A retrospective multicenter transnational observational study was conducted on patients admitted to SPDC during the first year of the pandemic. We involved three psychiatric services located in three different countries with similar socio-demographic settings by comparing hospitalization rate time-trends for each diagnostic group. The hospitalization trends observed in 2020 were evaluated in comparison to the pre-2020 period (average admission rate of 2017, 2018, and 2019).

January 1st to December 31st of years 2017–2018–2019–2020.

A comprehensive survey of all patients admitted to the SPDC during the study period was conducted in all three participating institutions.

Average age (Table 1), distribution of admissions by gender (Table 2, Figure 1), and by diagnosis (Table 3) are shown for each psychiatric service.

Inclusion Criteria: Any gender and ethnic background aged 18 or older.

Exclusion Criteria: Patients admitted as part of the so-called “revolving door phenomenon” (readmission within one month) were excluded, as these phenomena develop independently of the COVID-19 pandemic (see Table 4).

Diagnoses were grouped according to the ICD-10 classification system into the following macro-categories:

For Austria and Germany, separate Poisson GLMs were fitted to examine the effects of Year (pre-2020 vs. 2020), nonlinear temporal dynamics (Week, modeled with splines), Diagnosis, and their interactions.

For each diagnostic category, a separate within-country model was estimated to test the Week × Year interaction.

A combined GLM including the factor Country was used to assess differences between Austria and Germany in baseline diagnostic composition and temporal changes from pre-2020 to 2020.

Analyses were repeated using standard ICD-10 groupings with a finer scheme separating alcohol-related disorders (F10) from the rest of the F1 category (other substances).

As an initial descriptive step, we performed visual descriptive observations to assess whether the COVID-19 pandemic altered the temporal structure of psychiatric hospitalizations across countries and diagnostic categories.

Hospitalization counts were aggregated in 15-day intervals for each country. To ensure temporal comparability, Week 0 was set on 8 March 2020, corresponding to the first national lockdown in Italy (the first lockdown in Europe). Based on this reference point, two shared lockdown windows were defined:

First lockdown (time frames 0–5)

Italy: 8 March—23 May

Austria: 16 March—13 April

Germany: 23 March—10 May.

Second lockdown (time frames 15–17)

Italy: 23 October—30 November

Austria: 11 November—30 November

Germany: 2 November—30 November.

For each diagnostic category, the 2020 trajectory was visually compared with the mean pre-pandemic trend (2017–2019).

To explore whether temporal deviations differed across diagnostic groups, we generated diagnosis-specific trajectories for Austria and Germany based on the predicted values of the within-country Poisson GLMs (described in the Statistical Method section).

This allowed us to visually assess whether:

As an additional step, we plotted the weekly Germany/Austria ratio of admissions for each diagnostic category.

In the pre-2020 period, stable horizontal ratios were interpreted as reflecting consistent cross-country proportionality for that diagnosis (even if absolute volumes differed).

Deviations observed in 2020 were taken as visual indicators of pandemic-related divergence in system behavior.

Integrative role of visual descriptive observations.

These descriptive procedures provided a framework that guided but did not determine the inferential analyses. The quantitative modeling presented in the subsequent sections formally tested whether the qualitative patterns observed here corresponded to statistically significant diagnosis-specific or country-specific temporal effects.

To explore the influence of different healthcare system structures, a supplementary analysis compared Austria and Germany (hospital-based systems) with Italy (a community-based model).

We quantified the variation in total admissions in 2020 relative to the mean of the three preceding years (2017–2019).

To ensure comparability across countries, diagnostic categories F0 (organic and symptomatic disorders) and F1/F10 (substance- and alcohol-related disorders) were excluded, given substantial differences in structure and treatment organization across healthcare systems.

The analysis revealed significant main effects of diagnosis, year, and bi-weekly temporal variation (Week), all showing highly significant deviations (p < 0.001). The interaction between diagnosis and Week was also significant (p = 0.043), indicating modest differences in the temporal structure of hospitalizations across diagnostic categories. The Year × Week interaction was significant as well (p = 0.031), confirming that the overall temporal distribution of admissions in 2020 differed from that of the preceding 3 years. A significant Diagnosis × Year interaction (p = 0.001) further indicated that the relative diagnostic composition of admissions shifted during the pandemic year. In contrast, the three-way interaction Diagnosis × Week × Year was not significant (p = 0.411), suggesting that, although hospitalization timing changed in 2020, this deviation occurred homogeneously across diagnostic groups. Overall, the German sample exhibited a broad reorganization of temporal patterns and diagnostic composition during the pandemic year, but without evidence of diagnosis-specific temporal signatures.

The analysis revealed significant main effects of diagnosis, year, and bi-weekly temporal variation, each showing highly significant deviations (p < 0.001). The Year × Week interaction was also significant (p = 0.018), confirming that the temporal distribution of hospitalizations in 2020 differed from that observed across the previous 3 years. Importantly, both the Diagnosis × Year interaction (p < 0.001) and the Diagnosis × Week × Year interaction (p = 0.002) were significant. These findings indicate that the pandemic year introduced diagnosis-specific temporal fluctuations that were not present in the pre-2020 period. In contrast, the Diagnosis × Week interaction alone was not significant (p = 0.455), suggesting that temporal differences among diagnostic groups emerged only in relation to the pandemic context, rather than reflecting stable pre-existing differences across diagnoses. In summary, the Austrian dataset shows a marked diagnosis-specific reorganization of inpatient admissions during 2020, with heterogeneous and diagnosis-dependent temporal trajectories—in clear contrast to the more homogeneous temporal response observed in Germany.

To determine whether the temporal alterations observed at the aggregate level also emerged within specific diagnostic categories, we examined the predicted trajectories from the Poisson GLMs for Austria and Germany separately (Figure 3). In the pre-2020 period, both countries exhibited smooth and consistent temporal trajectories across diagnostic categories. In 2020, however, lockdown-related disruptions produced visible and heterogeneous distortions, with distinct profiles in Austria and Germany. In Germany, most diagnostic categories showed a relatively homogeneous response: the disappearance of the spring peak was observed across diagnoses, along with a shared increase around November, corresponding to the second lockdown. In Austria, temporal deviations were more pronounced and clearly diagnosis-specific. Several diagnostic groups diverged substantially from their pre-2020 trajectories, indicating a less uniform systemic response. Alcohol-related disorders (F10) displayed the most distinctive pattern, characterized by an inverted biphasic trajectory—a marked reduction during both lockdowns followed by a rebound in summer (around August). Other diagnostic groups exhibited heterogeneous alterations, reflecting a stronger diagnosis-specific reorganization of inpatient care. Together, these findings suggest that the temporal reorganization of psychiatric admissions during 2020 varied both by diagnosis and by healthcare system. This provides a first exploratory visual evidence—later confirmed by the quantitative analyses—that the impact of the pandemic on inpatient psychiatric care was not uniform but shaped by diagnosis-specific and system-level factors.

To quantify diagnostic-specific differences between Austria and Germany and their evolution over time, we modelled the weekly ratio of admissions (Germany/Austria) for each ICD-10 diagnostic category (Figure 4). A value of 1 indicates parity between the two countries, values above 1 indicate higher admission rates in Germany, and values below 1 indicate higher rates in Austria. In the pre-pandemic years (red lines), the curves remained largely stable over time, but not necessarily centered on 1. This indicates that, although the seasonal trajectories were comparable across the two countries, the absolute hospitalization levels differed systematically for certain diagnoses—for instance, F0 showed stable ratios around 0.25, meaning that Germany consistently admitted about half as many patients as Austria, despite similar temporal patterns. In contrast, during 2020 (cyan lines), several diagnostic categories exhibited marked temporal deviations, reflecting system-specific alterations under pandemic conditions. The most evident divergence was observed for alcohol-related disorders (F10), where Germany showed a relative increase in admissions compared to Austria during the summer and autumn months, suggesting differences in how each system reorganized addiction-related care during the pandemic. Other diagnostic categories displayed subtler but still noticeable shifts, indicating that the pandemic altered the long-standing equilibrium between the two systems. Overall, these findings show that although Austria and Germany shared similar pre-pandemic temporal organizations of psychiatric care, the pandemic year introduced new asymmetries, particularly for diagnostic categories sensitive to service accessibility and treatment organization, such as substance- and alcohol-related disorders (Figure 4).

This study has limitations. The cross-national comparison involved two hospital centers rather than national datasets, and larger multicenter cohorts may reveal more nuanced diagnosis-specific effects. Larger samples would also allow more granular analyses, especially within substance-related disorders, where patterns may differ across subgroups (26–28). Future studies should incorporate demographic variables, socioeconomic indicators, and measures of lockdown stringency, which have been shown to modulate mental health outcomes during the pandemic (1–5). Although formal stringency indices report that Italy implemented slightly stricter measures during the first lockdown, Austria and Germany followed a broadly comparable two-phase pattern of restrictions (spring and autumn 2020). Given our biweekly temporal resolution and the organizational focus of our analysis, these moderate differences in restriction intensity were unlikely to introduce systematic bias, which is why no explicit “stringency” covariate was included. Extending analyses beyond 2020 will help determine whether these disruptions persisted or gradually normalized.

Despite these limitations, the analytic framework applied here—integrating diagnostic, temporal, and cross-national dimensions—offers a model for understanding how psychiatric systems respond to acute external disruptions. Our findings highlight the importance of resilient organizational planning, the maintenance of structured care pathways, and flexible resource allocation during public health emergencies. Diagnostic groups relying on planned admissions, such as alcohol-related disorders, may require particular attention to ensure continuity of care during future crises (25–28). Future research should extend this approach to post-pandemic longitudinal outcomes and broader service metrics to assess whether the apparent resilience of community-based systems translates into sustained mental-health benefits and more equitable access to care.