. We investigate whether distress was higher in the COVID-19 period compared to pre-pandemic levels, and whether any increases in distress were greatest for parents living in Victoria (who had entered their second prolonged lockdown).

Methods

A community cohort of Australian working parents (n = 5197) was recruited online in 2016. In August 2020, 25.9 % (n = 1348) completed a follow-up survey. Analyses were restricted to those employed at both time-points (n = 1311). Random effects longitudinal models examined the association between time (i.e. pre vs. during-pandemic) and distress (K6 scale). Fixed effects models specifically tested change between time periods in association with change in distress.

Limitations

The initial sample were recruited online with highly educate parents over-represented. Attrition between survey time-points may also limit generalisability.

Results

All models consistently showed that the pandemic period was associated with greater distress. Overall, serious mental illness (i.e. K6 score ≥ 18) increased by 5.3 percentage points (from 8.0 to 13.3). This increase was greater (by 4.7 percentage points) for those parents in Victoria.

Conclusions

This study is one of few to longitudinally assess mental health pre- to during the pandemic. Psychological distress and serious mental illness increased for Australian working parents, and this effect was greatest for those experiencing a prolonged lockdown in Victoria.

1. Introduction

A public health emergency of international concern was declared by the World Health Organization on the 30th January 2020 in response to an outbreak of COVID-19. By March 2022, >6.1 million deaths had been reported globally (World Health Organisation, 2022). In Australia, alongside concerns about the proximal physical health impacts of COVID-19, there has been much discussion about the broader economic and social consequences resulting from prolonged uncertainty, restrictions and lockdowns. Much of the public discourse focused on the unfolding mental health consequences – including discussion on social media, expert commentary and individual stories in the mainstream media, and reports from community mental health organisations (e.g. Lifeline's three busiest days on record have all been this month - 9news.com.au).

The research community in Australia has also paid considerable attention to investigating the possible deterioration in mental health over the COVID-19 period and if so, for whom. Much of this research has been cross-sectional – and the majority of Australian surveys that have compared levels of depression, anxiety, distress and loneliness to pre-pandemic benchmarks have reported significant increases (e.g. Dawel et al., 2020; Fisher et al., 2020; Holton et al., 2020; Li et al., 2021; Newby et al., 2020; Van Agteren et al., 2020). However, while there is a plethora of cross-sectional research, published longitudinal studies with measures of mental health both pre- and during the pandemic from the same sample or cohort remain scarce. Our review of Australian community-based studies examining mental health in the COVID-19 period found studies with repeated cross-sectional snapshots (e.g. Taking the Pulse of the National (TTPN)) (Broadway et al., 2020; Botha et al., 2022) and the Life in AustraliaTM study (Biddle et al., 2020) (although the Life in Australia study has a longitudinal sub-sample), some with planned longitudinal assessments during the pandemic (Dawel et al., 2020; Westrupp et al., 2021a), but only a few longitudinal studies with pre-pandemic baseline data (e.g. adolescent samples Magson et al., 2021 and Munasinghe et al., 2020; population-based representative sample Butterworth et al., 2022). Further longitudinal cohort studies including pre-pandemic data would aid in robustly attributing any change in mental health over time to the pandemic context rather than individual or sample differences (Butterworth et al., 2020).

If we look internationally, published longitudinal studies with baseline pre-pandemic data are still rare, but are more common in Europe. For example, a recent 2022 systematic review and meta-analysis (Robinson et al., 2022) of longitudinal cohort studies comparing mental health pre and during the pandemic reported data from 65 studies, but only two of these were from Australia – both of which included samples from adolescents (Magson et al., 2021; Munasinghe et al., 2020). This systematic review reported a small increase in mental health symptoms early in the pandemic, but that this reduced and was comparable to pre-pandemic levels by mid-2020.

In terms of who has been most vulnerable to poorer mental health during the pandemic, parents have emerged as a likely at-risk group. Typically, research shows that being a parent is protective for mental health (Dehara et al., 2021; Helbig et al., 2006). However, throughout the pandemic period many parents' work and care roles collided in an unprecedented manner, with universal disruptions to how parents with dependent children managed their work-care routines. It is estimated that nearly 70 % of Australian workers worked at home under lockdown orders during 2020, with almost half of these working at home for the first time (Baxter and Warren, 2021). Irrespective of job type and working-at-home arrangements, parents in a range of contexts faced increased unpaid work and care loads, coupled with an absence of support from the usual formal and informal childcare, school, and family networks (Ruppanner et al., 2021). Studies focusing on parents' mental health during the pandemic propose that there has been a decline from pre-pandemic levels (Westrupp et al., 2021a), with some indications that this decline has been greater for parents than non-parents (Broadway et al., 2020). Higher rates of poor mental health have been found for people living in Victoria (versus living in other states) (Fisher et al., 2021; Butterworth et al., 2022), where a prolonged lockdowns was in place from July–October 2020, including for parents living in Victoria (Westrupp et al., 2021b).

The existing research base provides strong evidence that mental health has worsened for parents over the COVID period in Australia, especially for those restricted by the severe and prolonged lockdowns in Victoria. However, it is predominantly cross-sectional or restricted to change during the COVID period. There is an absence still of high-quality evidence from robust longitudinal studies including baseline pre-pandemic data. The current study addresses this gap, using data collected from a diverse sample of employed Australian parents assessed both prior to the pandemic in 2016 and then again during the peak of the Victorian lockdown in August 2020.

1.1. Aims

The current study aimed to investigate whether levels of distress for employed parents were greater in the COVID period (August 2020) compared to pre-pandemic levels of distress four years earlier. The study also explored whether any increases in distress during COVID were greater for working parents who were living in Victoria (i.e. those experiencing a prolonged lockdown during the survey period) versus those living in other states or territories (who were not under lockdown during this time).

2. Method

2.1. Sample and study design

This study uses data from the Families at Work (FAW) project, which originally surveyed 5197 employed parents in August 2016 (i.e. pre-pandemic baseline). Recruitment for the initial sample was online using a combination of paid Facebook and unpaid advertising targeting working parents. Parents were invited to complete an online survey about ‘balancing work and family’. Further details of recruitment procedures and sample are reported elsewhere (Bennetts et al., 2019). Eligible parents were residing in Australia; at least 18 years of age; parent of at least one child aged 18 or younger; and in paid employment. Approval was provided by La Trobe University Human Research Ethics Sub-Committee (S16–112). While the survey sample was recruited via convenience sampling (rather than a random sampling frame), fathers and single parents (who are typically under-represented in research about parents) were purposively recruited to improve their representation within the sample. Previous work comparing the FAW sample with the population-based Longitudinal Survey of Australian Children has indicated that the FAW sample is broadly representative of Australian parents, although those with tertiary education are over-represented and the level of psychological distress in the FAW sample is higher (K6 distress scale mean 11.5 vs mean 10.1; p < .001) (Bennetts et al., 2019).

In August 2020 during the pandemic, the original FAW sample was invited via email to participate in a follow-up study. During this time, most states in Australia were experiencing a reprieve from lockdowns, however Victoria had entered its second lockdown. In total, 1348 participants consented to participate in the follow-up (25.9 %), a robust number given participants had not been contacted since the first survey. The sample for the current study was restricted to those who were employed at both time-point 1 and 2 to reduce the influence of job loss during the pandemic (all participants were employed at Time 1 as this was part of eligibility to participate), leaving 1311 participants in the analysis sample. As it was four years since the original study, some parents had new children while for others their youngest child was now aged over 18 years (i.e., for 6.2 % their youngest child was aged 18+ years, with a maximum of 22 years old). Parents with young adult children were not excluded from the sample as these parents may also have been experiencing important stressors exacerbated by the pandemic (e.g., supporting their young adult children in contexts such as transitioning out of secondary school, returning home from tertiary study institutions/training, job loss or working from home).

Several preliminary analyses compared the characteristics of the sample who only participated in the original survey with those who also completed the follow-up (see Supplementary Table 1), to explore whether attrition may have biased the follow-up sample to over-represent those with poor mental health. First, these analyses showed that those who completed the follow-up were significantly more likely at baseline to be female, have a tertiary education, to live in Victoria and were less likely to work long (46+) hours (importantly there were no significant baseline difference in psychological distress). These characteristics were examined further to explore potential associations with psychological distress - neither being female or living in Victoria were significantly associated with distress at baseline, although the lowest level of education and working long hours were significantly associated with baseline distress. Further analyses examining whether these characteristics (low education and long work hours) might be over-represented specifically in the Victorian sample found that they were significantly under- rather than over-represented. Finally, interactions were examined between baseline sample characteristics and state location (i.e. living in Victoria vs. other state) to investigate any association with attrition. None were found to be statistically significant, suggesting that any differences in characteristics between the original and analytic sample were not markedly different between those living in Victoria and elsewhere.

2.2. Measures

Psychological distress was measured using the K6 Psychological Distress Scale (Kessler et al., 2010). The K6 is a well-established measure of psychological distress commonly used to screen for and identify serious mental health problems (most strongly related to affective and anxiety disorders). The K6 includes six items that ask about how often in the past 30 days participants felt: nervous, hopeless, restless or fidgety, so depressed that nothing could cheer them up, that everything was an effort, and worthless. Possible responses to all items were: 1 - none of the time, 2 - a little of the time, 3 - some of the time, 4 - most of the time, 5 - all of the time. A total scale score was calculated by summing all the items (range 6–30). In the current study, the items from the K6 scale demonstrated high internal consistency with a Cronbach Alpha of 0.86. The established threshold of ≥18 was used as a binary indicator of high psychological distress likely to indicate serious mental illness (National Comorbidity Survey, 2021), with previous Australian research demonstrating the sensitivity of the K6 to detect mood and anxiety disorders (Furukawa et al., 2003). The nationally representative longitudinal Household Income and Labour Dynamics in Australia (HILDA) study (see Leach et al., 2014 for more information on HILDA) collects data on the K6 every 2 years. These data indicate a normative very slight increase in the K6 score for parents with children aged ≤22 from a mean of 9.6 in 2015 to 9.8 in 2019 (across a 4 year period).

Covariates A number of variables measured at either Time 1 (pre-pandemic) or at Time 2 (during the pandemic) were included in the analyses to control for their influence on parents' mental health. At Time 1, participants reported their gender, age, education (Year 12 or below, post-school qualification, tertiary qualification), the state or territory they lived in (coded as Victoria vs all other), whether they were a single parent (yes, no), the number of children they had (coded as 1, 2, 3+), age of youngest child in the household (coded as child/ren ≤6 years vs ≥7 years), their work hours (coded as 0–14, 15–29, 30–45, 46+), and whether they felt their job was insecure (yes, no). All of these variables were again asked about at Time 2 during the pandemic, except for education level. At Time 2, parents were also asked about whether they were coping financially (yes, no), whether they owned their own business (yes, no), and whether they had worked at home during the COVID-19 period (yes, no). Those who were partnered were also asked whether they felt they were doing more than their fair share of domestic load (yes, no), whether they felt they were doing more than their fair share of childcare (yes, no), and whether their partner was employed (no, yes but not working any hours, yes and working hours).

2.3. Statistical analyses

Random effects models were initially conducted to examine the effect of time period (pre- vs during COVID-19) on overall levels of psychological distress. First, only time period was included in the model, with a random intercept to allow participants' initial distress level to vary. Subsequently, covariates that may have impacted on psychological distress during the pandemic (i.e. Time 2) were included – starting with socio-demographic factors (model 2), family structure (model 3), work characteristics (model 4) and domestic/care roles for those partnered (model 5). The analytic sample for all models was restricted to the model with the smallest sample to allow for comparability across models.

Fixed effects models were then conducted to specifically investigate within-person change in mental health over the time period (from pre to during pandemic). By focusing on within-person change, fixed effects models control the effect of all time-invariant characteristics, regardless of whether they are measured in the study (a source of bias in other cross-sectional studies including pre-pandemic benchmarks/estimates from different samples) (Singer and Willett, 2003). However, this means that models can only incorporate main effects for time varying characteristics. Our fixed effects models included a small number of time-varying covariates (partner status, work hours and job insecurity), but variables that changed uniformly one way were not included as they were too highly correlated with the COVID/time period-effect (e.g. parent age and number of children uniformly increased, and age of youngest child increased for the vast majority of the sample as few new children entered the families).

All random effects and fixed effects models were first conducted using the continuous K6 outcome measure to investigate change in distress levels. Models were then repeated using the binary indicator for serious mental illness to explore change in likely cases of mental disorder. An interaction between time period and gender (male vs female) was also briefly explored to examine whether change in distress differed between males and females. Subsequent to the initial main effect of time period (pre- vs during pandemic) being tested, models included an interaction between time period and location (Victoria vs non-Victoria) to examine change in mental health based on whether participants were in Victoria. The maximum amount of missing data for any one variable was 6 % (see Table 1 ), and no missing data were imputed. All analyses were conducted in STATA version 16 (StataCorp, 2019).

Table 1.

Descriptive statistics for pre and during the COVID-19 pandemic (n = 1311).

	Pre-pandemic (2016) Mean (SE) or n (%)	During Pandemic (2020) Mean (SE) or n (%)
Gender
Male	301 (23.0 %)	304 (23.2 %)
Female	1007 (76.8 %)	1006 (76.7 %)
Other	1 (0.1 %)	1 (0.1 %)
Missing	2 (0.2 %)	0
Age (W1 range 29–64 years)	40.5 (0.17)	44.3 (0.17)
Missing	0	0
Education
Secondary or below	86 (6.6 %)	–
Post-secondary	258 (19.7 %)	–
Tertiary	967 (73.8 %)	–
Missing	0	–
State/Territory
Other than Victoria	803 (61.2 %)	800 (61.0 %)
Victoria	508 (38.8 %)	510 (38.9 %)
Missing	0	1 (0.1 %)
Coping financially
Yes	–	1193 (91.0 %)
No	–	88 (6.7 %)
Missing	–	30 (2.3 %)
Single Parent
Yes	192 (14.7 %)	230 (17.5 %)
No	1119 (85.3 %)	1080 (82.4 %)
Missing	0	1 (0.1 %)
Young child ≤6 years
Yes	795 (60.6 %)	364 (27.8 %)
No	516 (39.4 %)	942 (71.9 %)
Missing	0	5 (0.4 %)
Number of children
1	355 (27.1 %)	267 (20.4 %)
2	658 (50.2 %)	709 (54.1 %)
3+	298 (22.7 %)	331 (25.3 %)
Missing	0	4 (0.3 %)
Work hours per week
0–14	47 (3.6 %)	97 (7.4 %)
15–29	324 (24.7 %)	283 (21.6 %)
30–45	736 (56.1 %)	746 (56.9 %)
46+	203 (15.5 %)	164 (12.5 %)
Missing	1 (0.1 %)	21 (1.6 %)
Own business
Yes	–	127 (9.7 %)
No	–	1176 (89.7 %)
Missing	–	8 (0.6 %)
Job insecure
Yes	243 (18.5 %)	222 (16.9 %)
No	1068 (81.5 %)	1059 (80.8 %)
Missing	0	30 (2.3 %)
WFH in COVID
Yes	–	956 (72.29 %)
No	–	315 (24.0 %)
Missing	–	40 (3.1 %)
High domestic loada
Yes	–	204 (18.9 %)
No	–	833 (77.1 %)
Missing	–	43 (4.0 %)
High childcare loada
Yes	–	188 (17.4 %)
No	–	850 (78.7 %)
Missing	–	42 (3.9 %)
Partner employeda
Yes, working hours	–	929 (86.0 %)
Yes, but no hours	–	40 (3.7 %)
No	–	91 (9.4 %)
Missing	–	20 (1.9 %)
K6 (range 6–30)	11.3 (0.12)	13.1 (0.13)
Missing	0	79 (6.0 %)
Serious mental illness (K6 ≥ 18)
Yes	97 (7.4 %)	158 (12.05 %)
No	1214 (92.6 %)	1074 (81.9 %)
Missing	0	79 (6.0 %)

^aDenotes items that were only asked of those who were partnered (n = 1080). WFH: work from home.

3. Results

3.1. Descriptives

The sample characteristics at each wave are shown in Table 1. The sample was two-thirds female with an average age of 40.5 years at Time 1. As previously described, everyone in the analytic sample was a parent, and all were working at both time-points. A major proportion of the sample had completed tertiary education (73.8 %), almost 40 % were living in Victoria, and almost 7 % reported having financial difficulties during COVID-19. In terms of family characteristics, the proportion of single parents increased slightly over time (14.7 % to 17.5 %), and as families aged over the four years they were less likely to have a child aged 6 or under and less likely to have only one child. In terms of job characteristics, there were more people working 0–14 h during the pandemic, but there was little change in the number of people reporting job insecurity. During the pandemic, just over two-thirds of the sample reported working from home. For those who had partners – about a fifth said they were doing much more than their fair share of domestic work and childcare during the pandemic, and 86 % had a partner who was also working. In terms of psychological distress, mean scores increased from pre to during COVID-19, as did the proportion of people scoring over the threshold indicating serious mental illness.

3.2. Distress across the sample from pre to during COVID

Table 2 shows the results for the random effects models examining the impact of the COVID-19 period on mental health. The table shows that levels of psychological distress were significantly higher during the pandemic than four years earlier. The initial model (1) showed that distress was 1.79 points higher on average during COVID-19, and subsequent models (2–4) confirm that this increase was sustained after controlling for other potential influences during the pandemic period. When additional variables representing the division of domestic/care and paid work between partners were included in the model (for only those partnered in model 5) the results continued to show an increase in distress during COVID-19 (B = 1.96, 95 % CI:1.69, 2.23). When models 1–4 were run with only those partnered, the estimate remained at approx. 1.95 across all models. The overall stability of the pandemic period effect suggests that the impacts on parent mental health were similar regardless of working from home, and the division of home, care and paid work (for those partnered).

Table 2.

Random effects models predicting K6 distress score (K6 range: 6–30) (n = 1266).

	Model 1	Model 2	Model 3	Model 4	Model 5
Period (2020)	1.79 (1.54, 2.05)	1.79 (1.54, 2.05)	1.79 (1.54, 2.04)	1.81 (1.56, 2.06)	1.96 (1.69, 2.23)
Gender (male)		−0.10 (−0.60, 0.41)	0.00 (−0.51, 0.51)	−0.03 (−0.54, 0.49)	0.26 (−0.31, 0.83)
Age (29–64)		−0.05 (−0.09, −0.02)	−0.06 (−0.10, −0.02)	−0.07 (−0.10, −0.03)	−0.06 (−0.11, −0.02)
Education (Year 12)		Reference	Reference	Reference	Reference
Tertiary		−0.61 (−1.45, 0.23)	−0.52 (−1.36, 0.33)	−0.65 (−1.49, 0.18)	−0.86 (−1.80, 0.09)
Postgraduate		−0.19 (−1.11, 0.73)	−0.17 (−1.09, 0.75)	−0.29 (−1.20, 0.62)	−0.82 (−1.84, 0.21)
Victoria (yes)		0.36 (−0.07, 0.78)	0.32 (−0.10, 0.74)	0.39 (−0.03, 0.81)	0.39 (−0.08, 0.86)
Financial coping (no)		2.30 (1.48, 3.12)	2.20 (1.37, 3.02)	1.84 (1.02, 2.65)	1.95 (0.93, 2.96)
Single (yes)			0.71 (0.15, 1.28)	0.66 (0.11, 1.22)	–
Child aged ≤6 (yes)			0.12 (−0.43, 0.66)	0.15 (−0.39, 0.69)	0.06 (−0.53, 0.64)
Number children (1)			Reference	Reference	Reference
2			−0.57 (−1.12, −0.02)	−0.53 (−1.07,0.01)	−0.55 (−1.18, 0.08)
3+			−0.59 (−1.23, 0.04)	−0.49 (−1.11, 0.13)	−0.47 (−1.17, 0.24)
Work hrs (30–45)				Reference	Reference
0–14				0.12 (−71, 0.95)	0.14 (−0.80, 1.07)
15–29				−0.17 (−0.69, 0.35)	−0.17 (−0.76, 0.42)
46+				0.27 (−0.37, 0.90)	0.21 (−0.47, 0.89)
Own business (yes)				0.48 (−0.22, 1.18)	0.47 (−0.30, 1.24)
Job insecure (yes)				1.80 (1.38, 2.22)	1.89 (1.42,2.36)
WFH COVID (yes)				0.05 (−0.44, 0.54)	0.23 (−0.33, 0.78)
High domestic (yes)					0.30 (−0.41, 1.00)
High child care (yes)					0.90 (0.16, 1.64)
Partner employed (no)					Reference
Yes, but no hrs					0.16 (−1.21, 1.53)
Yes, working hrs					0.54 (−0.28, 1.37)

Notes. All variables taken from Time 2 (2020) except for gender and education (from Time 1). n = 1266 models 1–4; n = 1033 model 5. Model 5 includes only those partnered. WFH – work from home.

A further fixed-effects model (n = 1266) was conducted, controlling for changes in the time-varying variables (i.e., partner status, work hours and job insecurity) and similar results were found. Compared to pre-pandemic levels, on average survey respondents reports a 1.84 point increase on the psychological distress scale during the pandemic (B:1.84, 95 % CI:1.58, 2.10). All models were repeated including an interaction effect for gender; however, no evidence of a statistically significant interaction effect was found.

Table 3 below shows the results for the random effects logistic regression models investigating the odds of having a serious mental illness (binary outcome) during COVID-19. The pandemic period was associated with a 233 % increase in likelihood of serious mental illness. The initial model (1) showed that the odds of serious mental illness was 2.33 higher during the pandemic, and subsequent models (2–4) confirmed that this increase was sustained after controlling for other potential influences. When additional variables representing the division of domestic/care and paid work between partners were included (in model 5) the odds were slightly greater at 2.73 (95 % CI: 1.69–2.23). When models 1–4 were re-run with only those partnered the odds remained at approx. 2.73 across all models.

Table 3.

Random effects models predicting odds of serious mental illness (K6: >18) (n = 1266).

	Model 1	Model 2	Model 3	Model 4	Model 5
Period (2020)	2.33 (1.68, 3.22)	2.33 (1.68, 3.23)	2.32 (1.68, 3.22)	2.37 (1.71, 3.30)	2.73 (1.69, 2.23)
Gender (male)		1.22 (0.77, 1.94)	1.23 (0.77, 1.97)	1.07 (0.66, 1.72)	1.46 (0.84, 2.53)
Age (29–64)		0.98 (0.95, 1.01)	0.98 (0.94, 1.01)	0.97 (0.94, 1.01)	0.99 (0.94, 1.03)
Education (Year 12)		Reference	Reference	Reference	Reference
Tertiary		0.46 (0.22, 0.95)	0.48 (0.23, 0.99)	0.43 (0.21, 0.89)	0.41 (0.18, 0.94)
Postgraduate		0.77 (0.35, 1.70)	0.78 (0.36, 1.72)	0.69 (0.32, 1.49)	0.50 (0.20, 1.23)
Victoria (yes)		1.49 (1.00, 2.21)	1.45 (0.97, 2.16)	1.63 (1.09, 2.43)	1.68 (1.06, 2.66)
Financial coping (no)		3.08 (1.57, 6.02)	3.13 (1.58, 6.19)	2.39 (1.21, 4.69)	2.01 (0.86, 4.71)
Single (yes)			1.06 (0.62, 1.80)	1.05 (0.62, 1.77)	–
Child aged ≤6 (yes)			1.14 (0.68, 1.91)	1.20 (0.72, 2.00)	1.19 (0.67, 2.08)
Number children (1)			Reference	Reference	Reference
2			0.60 (0.36, 0.99)	0.61 (0.37, 1.01)	0.72 (0.40, 1.31)
3+			0.65 (0.36, 1.17)	0.69 (0.39, 1.23)	0.80 (0.41, 1.57)
Work hrs (30–45)				Reference	Reference
0–14				0.81 (0.38, 1.70)	0.79 (0.33, 1.90)
15–29				0.67 (0.39, 1.12)	0.75 (0.41, 1.38)
46+				1.53 (0.87, 2.71)	1.33 (0.70, 2.51)
Own business (yes)				1.66 (0.89, 3.11)	1.61 (0.81, 3.23)
Job insecure (yes)				2.98 (1.96, 4.52)	3.21 (1.99, 5.20)
WFH COVID (yes)				0.80 (0.50, 1.28)	1.00 (0.57, 1.75)
High domestic (yes)					0.87 (0.44, 1.71)
High child care (yes)					3.02 (1.50, 6.11)
Partner employed (no)					Reference
Yes, but no hrs					0.77 (0.36, 1.65)
Yes, working hrs					0.44 (0.11, 1.81)

Notes. All variables taken from Time 2 (2020) except for gender and education (from Time 1). n = 1266 models 1–4; n = 1033 model 5. Model 5 includes only those partnered. WFH = work from home.

A further fixed-effects logistic regression model examined change in time period in association with change in serious mental illness, again controlling for time-varying partner status, work hours and job insecurity (as well as any between-person differences). The results showed a similar increase in the likelihood of serious mental illness by 244 % (OR: 2.44, 95 % CI: 1.72, 3.45). Although, the sample for this model was small (n = 170) as only participants who change on the serious mental illness variable are included (because fixed effects logistic regression models will only include respondents who report change in the outcome measure over time). The fixed effects model was re-run with a model adopting linear rather than logistic regression to avoid this constraint in the analysis sample. These results showed a 5.3 percentage point increase in parents reporting with serious mental illness from pre to during the pandemic (i.e., from 8.0 % to 13.3 %; B: 0.05, 95 % CI: 0.03, 0.07) (n = 1266). Again, models were repeated including an interaction effect for gender, with no evidence of a significant interaction found.

3.3. Lockdown effect in Victoria

To explore whether there was any additional impact for parents experiencing the extended lockdown in Victoria, the interaction between time period and location (Victoria vs. elsewhere) was examined. A random effects model with all covariates (i.e. repeating Model 4, Table 2) showed there was a significant interaction effect, such that the increase in psychological distress from pre- to during COVID was greatest for those in Victoria (B interaction estimate: 0.71, 95 % CI: 0.19, 1.23). While the pandemic was associated with an increase in levels of psychological distress of 1.54 points in the rest of Australia, the increase in Victoria was 2.25 points on the K6 scale. This effect was similar when a fixed effects model was conducted (B interaction estimate: 0.74, 95 % CI: 0.22, 1.26).

Random and mixed effects models showed that the increase in serious mental illness was 4.7 percentage points greater for those in Victoria than elsewhere in Australia (B: 0.05, 95 % CI:0.01, 0.09, p = .045). The increase in serious mental illness was 3.5 percentage points in the rest of Australia (from 7.6 to 11.1) compared to an increase of 8.2 percentage points in Victoria (from 8.0 to 16.2).

4. Discussion

The current study is one of very few in Australia to prospectively assess changes in psychological distress from prior to during the pandemic, for a cohort of employed parents. The findings indicate robustly and consistently, across several analyses and model types, that there was an increase in both levels of distress and likelihood of serious mental illness in the second half of 2020, when uncertainties around COVID and related public health orders were impacting most Australians. The estimates show a 5.3 percentage point increase overall in the population reporting serious mental illness (from 8.0 to 13.3), and that this increase was greater for those parents in Victoria compared to parents in other states.

The findings confirm and extend previous research showing that mental health has deteriorated in Australia during the pandemic. While the focus of this study was on employed parents, the results from other studies with various population groups including adolescents (Li et al., 2021; Magson et al., 2021; Munasinghe et al., 2020), medical students (Lyons et al., 2020), and health workers (Digby et al., 2021; Dobson et al., 2021; Holton et al., 2020), similarly suggest that it has been a difficult and stressful time for Australians, resulting in poorer mental health across the whole and key populations. Australian research using population-based representative repeated (cross-sectional) samples and similarly assessing change on the K6 psychological distress scale also found that scores increased during the pandemic (from 10.2 pre-COVID in 2019 to 12.6 in the first year of the pandemic) (Botha et al., 2022). The same study showed an 11.4 percentage point increase in those meeting the threshold for serious mental illness (from 6.3 % to 17.7 %). These figures are somewhat higher than those reported in the current study, which may reflect the socio-economic advantage in the current sample of employed parents or the repeated cross-sectional nature of Botha et al.'s study (with different samples pre and during COVID). Other representative longitudinal Australian research reported a significant but more modest increase in the proportion of people with serious mental illness (categorised using the K6 threshold) from 8.4 % (pre-COVID 2017) to 10.6 % (April 2020) (Biddle et al., 2022).

There have been mitigation strategies in place to reduce the physical health and economic impacts of COVID, such as lockdowns, masks, personal protective equipment (PPE), social distancing, and COVID support and emergency payments. Other initiatives have focused on preventing mental health decline (e.g. Medicare subsidised telehealth consultations for mental health professionals, and increased funding for mental health crisis lines and services). However, there has been less support available specifically for parents and those with caring responsibilities. Indeed, some of the mitigation strategies put in place to protect the economy and reduce physical morbidity and mortality have been at odds with the social and psychological supports parents need to maintain good mental health and wellbeing. For example, while closing/limiting schools and childcare centres reduced the spread of COVID infections, this left parents with tremendous increases in childcare and home-schooling responsibilities. Evidence suggests these strains were felt by parents of all genders, although some research indicates this was particularly so for women, and women with young children (Craig and Churchill, 2021; Butterworth et al., 2022), while another study suggests the mental health impacts were greater for men (Botha et al., 2022).

The current study found no evidence that the mental health effects of COVID were magnified for mothers in comparison to fathers, suggesting it may be parenting during the pandemic (i.e., an active parenting role, rather than gender roles) that created additional psychological strain. This links back to the original pre-COVID study design which invited ‘parents’ as participants, likely fostering a sample of mothers and fathers who are active and engaged in their parenting. We acknowledge that the current sample is socio-economically advantaged compared to the general population, with all participants employed and an over-representation of tertiary educated parents. However, while the findings may not be generalisable to all parents, the sample description matches hundreds of thousands of working Australian parents during COVID, and our findings reflect that relatively socio-economically resourced parents, who likely have an active and engaged parenting role, experienced declines in mental health during the first year of the pandemic.

The work-family-health literature has for many years consistently shown that work-family strains and conflicts (i.e. difficulties managing work and family demands) are associated with poorer mental health both for parents and children (Borgmann et al., 2019; Cooklin et al., 2016; Leach et al., 2021). Similarly, we know from prior research that while job flexibility (such as working from home) is predominantly good for working parents' mental health (Chandola et al., 2019) involuntary flexible work (variable schedules, involuntary working at home) is linked overall to poorer mental health, burnout and stress (Kaduk et al., 2019; Oakman et al., 2020) as is an absence of routines or boundaries between the work and family domains (Mandeville et al., 2022). Parents' usual work and family roles and boundaries were greatly disrupted during the COVID period, and the usual formal and informal supports available to parents to maintain mental health and wellbeing, including outsourcing some of their unpaid workload, were removed or restricted. Further, evidence suggests that for many families, inter-parental conflict and family violence escalated within households under lockdowns (Carrington et al., 2021), so it is plausible that these factors all contributed to the worsening of mental health we report here.

4.1. Limitations

There are several limitations to the current study that need to be acknowledged. First, it should be noted that the sample is a socio-economically advantaged group. The sample were employed at both time-points in the study and there was a high number of parents in the analytic sample with tertiary education. Those without tertiary education and/or who are unemployed are more likely to experience mental health problems and to bear the brunt of poorer social and economic circumstances during the pandemic. It is possible that this study underestimates the impact of the pandemic on parents' mental health compared to Australian parents' experiences more broadly. However, on the other hand, the studied cohort were those most likely to be combining family responsibilities with working from home in the pandemic, suggesting they faced important, unique risks.

Related to sample representation, the initial baseline study used a convenience sample that was recruited online (resulting in a baseline/pre-COVID sample that was more socio-economically advantaged, but that also had slightly poorer mental health, than parents in the general population) (Bennetts et al., 2019). While this has implications for the generalisability of the study findings (i.e. online recruitment typically results in over-educated samples (Bennetts et al., 2019)), it is worth noting that most of the studies conducted during the early period of the pandemic similarly used online recruitment (Craig and Churchill, 2021; Fisher et al., 2020; Westrupp et al., 2021a). Importantly the current study, regardless of initial recruitment method, was able to follow the same cohort over time and reduce bias introduced by sample differences. While the drop-out between the first and second time-point was considerable, this is not unexpected as the sample had not been contacted since their initial survey four years earlier. As noted earlier there were some demographic differences between the sample that dropped out and those that continued for two time-periods, but mental health at baseline did not differ.

A final point to acknowledge is that in the fixed effects models other factors or life events that occurred uniformly across the cohort (i.e. the changed all in one direction) were not able to be modelled/adjusted as these changes could not be disentangled from time effects. For example, ageing four years and increases in the number of children within families may be associated with either declines or improvements in mental health over time, but these variables could not be included as they were too highly correlated with (and almost synonymous with) time in these analyses. They were however included in the random effects models (which include both between and within-person differences) and these showed effects in the opposite direction to the COVID effect – i.e. that being older and having more children were associated with lower levels of distress.

4.2. Conclusions

This study found that psychological distress and likelihood of serious mental illness increased for employed parents during the pandemic, compared to four years earlier. This study is one of few in Australia to examine change in mental health within the same individuals from pre- to during the pandemic. Across multiple models and analyses we show consistently that mental health deteriorated for parents and that this is attributable to the pandemic effect. This was the case when we controlled for pre-pandemic differences in sample patterning and individual risk for poor mental health (e.g. socio-economic disadvantage, insecure work, family structure); as well as changes that occurred during or because of the pandemic (e.g. involuntary working from home, increased parenting demands, relationship breakdown). Together, our findings show a robust, independent pandemic effect whereby parents' mental health worsened consistently, compared to four years prior.

The following is the supplementary data related to this article.

Supplementary Table 1

Exploration of attrition effects, sample differences in initial baseline mental health, and sample differences in initial state location.

Funding

The “Families at Work” study was funded by La Trobe University, Transforming Human Societies Research Focus Area and was conducted by AC, SB, SC, SH, JL (La Trobe University), and LL (Australian National University). Authors SB, AC, SC, SH, JL, and JN were additionally supported by the Roberta Holmes Transition to Contemporary Parenthood Program at La Trobe University. AC is additionally funded by an ARC Future Fellowship (FT200100209).

CRediT authorship contribution statement

L.L, A.C, S.H, S.B, J.L, S.C designed the study and managed the data collection. L.L, J.L, S.H and P.B undertook the data preparation and analyses. L.L wrote the initial draft of the article and all authors participated in critical revisions and approved the final draft.

Conflict of interest

The authors declare that there are no conflicts of interest.