Worsening psychological wellbeing of Australian hospital clinical staff during three waves of the coronavirus (COVID-19) pandemic

Design, setting and participants

The study compared data from a brief, self-administered, anonymous, online cross-sectional survey completed by hospital clinical staff (nurses, midwives, doctors and allied health staff) employed at the study health service at three different timepoints during the COVID-19 pandemic (T1: 15 May–10 June 2020, first recognised ‘wave’ of the pandemic in Australia; T2: 5 October–2 December 2020, second wave; and T3: 30 November 2021–24 January 2022, third wave).

The methods and some findings from surveys 1 and 2 have been previously reported.^17,19,20 Findings from survey 3 or an examination of differences between staff’s wellbeing across the three survey timepoints have not been previously reported.

The study health service, Western Health, is a large public health service located in metropolitan Melbourne, Australia. Throughout the data collection, the health service provided inpatient care for COVID-19 patients. At T1 the number of clinical staff employed at the health service was approximately 4530.

Procedure

The surveys were available in Qualtrics (Qualtrics, Provo, UT, USA), an online survey platform. Convenience sampling was used at each timepoint. An invitation was sent to the group email address for each discipline (nursing/midwifery, medical and allied health staff) at each timepoint, followed by a reminder email 2–3 weeks later. The first survey was open for 4 weeks and the second and third for 8 weeks. The email invitation included the link to the survey and a plain language statement; completion of the survey was taken as informed consent.

The self-report questionnaire was informed by published studies on the impact of similar infectious diseases (SARS; Middle East Respiratory Syndrome Coronavirus, MERS-CoV) on health service staff’s psychosocial wellbeing,^19–24 and the clinical experience of the research team.

The survey took approximately 15 min to complete and included mostly fixed-response questions and assessed respondents’ sociodemographic characteristics, health status and psychological distress (Table 1).

The survey was the same at each timepoint (except for the assessment of resilience in T2 and T3, and burnout only in T3) and for each clinical group except for discipline appropriate sociodemographic questions.

The surveys were anonymous but to enable longitudinal matching, respondents to each survey were asked to create a unique identification code using a specific combination of letters and numbers from their personal details (e.g. name and date of birth). The same instructions for generating this code were included in all three surveys as well as examples.

Data management and analysis

Data were analysed using IBM SPSS Statistics version 26.

Descriptive statistics were used to summarise the data.

In order to compare responses from all three surveys, data were first matched using the unique identification codes generated by the respondents. Only 116 respondents completed more than one survey and responded to requests to generate the code. Therefore, the data from the three surveys were treated as independent samples.

The distributions of respondents’ sociodemographic and health characteristics in T1, T2 and T3 were compared (using chi-squared tests for categorical variables and Mann–Whitney U-tests for continuous variables) to identify significant differences.

DASS-21 subscale scores and proportion scoring in clinical ranges were calculated as outlined by the instrument’s authors²⁵ in order to determine the clinical staff who have experienced ‘normal’, ‘mild’, ‘moderate’, ‘severe’ or ‘extremely severe’ depression, anxiety or stress. Cohen’s d is reported, along with qualitative descriptors: small (0.20), medium (0.5), large (0.8) and very large (1.3).²⁶

Differences in respondents’ resilience, general health status and DASS-21 scores across the three survey timepoints were analysed using Kruskal–Wallis tests.

These categories were reduced to a binary variable (‘normal/mild’ vs ‘moderate/severe/extremely severe) for comparison with sociodemographic and health variables. The proportion scoring in these categories across the three waves was compared using chi-squared tests.

Associations were investigated between DASS-21 subscale scores and demographic variables, employment characteristics, COVID-19 contact status, discipline group and self-rated general health status. DASS-21 subscale scores were all significantly non-normally distributed; therefore, Mann–Whitney U-tests, Kruskal Wallis tests or Spearman’s ρ-coefficients were used as appropriate. For post hoc pairwise comparisons, significance values were adjusted by the Bonferroni correction for multiple tests.

Variables significantly associated with scores on any of the subscale scores (P < 0.05) were included in general linear models with DASS-21 Depression, Anxiety and Stress subscale scores as outcome variables. Preliminary analyses were conducted to ensure that assumptions of multicollinearity were not violated.

Ethics

Ethics approval was granted by the Western Health Low Risk Ethics Panel, HREC/20/WH/62913, 5 May 2020.

Sample and response

A total of 668 (14.7%) clinical staff employed at the study health service completed the first survey (391 nurses and midwives, 139 allied health staff and 138 doctors), 358 (7.9%) completed the second survey (184 nurses and midwives, 74 allied health staff and 100 doctors), and 444 (9.8%) completed the third survey (285 nurses and midwives, 100 allied health staff and 59 doctors).

The respondents’ sociodemographic characteristics were similar across the three survey timepoints and by discipline group. Most respondents were nurses or midwives and there was a statistically significant decrease in the proportion of nurses and midwives who were born in Australia (P < 0.001) and increase in those that had school-aged children (P = 0.049) across the survey timepoints (Table 2).

COVID-19 contact status

There was a statistically significant increase in the proportion of respondents, both overall and by professional discipline group, who had direct contact with a person with a COVID-19 diagnosis or had been diagnosed with COVID-19 from the first survey timepoint to the third (P < 0.001) (Table 2).

Self-reported general health status

Respondents’ general health status significantly declined across the three survey timepoints (P < 0.001) and by discipline group (nurses/midwives P < 0.001; allied health staff and doctors P = 0.005) (Table 2).

Psychological wellbeing

The sample’s mean resilience score significantly decreased from the second survey timepoint to the third, and almost two-thirds of the survey three respondents reported one or more symptoms of burnout (Table 2).

The total sample’s mean score on each of the DASS-21 subscales significantly increased from the first survey timepoint to the third (all P < 0.001) (Table 3). For nurses and midwives all DASS-21 subscale mean scores were significantly higher at the third timepoint than the first (depression P < 0.001, anxiety P < 0.001, stress P = 0.018); for allied health staff there was a significant increase on the Anxiety and Stress subscales (depression P = 0.348, anxiety P = 0.001, stress P = 0.016); and for doctors on the Depression and Anxiety subscales (depression P = 0.007, anxiety P = 0.004, stress P = 0.198) (Table 3).

For the total sample, the proportion of respondents who reported moderate to extremely severe symptoms of depression, anxiety and stress significantly increased from the first survey timepoint to the third (all P < 0.001). There was also a significant increase in the proportion of nurses and midwives and allied health staff who reported symptoms of depression, anxiety or stress over the three survey timepoints. For doctors there was a significant increase only in anxiety (P = 0.047) (Table 4).

In the general liner models, the main effect for survey timepoint was significant for all three DASS-21 subscales. Compared with the first survey timepoint, the second and third timepoints were associated with significantly higher depression (P < 0.001 for both T2 and T3), anxiety (P < 0.001 for both T2 and T3) and stress (T2: P < 0.001, T3: P = 0.003) mean scores. Older age and better self-rated general health status were significantly associated with lower scores on all subscales. Living with school-aged children was significantly associated with lower depression (P = 0.002) and anxiety (P = 0.033) scores. Compared with doctors, nurses and midwives had significantly higher depression (P = 0.024) and anxiety (P < 0.001) scores (Table 5).

Strengths and limitations

Large and diverse samples of hospital clinical staff including nurses, midwives, doctors and allied health staff were surveyed at three timepoints during the COVID-19 pandemic, and validated instruments were used to assess psychological wellbeing. This provided evidence about the long-term impact of the pandemic on the psychological wellbeing of hospital clinical staff.

Although the response rate for each survey was relatively low, the rates are similar to those of other studies that used unsolicited surveys during an infectious disease outbreak.²² Due to infection control protocols at the health service, staff could only be invited to participate via email and the surveys had to be completed online. It was not possible to accurately determine the number of staff who received the link to the surveys, thus our conservative estimation of the response rate at each timepoint was based on the total number of clinical staff in the health service.

The study was conducted at a large metropolitan health service in Australia; therefore, the results may not be generalisable to other health services or settings.

Due to the small number of respondents who generated a unique identification code and completed more than one survey, responses were not matched across all three survey timepoints. Accordingly, the sample is different at each survey timepoint, but respondents’ sociodemographic characteristics were similar across all three timepoints.

It is possible that other factors, which were not assessed in this study, contributed to the worsening psychological wellbeing of hospital clinical staff as the pandemic continued. These may include high workload, staff turnover and potential economic impacts of COVID-19. Nevertheless, the study has identified several variables significantly associated with poorer wellbeing.

Implications for health policy and practice

The findings of this study indicate that there was a significant and persistent negative impact on the psychological wellbeing and general health status of hospital clinical staff across waves of the COVID-19 pandemic. Health services should be cognisant of the potential for hospital clinical staff to experience deteriorating mental health during and between pandemic waves,¹³ and provide ongoing and continued wellbeing support.