Psychiatric care is shifting from institutionalisation to outpatient care and day treatment services.¹ This shift was driven by multiple factors, including the increasing cost of hospital admissions, the introduction of effective psychiatric medications and growing public awareness of human rights violations faced by individuals with bipolar disorder (BD), schizophrenia and other psychiatric illnesses collectively known as serious mental illness (SMI).^1,2

In this context, understanding the factors that influence the length of hospital admission (LOA) for psychiatric inpatients is important. Long LOA is associated with substantial clinical, psychosocial and economic burdens, especially in resource-limited settings, such as South Africa.^1,3 Length of hospital admission directly influences healthcare costs, making it an important consideration in financial planning for psychiatric services.

Research findings on the ideal LOA for patients with SMI are inconsistent. Some studies have found that shorter admissions have similar relapse and readmission rates to longer admissions, while others have suggested that shorter admissions have a negative impact on these outcomes.^4,5 The average LOA for patients admitted to psychiatric wards differs by country. In high-income nations, the typical duration of hospitalisation in acute psychiatric settings ranges from 10 days to 43 days.^6,7 South African statistics indicate a longer LOA in designated psychiatric hospitals, averaging 219 days, compared to 11 days in general hospitals and 7 days in district hospitals.⁸ Factors associated with longer LOA included older age, being female, being unmarried, involuntary admission, psychosis, unemployment, homelessness and residence in areas lacking community psychiatric services.^9,10,11

While existing data have focused on the LOA in patients with SMI, with particular emphasis on schizophrenia,¹¹ limited data exist on the LOA in patients with BD in South Africa. Bipolar disorder is a mood disorder that affects 44 million people worldwide and was ranked among the top 10 chronic conditions in South Africa in 2016.^12,13 Hospital admission is often required for acute manic and severe depressive episodes.¹³ The duration of hospitalisation for individuals with BD is influenced by factors such as the course of the illness, response to pharmacological treatment and the presence of comorbid psychiatric conditions such as substance use disorders and medical conditions such as human immunodeficiency virus (HIV).^14,15

According to the World Health Organization (WHO), South Africa has a high prevalence of HIV, with an estimated 3.9 million people living with HIV in 2023.¹⁶ People living with HIV may experience worse mood swings owing to medication side effects, neuroinflammation or opportunistic infections.¹⁷ Substance use, particularly alcohol, stimulants and cannabis use, is prevalent among patients with BD and is also associated with higher relapse rates and reduced efficacy of mood stabilisers.^18,19 Substance use is associated with shorter duration of admission in patients with BD.¹⁹ It is possible that this shorter admission may be because of intoxication, with symptoms resolving when the substance wears off.

In September 2018, the South African Constitutional Court decriminalised the private use and cultivation of cannabis by adults, which may have led to an increase in cannabis use.^20,21,22,23 Mona et al.’s study in KwaZulu-Natal found that up to 50% of patients admitted with SMI reported cannabis use.²⁴ Understanding how cannabis use post-decriminalisation affects LOA in patients with BD is important for mental health planning. If cannabis use extends the LOA, this could lead to increased healthcare costs, necessitating targeted interventions. However, if the LOA is reduced but the risk of relapse increases, different adaptations will be necessary.

Given the limited data on LOA among South African patients with BD, this study aimed to determine the factors associated with hospital stay duration, focusing on those who use cannabis. Understanding these relationships could inform resource allocation, intervention strategies and policies in mental health services.

Altogether 192 patients were admitted, and 214 admissions were recorded (as some patients were admitted multiple times). Of the 192 patients, 90.62% were admitted once, 7.29% were admitted twice and 2.08% were admitted three times. More female patients were admitted (N = 159, 74.29%) than male patients (N = 55, 25.70%). Of the total admissions, the median (IQR) age was 36 (from 29 to 51) years. Most of the admissions were between 18 years and 35 years of age (N = 105, 49.07%). Most patients were not in a relationship (N = 162, 75.70%), most were unemployed (N = 186, 86.92%) and almost two-thirds (N = 132, 61.68%) had mania with psychotic features. Personality disorders were the most common psychiatric comorbidity (N = 32, 14.95%). The majority had comorbid borderline personality disorder (N = 17, 7.94%) and antisocial personality disorder (N = 8, 3.74%). Of the admitted patients with BD, almost half (N = 97, 45.32%) had a medical comorbidity, with a fifth (N = 46, 21.49%) being infected with HIV. Most patients had a diagnosis of bipolar I disorder (N = 207, 96.72%), with the minority (N = 7, 3.27%) being diagnosed with bipolar II disorder (Table 1).

The median (IQR) LOA of the 214 admissions was 23 (range: 13–38) days. Eighteen patients with multiple admissions stayed in the hospital for a median (IQR) of 45 (range: 32–80) days over the duration of data collection. All patients with multiple admissions had a history of substance use and reported ongoing cannabis use. Of the patients admitted, 80 (37.38%) used cannabis, 38 (17.76%) used crystal methamphetamine, 29 (13.55%) used alcohol, 23 (10.75%) used nicotine and 5 (2.34%) used heroin. Over a quarter of the patients (N = 58, 27.10%) used two or more substances. Of those who used cannabis, 28 (35.00%) were identified by the MDUT, while 52 (65.00%) were self-reported. No MDUT was performed in 46 (57.50%) patients who reported cannabis use. Of those who used other substances, 21 (22.11%) were confirmed to use other substances via the MDUT, while 74 (77.89%) were self- or family-reported. As a result of a lack of resources, some patients did not get an MDUT (Table 2).

No significant association was found between LOA, age, relationship status and employment. Psychiatric, medical comorbidities, cannabis and other substance use were not associated with LOA.

This study examined the factors associated with LOA in patients with BD, as well as the relationship between LOA and cannabis use in patients with BD. The decriminalisation of cannabis use in South Africa could increase its use, which may have important clinical implications. The median LOA was 23 days, which aligns with the duration of admission for patients with SMI from other local studies and healthcare facilities in sub-Saharan Africa.^15,19,25 Some studies have reported shorter LOAs, ranging from 9 days to 14 days, in patients with mood disorders.^26,27 These studies focused on all mood disorders (such as major depressive disorder); therefore, their results may not be directly generalisable to this study population. Designated psychiatric hospitals in South Africa have reported longer LOA.^28,29,30 This may be because these hospitals are referral centres providing care to patients requiring longer admissions, while the setting of this study was in acute psychiatric care wards.

Approximately 8.41% (N = 18) of the cohort had multiple admissions and stayed longer in the hospital when the duration of all admissions during the study period was considered. They had a median hospital stay of 45 days (IQR: 32–80). In South Africa, the ongoing bed shortage in many facilities may necessitate a higher discharge rate. While this likely contributes to lower LOA, it may potentially contribute to an increase in readmission rates. According to Glick et al., the deinstitutionalisation and shift to community-based care led to lower LOA, thereby, minimising healthcare costs.³¹

The socio-demographic profiles of the patients were similar to those reported in other studies.^19,32,33 Most of the patients were young, female, single and unemployed. The median age at admission was 36 years (IQR: 29–51), with 49.07% (N = 105) (Table 1) of admissions occurring in patients aged 18–35. The high proportion of young people in this cohort is consistent with the typical age of onset of BD, which usually manifests in late adolescence and early adulthood.³⁴ Being older, female and having medical and psychiatric comorbidities were associated with a longer LOA.^34,35

Most patients were not in a relationship (N = 162, 75.70%) and were unemployed (N = 186, 86.92%) (Table 1). This may reflect the negative impact of BD on patients’ lives, affecting their ability to maintain stable employment because of mood instability, cognitive impairments and the stigma they experience.³⁶ In this study, relationship status was not associated with LOA. This result contradicts the literature, which reports that patients in a relationship have a shorter LOA than those who are not.³⁷ The shorter LOA was attributed to the partners’ involvement in patient management, enhancing early symptom detection and earlier admission.³⁷ The lack of association between relationship status and LOA may be explained by the fact that not all support comes from an intimate partner. Support from other family members may serve the same purpose and facilitate a shorter LOA.³⁷

More female patients (N = 159, 73.44%) were admitted than male patients (N = 55, 26.56%) (Table 1). This was consistent with the literature, which reported that more female patients with BD are admitted, despite a similar prevalence of BD in male and female patients.^38,39 Despite more women being admitted, gender in patients was not a significant predictor of the length of hospital stay (Table 3), which aligns with the findings of similar studies.^14,17 The higher number of female patients’ admissions could be explained by factors such as women being more likely to seek professional help from medical and mental health care⁴⁰ and the types of medications prescribed.⁴¹ Karanti et al. found evidence of gender bias in the treatment of BD and concluded that, at least in Sweden, clinicians’ treatment decisions could be unjustly influenced by their patients’ gender. Women were less likely to be prescribed lithium, an acknowledged gold standard for BD maintenance therapy that is associated with less relapses compared to other mood stabilisers.^41,42,43,44 They found that women were mostly prescribed antidepressants, which, when used without adjunctive therapies, worsen rapid cycling and can trigger mania, leading to hospital admissions.^41,45 This study did not explore medication choices among patients admitted with BD and thus could not comment on whether medication choice influenced the admission rate or LOA.

Chris Hani Baragwanath Academic Hospital has a maternal mental health clinic that screens and treats women with mental illnesses. This clinic serves patients with BD who are pregnant and post-partum from CHBAH and other surrounding community clinics. Consequently, there may be improved screening and identification of women who require admission. The post-partum status of some women may be a factor necessitating a lower threshold for admission, given the urgency to stabilise their condition and minimise risk to their infants. However, the reasons for admission and the pregnancy status of the women admitted were not captured in this study, and so the above interpretation remains speculative.

Over 95% of the admissions were diagnosed with bipolar I disorder. Almost two-thirds of the patients presented with mania along with psychotic symptoms, followed by over 30% of the patients who presented with manic episodes without psychotic symptoms. In contrast, patients admitted with depressive episodes and hypomania are rare. This is expected because admission of patients with BD is mostly indicated in the manic phase, which is difficult to treat on an outpatient basis because of poor treatment adherence or during severe depressive episodes with increased suicide risk.⁴⁶ Hypomania usually does not require admission; however, the presence of factors such as comorbid psychiatric disorders (e.g. personality disorders or substance use) may complicate the clinical picture, necessitating admission. Psychosis during manic or depressive episodes is associated with more severe impairments and predicts a more difficult-to-treat disorder; thus, it may be a risk factor for a longer LOA.⁴⁷ The clinical presentation of the patients with BD did not influence the LOA, which is consistent with the findings of Shi X et al.³²

Both medical and psychiatric comorbidities are common in patients with BD.⁴⁸ Medical comorbidities have been found to increase LOA in patients with SMI.^48,49 This study collected all medical and psychiatric comorbidities documented by the treating doctors giving the discharge summaries; however, it was beyond the scope of this study to examine how individual diagnoses may have impacted LOA. Over 21.49% of the admitted patients were HIV-infected. This is in keeping with the South African literature, which reports HIV prevalence to be between 20.3% and 26.5% in the South African population with SMI.^50,51 The prevalence of HIV in the South African general population is currently about 12.7%.⁵² Individuals with BD are at a higher risk of contracting HIV, as some patients may engage in risky sexual behaviour during manic and hypomanic episodes.⁵³ The variation in HIV prevalence in patients with SMI is likely explained by multiple factors, including the frequency of routine HIV testing in patients with SMI and the different sampling strategies used during the study.^53,54 Although HIV comorbidity has been associated with a longer LOA,^15,16,17,49 the results of this study did not correlate with this association. It is possible that virological suppression with advances in HIV treatment may have affected this finding, as individuals who are virologically suppressed with higher Cluster of Differentiation 4 (CD4) counts have fewer and shorter LOAs.⁵⁵

Personality disorders were the most common psychiatric comorbidity found in this study; the findings are in consistent with previous literature reporting a higher incidence of borderline and narcissistic personality disorders.⁵⁶ Although personality disorders did not affect the LOA in the present study, other studies have demonstrated that the patients with comorbid personality disorders often experience a more severe course of illness.⁵⁷

Cannabis was the most used substance, reported by one-third of patients, followed by crystal methamphetamine and alcohol. Although cannabis was the most frequently used substance, it was not a significant predictor of LOA (Table 3). This finding is relevant because recent South African literature reports that cannabis use may be increasing among patients admitted with SMI following the decriminalisation of home cannabis use in 2018.^23,24 Cannabis use has been associated with several adverse mental health outcomes, including poor treatment adherence and multiple readmissions.⁵⁸ However, LOA for a single admission may not capture a clear picture of how cannabis use may impact healthcare costs, as other factors such as the number of admissions, cost of medication and the need for frequent reviews or investigations for monitoring patients, also contribute to overall healthcare expenditure.⁵⁹ It is notable that all patients with multiple readmissions used cannabis, resulting in these patients having a net longer duration spent in the hospital. These findings highlight the need for further research with a larger sample size over an extended duration of study period.

Although the literature examines LOA in patients with SMI in South Africa, there is a paucity of literature specifically examining patients with BD. The findings of this study are similar to those of South African and other low- to middle-income countries and add to the body of knowledge on this topic.¹⁹ This study lays the foundation for exploring the correlations, such as how demographic factors, comorbidities, and substance use can affect LOA. Future research can build on this data to explore intervention strategies that may improve treatment outcomes after establishing the LOA determinants.

Although limited evidence exists regarding LOA in the South African patients with BD, to the best of our knowledge, this study is the first to examine cannabis use and its impact on LOA in this population. Cannabis was the most commonly used substance in this study; however, its use did not affect the LOA. South Africa is a resource-constrained country that requires targeted interventions to reduce LOA. These results do not suggest that reducing cannabis use decreases the LOA in patients with BD. However, this does not mean that decreasing cannabis use would not benefit patients and the healthcare system. Standardised treatment protocols, unmeasured socio-economic factors, medication adherence and comorbid substance use may have confounded the results. Additionally, episode severity and hospital resource constraints could also have influenced the LOA independent of cannabis use. Although these variables may have been confounding factors in this study, the research remains valuable in that it guides future research by identifying patients, diagnostic and treatment facility factors that warrant further exploration as potential factors influencing LOA in patients admitted with BD.

A higher proportion of women with BD were admitted compared to other studies.^19,39,40 This could be because of the increased screening of women with SMI at the CHBAH maternal mental health clinic. It is possible that the high number of female patients admitted with BD was because of the maternal mental health clinic improving screening and detection of women needing admission; this may speak to a lower threshold for admission in women, increased prevalence of women with BD or speak to women with BD having high risk of needing admission, to name a few reasons. This highlights a need for further research examining the reasons for admission of women with BD in this setting.

Also, it is noteworthy that all patients with multiple admissions during the study period had a history of substance use. This suggests a need for further research to evaluate the impact of cannabis use on readmission rates and the cumulative duration of hospitalisation throughout the lifespan among patients with BD.