To our knowledge, this is the first study examining psychiatric emergency department (pED) visits and admissions in Berlin during the first wave of the Covid-19 pandemic.
We found a no significant change in pED presentations in absolute numbers (− 9.1%, Table 1) and in weekly rates (− 9.1%, p = 0.064, Table 1 and Fig. 1) during the Covid-19 period compared to the pre-Covid-19 period. Most studies on pED attendances during the first wave of Covid-19 report decreases in pED presentations ranging from 15 to 40% [9, 10, 15, 17, 19, 23, 24, 26, 28, 47], mostly from Europe, America, and the Pacific region. However, some studies report even higher decreases [11, 14, 18], others as Simpson et al. report a range of 4 to 9% decreases in pED attendances in different regions of the US, partly not reaching statistical significance [25]. This underlines the impact of regional factors on pED presentations. Potential reasons for a decline in presentations will be discussed in the following paragraphs for each statistically significant difference that we found in the respective subgroups.
We saw no significant differences in age, gender and homeless status between the Covid-19 and the pre-Covid-19 period (Table 1), which is in line with most studies [21, 24, 28, 48].
Covid-19 high-risk group
When comparing the two time periods regarding to risk factors for a severe Covid-19 infection that were known in the beginning of the pandemic, no difference in the number of pED attendances was found. One could have hypothesized that there were less self-referred presentations in the Covid-19 high-risk group during the first wave of Covid-19 for reasons of self-protection, which was not observed in our sample.
Provenance
The rate of presentations from our catchment area increased during the Covid-19 period compared to the pre-Covid-19 period (Table 1). This, in combination with an important decrease of presentations from Berlin outside our catchment area (63.9 vs. 56.9%, Table 1), suggests a concentration to more local help-seeking behaviour and has not been shown yet in other studies focusing on pEDs. It may be due to stay-at-home advice of the government and a reduced mobility of patients [26, 49] (cf. last paragraph of Introduction). This view is supported by the observed decrease in presentations from other parts in Germany and outside Germany (Table 1). However, we also found that more cases were redirected from other hospitals in the Covid-19 period vs. the pre-Covid-19 period (Table 1), redirection being a strong predictor of Covid-19 period in the logistic regression model (Table 4). Redirection means that a patient requiring hospital admission attends a pED outside his responsible district and is subsequently transferred to the pED of his responsible clinic. Supposedly, the concentration on more local psychiatric help is also due to this mechanism. From clinical experience, we know that redirection takes place more often when admission capacities are scarce (cf paragraph on “Discharge/hospital admissions” in the Discussion).
Mode of attendance
The decrease in self-referring attendances in the Covid-19 period in our sample is in line with other studies, which discuss that fear of infection with Covid-19 limits voluntary pED attendances [11, 21]. Attendances accompanied by family/friends decreased in our sample in the Covid-19 period compared to the pre-Covid-19 period, too. It could be discussed that this is due to social distancing measures and the contact ban during the observed Covid-19 period. A main finding of this study is that the total number and proportion of patients brought in police custody increased during the observed Covid-19 period when compared to the same period in 2019 (Table 1). This is backed by the finding that police custody was a strong predictor for Covid-19 period in the logistic regression analysis (Table 4). This finding is even more remarkable as overall case numbers decreased during the Covid-19 period. Our findings differ with previous research concerning mode of attendance, because others do not report [13, 18, 19, 49] or show no significant differences in presentations with police custody between the Covid-19 period and the pre-Covid-19 period [11, 21]. In our sample, presentations in police custody are overall more frequent compared to these studies (20.4% in the Covid-19 period, 16.3% in the pre-Covid-19 period in our study compared to 7 or 10% of presentations in police custody in other studies [11, 21]). Another difference to the studies from Switzerland and Australia is that in their samples pED presentations with ambulance did increase during the Covid-19 pandemic (from 45.4% vs. 23.3% in Switzerland [21] and 46.0% vs. 43.6% in Australia [11]). In our sample, however, there were no differences in presentations by ambulance between the pre-Covid-19 and the Covid-19 period (percentages around 20%, Table 1). The subgroup of cases in police custody in the Covid-19 period in our sample did not show more aggressive behaviour towards others but was more likely to be suicidal (Table 2). This in mind, it could be discussed that part of the increase in police custody presentations in our sample would be in other countries more likely to be seen in an increase in ambulance presentations. Further studies are needed to investigate if there are Covid-independent differences in mode of attendance of suicidal patients between countries. Also, it may be possible that other local-bound factors such as distinctions in psychiatric care systems and/or diagnostic categories of the sample determine the differences between studies.
Diagnostic categories
Presentations with schizophrenia and other psychotic disorders (SPD) did not differ during the Covid-19 period compared to the pre-Covid-19 period (Table 1). When focusing on paranoid schizophrenia alone, we could see an absolute increase of presentations during the Covid-19 period (137 vs. 115, + 19.1%, p = 0.020, Table 1). Paranoid schizophrenia was a predictor of Covid-19 period in the logistic regression analysis, too (Table 4). An increase in rates of presentations with SPD during the Covid-19 period was shown in various studies of similar design [12, 18, 19, 23, 49,50,51]. One may speculate that patients with chronic psychotic disorders and high need of psychosocial facilities have suffered more than other diagnostic groups from the lockdown restrictions which included closing or limited accessibility of many psychosocial [41] and psychotherapeutic [52] facilities. However, there are also studies, which do not show an overall increase in presentation rates in patients with SPD [15, 21, 28, 30], stressing local differences. An absolute increase in presentations with paranoid schizophrenia has only been reported in few studies [12, 51]. Interestingly, these studies have longer observation periods (6 months or longer) than most other Covid-pED studies yet. Earlier pED studies covered mostly 4 to 8 weeks. Our study has a rather long observation period with 12 weeks. In line with the finding from Jagadheesan et al. [12] the presentation rates with schizophrenia (Fig. 3) increase mainly at the end of the Covid-19 period.
It could be discussed whether this is due to weeks of strain with restrictions and Covid angst. However, a study from Ireland did not find an increase in pED presentations with psychotic disorders in their 1-year observation period [53]. Their study comprises more non-lockdown than lockdown periods, suggesting that the increase in psychotic pED presentations predominantly occurs in lockdown periods.
We observed a trend to an absolute increase in numbers of presentations with paranoid schizophrenia in police custody in the Covid-19 period compared to the pre-Covid-19 period (Table 3). This trend is in contrary to both a clinical [54] by Winkler et al. and a questionnaire-based [55] study by Kølbæk et al. that showed that patients with schizophrenia subjectively deteriorated less than other diagnostic groups during the first lockdown. One possible explanation is, that Kølbæk et al. [55] focussed on a less severely ill sample than ours as their study was based on the subjective mental state of patients who did stay in touch with the mental care system and were willing to answer several questions on the telephone, which is from clinical experience not typical for a patient presenting to the pED. As the pED presentations in the current study were mainly not self-referred (Table 1), another explanation may be a discrepancy between the subjective mental state and the evaluation of others: Winkler et al. states that although patients with schizophrenia reported less emotional strain, they formed the biggest group with crisis admissions [54]. Another explanation could be that the increase of presentations with paranoid schizophrenia was driven by newly diagnosed patients in the Covid-19 period since these were not included in the above-mentioned studies. There is some evidence from Israel [56] and Italy [57] that there was a potentially stress-induced rise in new-onset psychosis in the beginning of the pandemic. If these partially developed into schizophrenia is unknown and should be addressed in future research.
All studies, so far, found that absolute numbers of presentations with depressive disorders decreased during the first lockdown [12,13,14,15, 18,19,20,21,22, 26, 28, 30, 31, 48, 50, 51]. In the current study the decrease was 29.4% and hence was comparable to earlier studies (68.3% [18], 43.5% [21], 28.3% [12]). The finding from the logistic regression analysis that DD was a negative predictor for Covid-19 period fits these observations. In our sample the decrease was seen equally in severe and milder forms of depression (Table 1). These findings give reason for concern, as depressive symptoms have been described to rise in adult populations [1,2,3], thus revealing a discrepancy between need for treatment options and attendance of mental health services.
We found a strong decline in pED presentations by − 40.2% in the group of patients with personality disorders (PD). At the same time, PD was a negative predictor of Covid-19 period (Table 4). Many earlier studies did not mention personality disorders or reported no change [22, 30, 53]. In three south European studies a comparable important decrease has been found (− 44% [31], − 34.4% [18], − 29.6% [20]). In a study from Germany, there was an increase in pED presentations with PD during the first wave of Covid-19 [28]. The current study is the first where the number of presentations of the group of PD decreased more importantly than in all other diagnostic groups (Table 1). Interestingly, presentations with PD were the only diagnostic group in the pre-Covid-19 period who came predominantly from outside the catchment area (54.5%, Table 2s in supplementary section), suggesting a higher baseline mobility in help-seeking compared to other diagnostic groups. The restrictions in mobility during lockdown (cf introduction section) seem to have had a relevant impact on this help-seeking mobility in PD: the amount of provenance from outside our catchment area is in the Covid-19 period not higher than in other diagnostic groups (34.3%, Table 2s). In other words, the decrease in PD presentations in our sample seems to be driven by a significant drop in presentations with PD from Berlin but outside our catchment area (− 68.3%); in presentations from our catchment area the drop was only − 8.3% (Table 2s). Winkler et al. found that outpatients with PD were suffering more intensely from lockdown and social distancing measures than patients with other diagnoses [54]. This in mind, we shouldn’t simply interpret the decrease in presentations as a decrease in mental health emergencies. Table 3 indicates that albeit the important decrease in numbers of pED presentations with PD, the absolute number of presentations with PD brought by police, nevertheless, increased (+ 4.5%), suggesting a high severity. Until date, there is only very few research focusing on the fate of the vulnerable group of PD during the Covid-19 pandemic. More research is necessary.
In the logistic regression analysis, Bipolar and manic disorders (BMD) were a predictor for Covid-19 period (Table 4). If assessed, most studies report decreases in pED presentations with bipolar disorders [21, 31, 48]. Gómez-Ramiro et al. reported no statistically significant changes in pED presentation with bipolar disorder [20]. In most studies, BMD was not reported separately but as part of “mood disorders”. More research is needed to solve the question if patients with BMD were particularly at risk of deterioration during the first Covid-19 lockdown.
Police custody
The relative and absolute increase in presentations in police custody is a main finding of this study (Table 1, Fig. 2). In our sample, this increase is based on a rise in female cases in the Covid-19 vs. the pre-Covid-19 period (67 vs. 38, + 76.3%, p = 0.008) with a female rate of 40.4% vs. 26.0%. In comparison, on the level of all presentations per observation period the female rate in the Covid-19 period did not differ compared to the pre-Covid-19 period (38.3% vs. 42.8%, p = 0.055, Table 1). There is evidence of gender-dependent lockdown effects, hinting to a more important deterioration in mental health of women than in men during the Covid-19 period [58, 59]. In our sample, self-referral significantly decreased (cf paragraph “Mode of attendance” in the Discussion), only in women (− 27.9% p < 0.001data in supplements, Table 1s). As women were more likely to be referred in police custody, we interpret this decrease in help-seeking in women as an expression of mental health deterioration.
Police custody may present a straight-forward characteristic suggesting acuity of cases. However, in our sample, the potential increase in pED presentations in police custody in patients with paranoid schizophrenia and with PD during the Covid-period compared to the control period does not necessarily represent an increase in disease severity. For instance, Individuals suffering from schizophrenia and from some PDs with high impulsivity traits may have had more trouble adjusting to social distancing and other restrictions, thus soliciting police interventions. The psychopathological signs of severity such as signs of delusion, aggressive behaviour towards others, suicidal thoughts, suicidal plans and suicidal attempts were, as expected, often present in the subgroup of presentations with police custody (Table 2). Differential between the Covid-19 and the pre-Covid-19 period in this subgroup were suicidal attempts prior to pED attendance, which were significantly more frequent in the Covid-19 period and reported aggressive behaviour towards others which was less often present in the Covid-19 period.
To summarize: the increase in pED presentations in police custody in the Covid-19 period compared to the pre-Covid-19 period in women, with suicidal attempts, with less aggressive behaviour and predominantly with schizophrenia or PD suggests but does not prove an increase in severity in this subgroup of psychiatric patients in the early Covid-19 era. Future studies should address the number of new-onset diseases of this subgroup. Moreover, this subgroup might deserve particular concern of mental health care services in the ongoing Covid-19 pandemic.
Hospital admission rates were, as expected, high after presentation in police custody: 62.0% during the Covid-19 period and 72.6% in the pre-Covid-19 period. I.e., the increase in presentations with police during the Covid-19 period was not and against expectation, translated into more hospital admissions, but- in the contrary- in less hospital admissions (Table 2). This raises the question: why could the need of more presumably necessary admissions not be met (cf paragraph Discharge/ hospital admission in the Discussion)? The fact that there was no increase in involuntary admissions in the Covid-19 period is reassuring in this matter (Table 2).
Suicidality
In our sample, suicidal thought and suicide attempt prior to pED presentation are predictors of Covid-19 period (Table 4). Elsewhere, we report that overall suicidal thought, plans and suicide attempts increased during the first wave of Covid-19 and not during the second wave of Covid-19 [60]. In the subgroup of presentation in police custody, we reported an increase in presentations after suicide attempt (Table 2). Further research is necessary to further determine the reasons of this increase. As an explaining factor, an increase in disease severity may be discussed. A bias may have been introduced by a potential increase in parasuicidal behaviour in order to increase chances of inpatient treatment when bed capacities were limited [60].
Discharge and hospital admission
We found no statistically significant change in hospital admissions (− 16.9%) in the Covid-19 period compared to the pre-Covid-19 period. Nonetheless, hospital admission was a strong negative predictor of Covid-19 period in the logistic regression analysis (Table 4). A German multi-center study found a decrease of − 30.3% [32]. As the multi-center study focuses on admissions and does not provide data on pED presentations, we cannot compare our admission rate to theirs. In our sample, total admission rate dropped from 48.9 to 44.6%. There were important differences per diagnostic group. Significantly, less presentations with substance use disorder (SUD) were eventually admitted (decrease in admission rate from 57.2 to 49.9%, Table 1). Also, a significant decrease for SPD and particularly the group of paranoid schizophrenia was observed (Table 1). In the pre-Covid-19 period 67.0% of presentations with paranoid schizophrenia led eventually to a hospital admission, in the Covid-19 period the admission rate was 49.6%.
To understand the important differences in admission rates, it is indispensable to consider the bed capacity in these days. The bed capacity in the Covid-19 period was significantly lower (Table 1) as one of the seven wards was turned into a Covid-19 ward and other infection-avoiding guidelines were implemented (cf Study design in chapter Methods). Figure 4 shows the admissions via the pED per month in the two observation periods (the admissions in the control period are set to 100%).
Figure 4 shows that hospital admissions via the pED in our hospital did not decrease for all patients groups but for patients with SUD and SPD (with paranoid schizophrenia). The fact that overall bed capacity was clearly reduced in the Covid-19 period and overall admissions via the pED did not decrease significantly, can be explained by a significant decrease in scheduled admissions (not via the pED) during the Covid-19 period compared to the pre-Covid period (27 vs 110, data not shown). This paper is the first, taking the bed capacity into account. Its implication is that lower admission numbers during the Covid-19 period as seen in many studies and in subgroups in the current study, too, do not mandatorily mean that there was a decrease in severity of presentations, but the driving factor could be the higher admission threshold due to a lower admission capacity. The policy implemented to avoid Covid-19 infection and collaterally diminishing bed capacity was certainly well contemplated and there is by now very strong evidence that the then pandemically present Covid-19 subtypes can lead to life-threatening disease, especially in psychiatric patients [61] and infection should consequently be avoided. Additionally, we should consider that at the observation time period no vaccinations were available.
Above, we suggested with some caveats that patients presenting in police custody might represent a severely ill subgroup. However, less admissions were observed in this subgroup (admission rate decreasing from 72.6 to 62.0%, Table 2). What is more, in this subgroup (cf previous paragraph), suicide attempts prior to pED presentation were more frequent in the Covid-19 than in the pre-Covid-19 period. Together with the decrease in hospital admissions in patients with SUD and paranoid schizophrenia, this might represent rather a consequence to limited bed capacity than a reduction in required inpatient treatment.
The fact that admissions were shorter in the Covid-19 period (8 vs. 10 days, Table 1) is also most likely due to the need of creating admission capacity and not due to less treatment necessity. This shortening was driven by significantly shorter admissions for SUD, NSSD and PD (5 vs. 8, 5 vs. 12 and 3 vs. 10 days respectively, Table 3s, supplementary section).
Taken together, we conclude that a lower bed capacity (due to infection curbing measures) might has led to a higher threshold to admit pED patients for inpatient treatment. We assume that the phenomenon of limited bed capacity due to infection-avoiding policies was - at least in Germany - existent common problem during the first Covid-19 wave as there was a governmental financial support from March 16th 2020 on for leaving beds empty for prospective Covid-positive patients who were - in Germany and Nordic countries - scarce in the beginning of the pandemic (cf no Covid-positive patients in our sample).
Strengths and limitations
We want to highlight some strengths of our study: it is based on thorough clinical documentation of which each case was reviewed individually. We covered a relatively large observation period (12 weeks) with a comparably large number of assessed pED presentations and subsequent admissions. Due to knowledge from previous studies, we were able to select the independent variables for the logistic regression model in an informed manner. This study is the first considering bed capacity.
There are also some limitations to be considered. First, it is important to point out, that this study is limited to one site, where data was collected. Furthermore, the control data is limited to the previous year only. Extrapolation of results should therefore be done cautiously. Although we implemented measures to minimize inter-rater bias during the process of data extraction from clinical records, we cannot rule out, that inter-rater bias exists to some extent. Also, there might be a bias in attention to certain details from the primary examining staff. It seems plausible, that e.g. medical conditions leading to a Covid-19 high-risk group classification were more likely to be asked for and documented during the Covid-19 pandemic. Furthermore, obesity was not always documented, only in severe cases. Also, in our sample of the beginning of the pandemic the knowledge of high-risk groups was possibly not as widely spread as later in the pandemic. Generally, the extent of documentation varied between the individual emergency ward files, possibly leading towards a more detailed description of more severe cases. Since this was rather evenly spread over the two compared time periods, we only see a minor bias risk due to this factor. Another limiting factor is the absence of a severity scale.