Despite widespread deinstitutionalisation, the demand for inpatient psychiatric care still far outweighs supply, particularly for patients who pose a significant risk, refuse necessary treatment or have severe symptoms impairing daily functioning.^1,2,3,4 In South Africa, deinstitutionalisation following the advent of democracy in the mid-1990s resulted in a substantial reduction in inpatient beds, with chronic psychiatric beds in Gauteng halving between 1994 and 2004.⁵ This reduction in inpatient capacity has led to high bed occupancy, shortages and shorter hospital stays.^1,2,6,7 Admission practices have been influenced, with inpatient units increasingly admitting patients with severe psychopathology, particularly psychotic and substance use disorders, while fewer patients with non-psychotic presentations are admitted.^8,9 Even in settings with relatively good community support, bed shortages contribute to declining patient well-being in the community, longer lengths of stay once admitted, and pressure to discharge patients who remain symptomatic.^6,10,11 As a result, only the most severely ill patients are admitted,⁹ and they are often discharged rapidly, transforming inpatient units into short-term crisis-intervention services.^12,13

Ongoing shortages of adequate community mental health resources, particularly in rural areas, remain a major constraint in South Africa.^14,15 Consequently, despite policy aims to move away from a hospital-centred model, inpatient psychiatric care remains a critical component in South Africa and other low- to middle-income countries (LMICs), highlighting the importance of efficient use of inpatient services in resource-limited settings.^1,14,16 Some international organisations have set the maximum bed occupancy in a psychiatric unit at 85% to ensure patient and staff safety.¹⁰ However, psychiatric hospitals in the Western Cape typically operate at 100% or higher bed occupancy, according to the Mental Health Dashboard on the Single Patient Viewer platform.¹⁷

Inpatient clinical teams may be required to periodically review and adjust their practices in response to ongoing bed-capacity constraints within the South African state psychiatric sector. Stikland Hospital follows a ‘crisis discharge policy’, adopted by the associated psychiatric hospital management team in the Western Cape province, South Africa, which governs the practice of premature discharge, as standard operating procedure since 2003.¹⁸ A ‘crisis discharge’ is made when no inpatient beds are available, but a patient requires urgent admission. Some safeguards exist to limit risk with crisis discharges, including that the patient should be the most stable in the ward, not pose any immediate risks to him-/herself or others, be less unwell than the patient requiring admission and be able to follow up with practical plans in place.

Given the above, it is not surprising that readmission rates for psychiatric patients remain high in South Africa. Almost one in four patients is readmitted within 3 months of psychiatric discharge, with readmission costing an estimated 1.5 billion rand in 2016/2017, equal to almost a fifth of the mental healthcare budget.¹⁴ Crisis discharges and similar practices have the potential to contribute to relapse, which disrupts psychosocial rehabilitation, employment and other psychosocial support, and increases distress for the patient and family, risk for treatment resistance, and costs of care.^19,20,21,22 Demographic and clinical factors that have been associated with a higher risk of relapse include male gender, a schizophrenia diagnosis, a history of an admission for stimulant use disorders before the index admission with psychosis, psychiatric comorbidity, treatment adherence, previous hospitalisation and lack of housing.^23,24,25 There are also system-related factors associated with relapse, which include limitations of available inpatient care programmes, lack of availability of prescribed medication in the community, poor post-discharge care, lack of comprehensive community care, lack of coordination with community mental health workers, disruption to social security benefits and locations of the hospitals and patients.^11,24,25 Readmission is commonly used as a measurement of relapse.^26,27 Readmission can be conceptualised as the outcome of an individual’s specific needs exceeding the capacity of the environment and support system.⁷ Early readmission is therefore often seen as an indication of too early discharge, poor quality of psychiatric care or lack of discharge coordination.^14,28

The Western Cape province has the highest documented lifetime prevalence of common mental disorders in South Africa. Despite this, the Western Cape has only 0.89 psychiatrists per 100 000 of the uninsured population compared to a global average of 3.96 per 100 000 of the population.¹⁴ The majority of South Africans with serious mental illness (SMI) receive little to no psychiatric care. In settings with limited community resources, this places an additional responsibility on inpatient services to stabilise patients adequately.²⁹

Stikland Hospital is one of three specialised state psychiatric hospitals in the Western Cape province of South Africa. The Tygerberg South substructure falls within the service area of this specialised facility. However, it currently lacks a dedicated district-level psychiatric admission facility, resulting in this specialised hospital also providing services that are typically delivered at the district-level for this substructure. Consequently, some patients are admitted directly from the community to specialised-level care, a process commonly referred to as a ‘direct admission’, in other words, without a preceding admission at a district hospital. District-level facilities would ordinarily assist with the initial stabilisation of medical comorbidities and the early management and containment of acute psychiatric presentations before transfer to specialised services. Direct admissions likely contribute to the emergency management demands placed on tertiary acute psychiatric units and could add to the workload of clinical staff in these settings.

Understanding the current challenges facing this mental health system is therefore important not only within our own context but also in the broader public mental health service in South Africa and similar LMICs in general. Challenges such as high bed occupancy, high demand for beds, and limited community resources place conflicting pressures on acute services, leading to rapid discharges and, in our setting, an official crisis discharge policy. A description of this cohort and an evaluation of readmission rates within this service may provide important insights to inform current clinical practices and guide the development of strategies aimed at reducing readmissions.

A total of 712 discharges were identified from hospital records. Of these, 10 discharges were transferred to a different hospital for further tertiary care, six discharges were found to have incomplete records, two discharges had duplicate records, and one discharge died during admission. A total of 693 discharges were analysed. The median (IQR) age of the sample was 29 (23–35) years (range 18–59). Most of the participants were not in a relationship (n = 617, 89%) and lived with family (n = 602, 86.9) (Table 1). Just over two-thirds of participants did not complete schooling (n = 453, 65.4%), most were unemployed (n = 650, 80.8%) and only 23.2% (n = 161) were recipients of a disability grant.

Almost half (n = 331, 47.8%) of all index admissions were previously admitted to Stikland Hospital (Table 2), with patients having a median (IQR) of 3 (1–5) previous admissions. The duration of previous Stikland Hospital admissions ranged between 39 days and 45 days, and the median (IQR) total length of stay for previous Stikland Hospital admissions was 123 (56–224) days. More than two-thirds (n = 480, 69.3%) of the patients were referred from other hospitals to Stikland Hospital for their index admission. The median (IQR) length of stay of the index Stikland Hospital admission was 36 (27–53) days. A total of 30.7% (n = 213) of all male acute admissions were admitted directly into the service without being admitted for a period of time in a referral centre.

For readmission to Stikland Hospital within 1 year after their index admission, 191 patients had at least one post-index readmission, and 40 patients had four post-index readmissions (Table 2). The median (IQR) time to first readmission was 137 (68–238) days. The median length of stay of post-index readmissions to Stikland Hospital ranged between 37 and 43 days. For readmission to a psychiatric unit at a district hospital, 158 patients had at least one post-index readmission, and two patients had five post-index readmissions. The median (IQR) time to first readmission was 180 (66–244) days. The median length of stay of post-index readmissions to a district hospital ranged between 8 days and 18 days. Eight patients were readmitted to another psychiatric hospital (other than Stikland Hospital) for a median (IQR) length of stay of 29 (19–54) days after their index admission.

The most common primary psychiatric diagnoses at Stikland Hospital index discharge were schizophrenia (n = 292, 42.1%), substance-induced psychotic disorder (n = 135, 19.5%) and schizophreniform disorder (n = 87, 12.6%) (Table 3). An ‘other’ (n = 162, 23.4%) and substance use disorder (n = 134, 19.3%) were the most common secondary psychiatric diagnoses. Just over one-fifth of participants (n = 159, 22.9%) presented with a first episode of psychosis. Of these patients with a first episode of psychosis, 30 (18.9%) were readmitted to a psychiatric hospital or unit within 1 year post-index discharge. Forty per cent of participants (n = 280) had untreated psychosis for a median (IQR) of 30 (7–90) days. Almost two-thirds of participants were crisis-discharged (n = 440, 63.5%).

Co-morbid medical conditions were infrequent, with hypertension (n = 52, 7.5%), COVID-19 (n = 50, 7.2%), pulmonary TB (n = 41, 5.9%) and syphilis (n = 34, 5.0%) being the most frequently observed (Table 4). A diagnosis of human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) was documented in 3.5% of admissions.

A high prevalence of substance use was noted in this population, with cannabis, methamphetamine and methaqualone being the substances most frequently used (Table 5). The prevalence of lifetime use of cannabis was 81.5% (n = 565), methamphetamines was 54.4% (n = 377) and methaqualone was 47.5% (n = 329). The prevalence of prior-month use of cannabis was 29.6% (n = 205), methamphetamines was 20.8% (n = 144) and methaqualone was 29.6% (n = 205).

Almost all patients were prescribed antipsychotics during their index admission/discharge (n = 688, 99.3%), while only 7.2% (n = 50) were prescribed an antidepressant (Table 6). Of those prescribed an antipsychotic, the median (IQR) number of antipsychotics was 2 (2–2), 73.4% (n = 505) were prescribed a long-acting injectable (LAI) and 22.7% (n = 156) were prescribed clozapine.

Stikland Hospital is a specialised psychiatric hospital in Bellville, Western Cape province, South Africa. The male acute service is the most highly pressured unit within the hospital, and this study aimed to provide a descriptive summary of the demographic and clinical characteristics of patients utilising this service. Most of the admissions were young, single and unemployed with no secondary-level education. Almost a third of all admissions were admitted directly into specialised care, and most patients presented with a diagnosis of schizophrenia. Most admissions had a short length of stay, were discharged as crisis discharges and were admitted as substance users. While the use of LAI was highly prevalent, readmissions remained a significant challenge within the service.

The index admissions had a median length of stay of 36 days, which is similar to the UK (35.9 days).¹⁰ A shorter length-of-stays (LOS) has been associated with substance abuse and psychiatrists managing high-volume caseloads, which are both significant factors in this population.³³ Average LOS in most industrialised countries varies from 6.4–13 days in the US to about 35.9 days in the United Kingdom, compared to 54.4–79.7 days in Japan.^10,12,34 The lack of post-discharge community coordination and transitional care in the Japanese system leads to a greater emphasis on full stabilisation of the patient’s condition during the inpatient admission and subsequently longer stays.¹² The situation with regard to the lack of community resources is similar in many developing nations. In China, for example, psychiatric beds are limited and unequally distributed, which has been linked to long psychiatric admissions, with an average LOS of 50–60 days.¹⁶ A longer LOS is associated with the availability of beds, which is a significant limiting factor in our population. Stikland Hospital’s male acute unit always runs above 100% bed capacity with a long waiting list for admission. The continuous bed pressure leads to crisis discharges and, therefore, a shorter LOS. Considering the lack of community resources, a longer LOS would ideally assist with greater stabilisation before discharge; however, system-related factors such as bed pressure render a longer LOS unlikely.

Almost half of all admissions in this study were patients previously admitted to Stikland Hospital, and more than one quarter were readmitted within 1 year following discharge. A readmission rate of less than 15% within 28 days was previously suggested as an acceptable standard.¹ Although relapse is expected as a part of the natural course of psychotic illness,^16,20 almost a third of patients are readmitted into acute psychiatric services within 3 months of their discharge.³⁵ In the context of increasing pressure on acute psychiatric services, readmission rates represent a critical area for targeted intervention, with efforts to reduce readmissions offering a potential means of alleviating strain on healthcare systems and improving continuity of care.

Niehaus et al.¹⁸ also reported results from a Stikland Hospital study in 2008, which showed that patients discharged as usual had a far lower risk of readmission compared to those discharged as ‘crisis discharges’. Despite this finding, the crisis discharge policy remains practised because of ongoing bed shortages for acute inpatient admissions. Our study showed that 63.5% of discharges within our study period were documented as crisis discharges. This remains a significant factor to consider when assessing the risk of readmission.

Schizophrenia was the most predominant diagnosis listed on discharge in this cohort. Zhan et al.³⁶ analysed the results of the Global Burden of Disease study for schizophrenia, looking at trends from 1990 to 2021 and projections to 2050. The results found that between 1990 and 2021, the prevalence of schizophrenia increased from 13.62 million to 23.18 million, which reflects a 70% increase. The incidence of schizophrenia showed a 38.5% increase as it escalated from 883 000 to 1.223 million. The authors highlight the anticipated increase in the burden of schizophrenia as projections indicate a continuous rise through 2050. Policymakers are advised to revise policies and implement measures, which include the development of comprehensive systems to proactively address the anticipated increase.

Our study also showed that just over a fifth of the study population presented with a first episode of psychosis. Within the first episode psychosis group, 18.9% were readmitted within 1 year after discharge. The importance of early intervention in first episode psychosis remains an extensively studied topic, and it is well known that early intervention leads to improved prognosis. The review by Nikolai et al.³⁷ highlighted the importance of reducing the duration of untreated psychosis and providing specialised early intervention treatment in first episode psychosis.

Non-psychiatric medical comorbidity is an interesting and seemingly important marker of poor mental health, and possibly acts as a precipitant to acute psychiatric episodes.³⁸ Three of the four most frequently noted medical comorbidities in this study were infectious diseases. People with schizophrenia are known to be at an increased risk of infectious diseases.^39,40 Anecdotal evidence from South Africa suggests that the stress of the COVID-19 pandemic was linked to an increase in medication discontinuation and schizophrenia relapses.³⁹ In the initial COVID-19 post-pandemic period, there were increases noted in psychiatric admissions for psychosis.^41,42 It seems likely that psychotic disorders continue to predominate psychiatric admissions in the post-COVID-19 era, leading to higher numbers of involuntary admissions and greater pressure on psychiatric hospitals.⁴¹

The South African Stress and Health (SASH) study, a large population-based mental health epidemiological survey, showed that the Western Cape had a 20.6% lifetime prevalence of substance use, the highest compared to other provinces within South Africa.⁴³ Our study showed that substance use remains a highly prevalent condition within this population. Multiple substance use was previously associated with more frequent readmissions and a shorter LOS in schizophrenia spectrum disorders.⁴⁴ Florentin et al.⁴⁵ found that dual diagnosis (severe mental illness with a substance use disorder) patients had a significantly younger mean age at first hospitalisation and shorter average LOS per hospitalisation, but a greater total number of hospitalisations and hospital days when compared to non-substance users. Substance use interventions should be considered a priority in this service or post-discharge community planning, as they could be a contributing factor to readmission and worsening prognosis.

The use of antipsychotic polypharmacy has been shown to increase the risk for readmission when compared to antipsychotic monotherapy.⁴⁶ In this cohort, the majority of patients were discharged on a LAI antipsychotic, and the median number of prescribed antipsychotics was two. Patel et al.⁴⁷ showed that LAI antipsychotics used in psychotic disorders decreased readmission rates at 90 days and 1 year post-discharge at a state psychiatric hospital in the US. Only one-fifth of the cohort was prescribed clozapine, which is lower than the one-third of patients found to be on clozapine treatment in a tertiary hospital in Johannesburg, South Africa.⁴⁸ A retrospective cohort study conducted in Taiwan between 2006 and 2021 showed that patients on a clozapine and an LAI combination had a longer time to rehospitalisation when compared to clozapine monotherapy or clozapine with other antipsychotic regimens.⁴⁹ Further studies are required to compare the categories of medication use to readmission in this population.

A limitation of the study is the fact that conducting a retrospective study relies on the records kept by others. The accuracy of the data is dependent on several self-reported pieces of information and collateral information provided by family members. Retrospective data also limits inference of causality. Further studies are needed to provide deeper insight into factors associated with low-frequency and high-frequency service users.

Inpatient psychiatric care remains an essential service in South Africa, but current services are functioning as crisis intervention units. Most admissions are as a result of psychotic disorders, and substance use remains highly prevalent in this setting. Current system-related factors, such as bed shortages, lead to short LOS and crisis discharges, which likely contribute to a higher rate of readmissions. With the expected ongoing increase in prevalence of schizophrenia spectrum disorders, this study contributes to our understanding of the current patient profile. This will provide a foundation for developing context-specific, tailored strategies to be integrated into the discharge planning processes within the existing healthcare system.