Selection and inclusion of systematic reviews
The systematic search provided 46,284 record entries, reduced to 31,559 after deduplication. 939 records were selected for full text retrieval by screening titles and abstracts. Of these, nine were finally included after comparing full texts against inclusion and exclusion criteria. Figure 1 reports the details of the study selection process.
PRISMA flowchart of studies selection
Characteristics of included systematic reviews
Of the nine included reviews, five performed a meta-analysis on at least one of the outcomes of interest [18,19,20,21,22], while the remaining four were of a narrative nature [23,24,25,26]. Table 1 summarises the characteristics of the included reviews.
Systematic reviews with meta-analyses (MAs) generally had a search covering up to the first quarter of 2021. They included primary studies with mainly a cohort design, with the notable exceptions of the review by Liu and colleagues which considered case-series too and the review by Fond and colleagues which considered population-based cohort studies. MAs included a mean of 45 studies and a median of 21, the review by Liu and colleagues being an outlier with 149 included studies. Most of the MAs considered people with any mental disorder, possibly providing subgroups for specific disorders groups, while the review by Ceban and colleagues only focused on people with a mood disorder. Included studies covered a wide range of countries but with a general lack of representation from low- and middle-income countries. Infection by SARS-CoV-2 was defined by laboratory testing, ICD-10, electronic health records, or clinical judgment in the review by Ceban and colleagues and by laboratory results and diagnosis in conjunction with clinical presentation in the review by Liu and colleagues. Severe illness was defined as ICU admission, mechanical ventilatory support, oxygen therapy, extracorporeal membrane oxygenation, acute respiratory distress syndrome, and/or cardiopulmonary resuscitation by Ceban and colleagues and as hospitalization, ICU admission, or requirement for other special treatment (including oxygen therapy) by Liu and colleagues. For the review by Vai and colleagues, which does not have a “severe illness” outcome, we considered the outcome ‘hospitalisation’. For the review by Toubasi and colleagues, we considered the pooled mortality and severe illness outcome, as a separate estimate for mortality only was not available. Notably, no review was available to inform on the long-term physical symptoms after SARS-CoV2 infection. Four reviews reported estimates based on pooling adjusted ORs only [18, 19, 21, 22]. The review by Vai and colleagues reported a fully adjusted model only when considering people with any mental disorder, for the diagnostic groups we have then considered the ‘partially adjusted model’, where review authors considered aORs pooled together with crude ORs when an adjusted figure was not available from primary studies.
The four narrative reviews varied in their specific design, with two scoping reviews [25, 26], one rapid review [23], and one systematic review without meta-analysis [24]. They considered a wide range of study designs with the review by Lemieux and colleagues considering opinion pieces and other reviews; as for population of interest, they considered people with bipolar disorder [25], people with schizophrenia [24], people with mental illness in secure settings [23], and generally people with mental disorders [26].
Quality of included reviews
The AMSTAR 2 rated level of quality for all the MAs was “low”, with the exception of the review by Liu and colleagues with “high” and the review by Fond and colleagues with “critically low”. The review by Liu and colleagues did not have any weaknesses in critical items, while all other MAs did not report the list of excluded studies with reasons for exclusion; the review by Fond and colleagues also did not account for the impact of risk of bias in primary studies on the results. The level of quality for the narrative reviews was “critically low” with the exception of the review by Fornaro and colleagues (“low”), mainly because of a lack of risk of bias assessment and protocol registration. See the supplementary material for detailed AMSTAR 2 evaluation of the included reviews.
Risk of SARS-CoV-2 infection
Two MAs informed on the association between Sars-CoV-2 infection and having a pre-existing mental disorder compared to not having a pre-existing mental disorder (Fig. 2) [18, 19].
Risk of contracting SARS-CoV-2 infection. CI: confidence interval; K: number of included studies; OR: Odds ratio; n: total number of included participants; NR: not reported; *: partially adjusted model
For people with any mental disorder, Liu and colleagues found a statistically significant positive association (OR: 1.71, 95% CI 1.09–2.69). For people with anxiety disorders, Liu 2021 and colleagues in a partially adjusted model found a statistically significant positive association although this effect size relies on two studies only (OR 1.63, 95%CI 1.44–1.85). For people with mood disorders, Liu 2021 and colleagues in a partially adjusted model found a statistically significant positive association (OR: 2.02, 95%CI 1.08–3.76), while Ceban and colleagues did not find a statistically significant association (OR: 1.50, 95%CI 0.75–2.99). For people with schizophrenia spectrum disorders, Liu and colleagues, in a partially adjusted model, did not find a statistically significant association (OR: 1.72, 95%CI 0.62–4.77). The level of statistical heterogeneity was been generally very high, with most I2 statistics over 95%, with the exception of the estimate for people with mood disorders by Liu and colleagues (0%). I2 was not reported in Ceban et al., 2021.
Risk of severe illness
Three MAs informed on the association between a severe course of COVID-19 and having a pre-existing mental disorder compared to not having a pre-existing mental disorder (Fig. 3) [18,19,20].
Risk of severe COVID-19 illness. CI: confidence interval; K: number of included studies; OR: Odds ratio; n: total number of included participants; NR: not reported; *: partially adjusted model
For people with any mental disorder, both the review by Liu and colleagues and Vai and colleagues found a statistically significant positive association (OR: 1.32, 95%CI 1.19–1.46 and OR: 1.77, 95%CI 1.29–2.42, respectively). No study informed on people with anxiety disorders. For people with a mood disorder, Ceban and colleagues found no association (OR: 0.99, 95%CI: 0.80–1.24), Liu and colleagues in a partially adjusted model found a statistically significant positive association (OR: 1.34, 95%CI 1.08–1.67) while Vai and colleagues did not find a statistically significant association (OR 1.27, 95%CI 0.64–2.50). For people with a schizophrenia spectrum disorder, both the reviews by Liu and colleagues and Vai and colleagues did not find a statistically significant association (OR: 1.22, 95%CI 0.70–2.13 and OR: 1.38, 95%CI 0.61–2.94, respectively). The level of statistical heterogeneity was moderate to very high, with an I2 statistic between 65 and 100%, but low for the estimate for people with mood disorders by Vai and colleagues (23%). I2 was not reported in Ceban et al., 2021.
COVID-19 related mortality
Four MAs informed on the association between mortality and having a pre-existing mental disorder compared to not having a pre-existing mental disorder (Fig. 4) [19,20,21,22].
COVID-19 related mortality risk. CI: confidence interval; K: number of included studies; OR: Odds ratio; n: total number of included participants; NR: not reported; *: partially adjusted model; **: considers mortality and severe illness together
For people with any mental disorder all four MAs found a statistically significant positive association, ranging from an OR of 1.38 (95%CI: 1.15–1.65) for the review by Fond and colleagues to 1.52 (95%CI: 1.20–1.93) for the review by Toubasi and colleagues (which however considered in this outcome severe illness cases as well). For people with anxiety disorders both the reviews by Liu and colleagues, in a partially adjusted model, and by Vai and colleagues did not find a statistically significant association (OR: 1.16, 95%CI 0.75–1.79 and OR: 1.01, 95%CI 0.77–1.32, respectively). For people with mood disorders, all three informing MAs found a statistically significant positive association with ORs ranging from 1.36 (95%CI: 1.15–1.79, Liu and colleagues, partially adjusted model) to 1.57 (95%CI: 1.26–1.95 Vai and colleagues). For people with schizophrenia spectrum disorders both the reviews by Liu and colleagues, in a partially adjusted model, and by Vai and colleagues found a positive association (OR 2.28, 95%CI 1.40–3.73, and OR 1.68, 95%CI 1.29–2.18, respectively). The level of statistical heterogeneity has been generally moderate to considerable, with an I2 statistic between 60% and 81.4%, but for the review by Vai and colleagues for the estimate for people with any mental disorder (39%), and low for the estimate for people with anxiety disorders (0%) and mood disorders (22%). The I2 has not been reported in Ceban et al., 2021.
Narrative reviews
The narrative reviews corroborated meta-analytic findings by indicating that patients with serious pre-COVID-19 mental disorders show adverse health outcomes related to COVID-19 infection in terms of higher severity and mortality. Fornaro and colleagues [25] performed a scoping review on clinical and public health themes for people with bipolar disorder. They identified four major themes from the 14 included papers, among which was the impact of having bipolar disorder on the risk of contracting the SARS-CoV-2 infection. For this theme, one study reported an increased risk of infection contraction for people with bipolar disorder [27]. This study was considered by the MAs previously reported. Karaoulanis and Christodoulou performed a systematic review without meta-analysis on infection rates and mortality in people with schizophrenia spectrum disorders. The included studies suggest an increased infection and mortality risk; these studies have been included in the MAs previously reported. Lemieux and colleagues performed a rapid review on the management of COVID-19 for people with mental illness in secure settings. They considered a wide range of publications including opinion pieces and other reviews and report greater morbidity and mortality. Murphy and colleagues conducted a scoping review on the impact of COVID-19 and related restrictions on people with pre-existing mental disorders. They reported an increased infection risk in this population.
link