The Center for Acute Psychiatric Disorders, as part of the Psychiatric University Hospital of Zurich, is responsible for the inpatient treatment of adults aged 18–65 years in the city of Zurich and its surroundings, with a catchment area of ~500,000 habitants. It provides 100 beds distributed in eight wards to treat psychiatric crises and emergencies. Routine demographic and clinical data are recorded for each admission (and discharge). We drew on data from an 8-year period (January 1st, 2013- December 31st, 2020); the last year was solely used to observe service use after discharge. Therefore, only a patient's first admission during the observation period, with a 1-year follow-up after discharge, was included in the analysis. The Ethics Committee of the Canton of Zurich authorized the use of anonymized data for research and publication purposes (BASEC: 2018-01906).

The day and time of admission were recorded and categorized according to the daytime and weekday. Daytime was divided into three shifts: “daytime” (i.e., office hours from 8 to 17 h), “dusk” (from 18 h to midnight), and “dawn” (from midnight to 7 h). The days of the week were categorized as: “working days” (i.e., regular weekdays from Monday to Friday) and “weekends” (i.e., Saturdays, Sundays, and public holidays).

We operationalized “help-seeking behavior” 2-fold (9, 20, 21). First, as either “voluntary” or “involuntary” (i.e., by a compulsory admission order). Compulsory admission orders in the Canton of Zurich, Switzerland, may be issued by a qualified physician if the treatment or care of a mental disorder cannot be provided otherwise; in contrast to other countries, impending danger is not essential (22). Thus, allowing sensing the discrepancy in the need for treatment between medical judgment the patients' insight. Second, we further differentiated the pathway to admission according to referral as: “emergency services” (i.e., an ambulance or the emergency department of a general hospital); “mental health professionals” (i.e., psychiatrist and for voluntarily admitted patients also clinical psychologists); “health professionals” (i.e., any other medical discipline); thus allowing us to infer that care (i.e., therapy) was not possible at these sites. In addition, we established a fourth and exclusive category, “self-referral” (i.e., own initiative; walk-in: or advice by non-medical professionals) for voluntarily admitted patients.

Attending psychiatrists, psychiatry residents, or clinical psychologists carried out diagnoses and clinical ratings. Relevant information was derived from the clinical interview and reports of nursing staff, social workers, and significant others. Diagnoses were made according to the International Classification of Diseases, 10th edition (23) criteria and then grouped in one of eight diagnostic categories: “Dementia and Neurocognitive Disorders” (F0); “Alcohol Use Disorder” (F10); “Substance Use Disorder” (F11–F19); “Schizophrenia Spectrum Disorders” (F2); “Mania and Bipolar Disorder” (F30–F31), “Major Depressive Disorder” (F32–F34); “Anxiety and Stress-related Disorders” (F4); and “Personality Disorders” (F6). In addition to the main treatment diagnosis, we addressed psychiatric comorbidity with “Alcohol or Substance Use Disorder” (F10–F19); and “Personality Disorder” (F6). We chose diagnostic categories to be representative and large enough for statistical analysis and comparison.

For clinical rating, we used the Clinical Global Impressions Scale (CGI) and the Health of the Nation Outcome Scales (HoNOS). The CGI is a brief, easy-to-use measurement tool that assesses the severity and response to treatment in different subscales (24, 25). The Severity of Illness (CGI-S) scale is rated on a seven-point Likert-like scale response format which ranges from “1” (normal) to “7” (extremely ill). The Global Improvement (CGI-I) evaluates the change (i.e., either improvement or deterioration) in comparison to a previous CGI evaluation. It is rated on a seven-point Likert-like scale that ranges from “1” (very much improved) to “7” (very much worse), whereby a score of “4” indicates no change.

The HoNOS is a measurement instrument to assess the severity and the burden of a psychiatric disorder in 12 different domains (26, 27). Each item is rated on a five-point Likert-like scale ranging from “0” (no problem) to “4” (severe to very severe problem). The domains are defined at item level: 1. “Overactive, aggressive, disruptive or agitated behavior;” 2. “Non-accidental self-injury;” 3. “Problem-drinking or drug-taking;” 4. “Cognitive problems;” 5. “Physical illness or disability problems;” 6. “Problems associated with hallucinations and delusions;” 7. “Problems with depressed mood;” 8. “Other mental and behavioral problems” (including phobic; anxiety; obsessive-compulsive; mental strain/tension; dissociative; somatoform; eating; sleep; sexual and other), 9; “Problems with relationships;” 10. “Problems with activities of daily living;” 11. “Problems with living conditions” and 12. “Problems with occupation and activities.” There is a glossary with a clear and concise definition for each domain and rating instructions. We evaluated the HoNOS at scale level (i.e., sum score ranging from 0 to 48) and item level. We considered, items rated three or four as clinically significant and part of the patients' care and treatment plan (28–30).

We extracted the pharmacological and non-pharmacological treatments prescribed and coercive measures executed during hospitalization from the electronic health record. The pharmacological treatment comprises all medication approved by the national authority. We classified the pharmacological treatment according to their indication: “antipsychotics,” “antidepressants,” “mood- stabilizers” (i.e., lithium and antiepileptics), “anxiolytics and hypnotics,” “narcotics” (i.e., scheduled medication as amphetamines, or opioids), and “other” (pharmacological treatment for non-psychiatric disorders).

We defined the non-pharmacological treatments as “crisis intervention” (short-term treatment to mental, emotional, and behavioral distress due to a psychiatric disorder), “counseling” (professional guidance, information, and motivational counseling), “observation” (of symptoms and behavior for diagnostic appraisal), “psychotherapy” (either single or group), and “occupational therapy” (art, work, sport and leisure activities). Patients were entitled to receive more than one treatment modality.

Coercive measures were classified as: “isolation- restraint,” “forced medication,” and “compulsory retention;” patients could have undergone one or more coercive measures during their hospitalization. If a patient had been compulsorily admitted, we also recorded those that changed their willingness and remained voluntarily in psychiatric treatment.

We defined the type of discharge as either “regular;” “irregular” (discharge against medical advice, court order, or patients who leave the hospital area without further notice); “death or suicide.” Same hospital readmissions and thus, the total count and the rate of and time to readmission within 1 year after discharge were included.

According to the principle of independence, the analysis only included the first admission between January 1st, 2013, and December 31st, 2019. For the analysis, we considered the day (i.e., weekdays and weekend days) and time (i.e., day, dawn, or dusk) of admission as the independent variable. Descriptive statistics (mean, SD, median, percentage) were used for demographic and clinical characteristics. Differences in categorical variables were assessed with the Chi-square Test, with a subsequent Chi-square omnibus comparison for the day of the week (i.e., weekdays vs. weekends) and time of the day (i.e., day vs. dawn vs. dusk) of admission (31). Using the residuals, we calculated the percentage a given category contributes to the Chi-squared score. Outcomes of interest were additionally represented as Odds Ratios (OR). Logistic regression was used to control confounders; we adjusted OR for age, sex, education, marital status, admission status, diagnosis, and severity. ANOVA assessed differences in continuous variables, with a pairwise comparison with the Student's t-test. If assumptions about the distribution were not met, we additionally used an alternative non-parametric test (i.e., the Kruskal–Wallis test and the Mann–Whitney U-Test). All tests of significance were two-tailed. Due to the large sample size, we chose a significance level of p < 0.01. However, considering that the day and time categories have a hierarchical structure, we decided to introduce a correction (by a factor of six); thus, we consider only p-values < 0.001 statistically significant. Effect sizes were assessed with eta-square for continuous and phi for categorical variables (32, 33). Kaplan Meier time-to-event curves were drawn and analyzed for the length of stay and time to readmission. To test for statistical significance, we used the log-rank p-value. Statistical analyses were conducted using the statistical software R (v4.0.3).

During the study period, 12,449 direct patient admissions to an acute psychiatric ward were recorded. Most admissions (n = 10,452, 84.0%) occurred on a working day; the remaining (n = 1,997, 16.0%) were on a weekend or public holiday. Admissions occurred predominantly during daytime hours (n = 7,924, 63.7%). Correspondingly, admissions occurred during office hours (n = 6,886, 65.9%). The percentage on weekend days (n = 1,038, 52.0%) and outside office hours (n = 959, 48.0%) was similar.

The mean age of the sample was 48.05 ± 20.90 years, with 49.32% (n = 6,140) females. More than one-third of all admissions (39.7%, n = 4,950) were compulsory; the majority (41.0%, n = 2,030) of them was issued by a mental health professional (i.e., a psychiatrist); with almost equal proportions for emergency services (30.3%, n = 1,499) and non-mental health professionals (28.7%, n = 1,421). The mean CGI-S at admission scored 4.76 ± 1.10; the mean HoNOS sum score was 19.66 ± 7.42. The most frequent treatment diagnoses were: major depressive disorder (n = 3,232, 25.9%); anxiety and stress-related disorders (n = 2,300, 18.5%); and schizophrenia spectrum disorders (n = 2,226, 17.9%), which together accounted for two thirds of all admissions.

Patients admitted during the daytime on working days were slightly older. Females were more frequently admitted during the daytime (both working and weekend days) than their male counterparts. Swiss nationals were more likely to be admitted during office hours, while tourists and travelers were more frequently admitted at out-of-office hours. During office hours and daytime, fewer compulsory admissions were recorded: they steadily increased from dusk till dawn until a peak of compulsory admissions was reached around 5–6 a.m. The increase was more prominent on weekends (see Figure 1). The rate of voluntary admissions, especially self-referral, was higher during the daytime on weekends. Further demographic and clinical characteristics according to the day and time of admission are summarized in Table 1 and Figure 1.

The admission rate for certain diagnoses varied from working days to weekends, increasing until dawn. This was the case of alcohol and substance use disorders, either as the main treatment diagnosis or comorbid disorder (See Figure 2). The clinical severity and burden showed a slightly differing profile according to the day and time of admission (see Figure 3); HoNOS Items rating “overactive, aggressive, disruptive or agitated behavior” (Item 01); “non-accidental self-injury” (Item 02) were higher after day time: “problem drinking or drug-taking” (Item 03) was higher at dawn irrespective of the day of the week; and “physical illness or disability problems” (Item 05) was higher at office hours. The CGI-S severity score and the HoNOS sum scores were lower on weekends, irrespective of the time of the day. However, the effect sizes are very small.

For further detail, see Table 2 and Figure 2.

There are slight differences regarding pharmacological and non-pharmacological treatment perceived according to the day and time of admission. Patients admitted outside of office hours generally had fewer prescribed therapies (See Figures 4A,B). These differences disappeared after correcting for confounders (age, sex, education, marital status, admission status, diagnosis, and severity). The mean length of stay was 20.62 ± 28.52 days with a right-skewed distribution and a median of 13 (IQR: 29) days. Patients admitted during the daytime of a weekday had a longer length of stay than all others (24.70 ± 27.49). Compulsorily admitted patients almost doubled the length of stay of those voluntarily admitted [F_{(5, 12, 429)} = 4.539, p < 0.001]; although this difference remained significant only for those admitted during daytime (See Table 2, Figures 5A,B). Patients voluntarily admitted during the daytime on the weekend had a shorter length of stay with a median of 4 days. The rate of compulsory retention was similar among all groups. The rate of and time to readmission and the total count of subsequent readmissions were similar for all groups (For further details, see Table 2, Figures 5C,D).

Clinical ratings showed a significant improvement, regardless of the day and time of admission. The CGI-I ratings showed a robust clinical improvement, so did the HoNOS sum score difference among all groups between admission and discharge [F_{(11, 22, 610)} = 960.05, p < 0.001; η² = 0.30]. Nonetheless, patients admitted outside office-hours (i.e., all except daytime on working days) achieved a significantly higher sum score difference (office hours 9.05 ± 8.04 vs. cumulative outside office-hours 8.11 ± 7.60; F_{(5, 12, 443)} = 9.4, p < 0.001) and correspondingly a higher percentage (office hours 46.3 ± 38.1 vs. cumulative outside office hours 42.6 ± 40.7; F_{(5, 11, 374)} = 6.7, p < 0.001) of improvement (see Table 2). However, the effect size of the difference was small. Thus, the total count of clinically relevant domains could be reduced during hospitalization. The differences in the clinical ratings between the day and time of admission were also statistically significant but again showed an overall very low effect size (see Table 2). Patients admitted at evening and night shifts showed higher rates of coercive measures (i.e., forceful medication and seclusion/restrain) irrespectively of the day of the week (see Figures 6A,B).