Mental health impacts of COVID-19: A retrospective analysis of dynamic modelling projections for Australia

Abstract

Background: In early 2020, we developed a dynamic model to support policy responses aimed at mitigating the adverse mental health effects of the COVID-19 pandemic in Australia. As the pandemic has progressed, it has become clear that our initial model forecasts overestimated the impacts of infection control measures (lockdowns, physical distancing, etc.) on suicide, intentional self-harm hospitalisation, and mental health-related emergency department (ED) presentation rates.

Methods: Potential explanations for the divergence of our model predictions from observed outcomes were assessed by comparing simulation results for a set of progressively more refined models with data on the prevalence of moderate to very high psychological distress and numbers of suicides, intentional self-harm hospitalisations, and mental health-related ED presentations published after our modelling was released in July 2020.

Results: Allowing per capita rates of spontaneous recovery and intentional self-harm to differ between people experiencing moderate to very high psychological distress prior to the pandemic and those developing comparable levels of psychological distress only as a consequence of infection control measures substantially improves the fit of our model to empirical estimates of the prevalence of psychological distress and leads to significantly lower predicted effects of COVID-19 on suicide, intentional self-harm hospitalisation, and mental health-related ED presentation rates.

Conclusion: Accommodating the influence of prior mental health on the psychological effects of population-wide social and economic disruption is likely to be critical for accurately forecasting the mental health impacts of future public health crises as they inevitably arise.

Keywords: Pandemic; Psychological distress; Social disconnection; Suicide; System dynamics; Unemployment.

Fig. 1

Simulated unemployment rates for Model-empirical and Model-RBA (see Table 1 for details). Monthly Labour Force Survey data (Data) published by the Australian Bureau of Statistics [19] were used to fit Model-empirical; Model-RBA was fitted to the Reserve Bank of Australia's May 2020 unemployment forecast (RBA forecast) [26]. Simulated unemployment rates for Model-ABS are nearly identical to those for Model-empirical. The dotted vertical line in each plot indicates the start of the COVID-19 pandemic (30 January 2020).

Fig. 2

Simulated prevalence of moderate to very high psychological distress (15 years and above) and numbers of suicides, intentional self-harm hospitalisations, and mental health-related emergency department (ED) presentations per year for the three models described in Table 1. Pointwise 50% and 95% intervals derived from the sensitivity analyses (see Methods) are indicated with dark and light shading, respectively. Data on the prevalence of moderate to very high psychological distress are from the National Health Survey (pre-pandemic) [27] and the Household Impacts of COVID-19 Survey [20]; data for the remaining outcomes were obtained from online reports published by the Australian Institute of Health and Welfare [28,43]. The vertical dotted line in each plot indicates the start of the COVID-19 pandemic (30 January 2020). Results for adolescents and young adults (15−24-year-olds) are in Supplementary appendix 4.

Fig. 3

Model-based and empirical estimates of suicide and intentional self-harm hospitalisation rates (all ages and 15−24-year-olds) for the pre-COVID-19 period. Model-based estimates were obtained by fitting the baseline model to data for the period 2011–2015 and predicting suicide and intentional self-harm hospitalisation rates for 2016–2019. Empirical estimates (Data) were derived from the Australian Institute of Health and Welfare's Suicide and Self-harm Monitoring web report [28]. The vertical dotted line in each plot indicates the start of the prediction period (i.e., 2015).

Fig. 4

Reductions in total (cumulative) numbers of suicides, intentional self-harm hospitalisations, and mental health-related emergency department (ED) presentations for selected mental health promotion and services interventions calculated using each of the models in Table 1. Descriptions of each intervention are provided in Supplementary appendix 3. Mean numbers of suicides, hospitalisations, and ED presentations prevented (reported in the third column) and mean percentage reductions and 95% intervals (fourth column and plot) were derived from the distributions of outcomes calculated in the sensitivity analyses (see Methods section for details). Note that the 95% intervals provide a measure of the impact of uncertainty in the assumed intervention effects but should not be interpreted as confidence intervals.