Introduction
Psychotic disorders are a major cause of morbidity and premature mortality affecting approximately 3% of the general population (van Os et al., Reference Van Os, Linscott, Myin-Germeys, Delespaul and Krabbendam2009). They are associated with a significant public health burden worldwide (Knapp et al., Reference Knapp, Mangalore and Simon2004) with approximately half of the costs attributable to hospitalisation (Sledge et al., Reference Sledge, Tebes, Wolff and Helminiak1996; Kennedy et al., Reference Kennedy, Altar, Taylor, Degtiar and Hornberger2014).
Even though hospitalisation for psychosis has been a common outcome measure in longitudinal studies for the past 40 years, it remains unclear how many patients require hospital admission in the years after FEP. Some studies have reported that 30% or fewer patients with FEP are hospitalised at least once during their illness course (Stirling et al., Reference Stirling, White, Lewis, Hopkins, Tantam, Huddy and Montague2003; Ucok et al., Reference Ucok, Polat, Cakir and Genc2006; Salem et al., Reference Salem, Moselhy, Attia and Yousef2009) while others found that as many as 90% required hospital care after their first contact with mental health services (Berg et al., Reference Berg, Lindelius, Petterson and Salum1983; Lehtinen et al., Reference Lehtinen, Aaltonen, Koffert, Rakkolainen and Syvalahti2000). Similarly, wide variations in the length of stay (LOS) in psychiatric inpatient units have been reported with average durations ranging from 20 days to 740 days (Turner et al., Reference Turner, Boden, Smith-Hamel and Mulder2009; Fraguas et al., Reference Fraguas, Del Rey-Mejías, Moreno, Castro-Fornieles, Graell, Otero, Gonzalez-Pinto, Moreno, Baeza, Martínez-Cengotitabengoa, Arango and Parellada2014). Methodological variations accounting for some of this heterogeneity preclude the development of a reliable picture of hospital use in patients after FEP (Eaton et al., Reference Eaton, Mortensen, Herrman, Freeman, Bilker, Burgess and Wooff1992). There is also the question of the generalisability of studies as a large proportion were conducted in high-income counties (Saxena et al., Reference Saxena, Paraje, Sharan, Karam and Sadana2006a, Reference Saxena, Sharan, Garrido and Saraceno2006b; Patel et al., Reference Patel, Araya, Chatterjee, Chisholm, Cohen, De Silva, Hosman, McGuire, Rojas and van Ommeren2007). Thus, the current depiction of illness course is driven by findings obtained in the countries that are known for superior health-care rather than being globally representative.
It is important to provide unbiased and generalisable estimates of how many FEP cases will require hospitalisation after their first contact with services and of the time they will spend in inpatient care during their illness course. This will contribute to a better understanding of treatment needs for these individuals and aid service development and planning (Friis et al., Reference Friis, Melle, Johannessen, Røssberg, Barder, Evensen, Haahr, Ten Velden Hegelstad, Joa, Langeveld, Larsen, Opjordsmoen, Rund, Simonsen, Vaglum and McGlashan2016). It is equally important to identify moderating factors for these outcomes which may help to identify those FEP cases which may be at greater risk of poor long-term outcomes (Friis et al., Reference Friis, Melle, Johannessen, Røssberg, Barder, Evensen, Haahr, Ten Velden Hegelstad, Joa, Langeveld, Larsen, Opjordsmoen, Rund, Simonsen, Vaglum and McGlashan2016; Lally and Gaughran, Reference Lally and Gaughran2018). However, no previous study has conducted a meta-analysis incorporating global data and considered the moderators of hospital admission and LOS with meta-regression, which may identify important variables that influence these outcomes.
Therefore, the aims of the study were to conduct a systematic review and meta-analysis of all longitudinal studies that investigated the proportion of people with FEP who were hospitalised at least once during follow-up and/or reported average LOS during this period; and further to identify the moderators for these outcomes. Given the drive to reduce LOS and hospital admission, we hypothesised that the number of patients who required inpatient care, and the average LOS during follow-up would be significantly lower in the studies conducted in the last 20 years compared to earlier studies.
Methods
This systematic review was conducted and reported according to the Meta-analysis of Observational Studies in Epidemiology guidelines (Stroup et al., Reference Stroup, Berlin, Morton, Olkin, Williamson, Rennie, Moher, Becker, Sipe and Thacker2000) and the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) (Moher et al., Reference Moher, Liberati, Tetzlaff and Altman2009).
Search strategy
Two independent authors (JL, OA) searched PubMed, Medline, and Scopus without language restrictions from database inception to 1 October 2017. Key words used were (‘first episode psychosis’ OR ‘early episode psychosis’ OR ‘FEP’ OR ‘schizophrenia’ OR ‘schiz*’) AND (‘admission’ OR ‘hospitalisation’ OR ‘hospitalization’ OR ‘hospital*’ AND ‘outcome’ OR ‘follow-up’). A manual search of the reference lists of the retrieved articles was conducted.
Articles were initially screened based on title and abstract. The full texts of potentially eligible articles were independently inspected by two of the authors (O.A., J.L.). When data were incomplete, the corresponding author was contacted and invited to send additional information. When studies reported on overlapping samples, details of the study with the longest follow-up were included. If this was unclear, studies with the largest study sample for each respective outcome were included. We included multi-site studies and retained data for the entire cohort and not for individual sites.
Inclusion and exclusion criteria
We included longitudinal studies, incorporating both retrospective and prospective study designs, which were conducted in patients with FEP (including first episode schizophrenia and first episode affective psychosis) irrespective of clinical setting (i.e. inpatient, outpatient or mixed) that fulfilled the following criteria: (1) studies reporting the (a) proportion of patients who were hospitalised at least once during the follow-up period; and (b) average LOS in psychiatric hospitals during the entire follow-up period; (2) studies including individuals with FEP who were making their first contact with mental health services for psychosis; (3) studies using a specified standardised diagnostic system (e.g. International Classification of Diseases (ICD versions 8, 9 and 10), Diagnostic and Statistical Manual of Mental Disorders (DSM versions III and IV), and the Research Diagnostic Criteria (RDC); (4) studies with a follow-up period ⩾12 months; and (5) English language articles published in peer-reviewed journals.
We excluded studies if they: (1) were Randomised Control Trials, due to the potential that any structured intervention beyond routine care could influence the primary outcomes outlined in this meta-analysis; (2) assessed the feasibility and effectiveness of different treatment strategies for psychotic disorders; (3) were of organic psychosis due to medical conditions (i.e. psychosis secondary to medical condition, such as encephalitis or epilepsy) or non-FEP cohorts; and (4) did not report quantitative data;
Data extraction
Three authors (J.L., O.A., E.F.) extracted all data using a predetermined data extraction form and any inconsistencies were resolved by consensus. The data extracted included first author, study participant details, including mean age (years) at illness onset and first contact with mental health services, gender, country, setting [i.e. inpatient, outpatients (community), mixed, in- and out-patient settings], population, study design (i.e. prospective, retrospective), diagnostic classification method, assessment type, economic income status of the countries, duration of untreated psychosis (DUP), socio-demographic characteristics of the sample at the time of recruitment (i.e. proportion of patients who were employed, single or in a stable relationship at the study entry), baseline psychotic symptoms (mean scores), length of study follow-up, attrition, proportion hospitalised and average LOS, the proportion of patients who were taking antipsychotic medications at the study entry and at the end of follow-up, compliance with antipsychotic medications during the follow-up period, and socio-demographic characteristics at the end of follow-up (i.e. proportion of patients who were employed, single or in a stable relationship at the end of the follow-up period). A more detailed definition of these variables is provided in online Supplementary Materials.
Definitions of outcomes
The co-primary outcomes were:
(1) the proportion of people with FEP who were hospitalised at least once during the follow-up period (excluding any hospitalisation which occurred during the first contact for FEP)
(2) the average LOS in psychiatric hospitals defined as the average (mean and the standard deviation measured in days) time spent in hospital during the follow-up period excluding any hospitalisation which occurred during the first contact for FEP.
Statistical analysis
All analysis was conducted with Comprehensive Meta-Analysis software (CMA, Version 3) and RStudio version 3.4.4 (Integrated Development for R. RStudio, Inc., Boston). The pooled prevalence of hospitalisation and average LOS was calculated using a random-effects model (Borenstein et al., Reference Borenstein, Hedges, Higgins and Rothstein2010). The random-effects model was chosen to account for the influence of the context of care on these outcomes. To examine potential effects of specific factors on the primary outcomes, we further stratified these analyses according to: (1) baseline diagnosis, (2) assessment types; (3) length of follow-up; (4) study region; (5) study settings, and (6) economic income status of the country in which the study was conducted. The summary statistics were illustrated with a forest plot and funnel plot (Duval and Tweedie, Reference Duval and Tweedie2000; Phan et al., Reference Phan, Xie, Di Eusanio and Yan2014).
To investigate the variables that may influence the outcomes we conducted an unrestricted maximum likelihood meta-regression. The included moderating factors were age at illness onset, age at first contact with mental health services, male gender, ethnicity, baseline psychotic symptoms (mean scores), relationship and employment status at baseline, DUP, duration of follow-up, attrition rate, study year, treatment with antipsychotic medications at baseline and during follow-up, and compliance with antipsychotic medications during the entire follow-up period.
Publication bias was assessed with the funnel plot, Egger regression test (Opjordsmoen et al., Reference Opjordsmoen, Friis, Melle, Haahr, Johannessen, Larsen, Røssberg, Rund, Simonsen, Vaglum and McGlashan2010). We also adjusted for the presence of any publication bias calculating the Duval and Tweedie ‘trim-and-fill’ method (Tohen et al., Reference Tohen, Tsuang and Goodwin1992). Heterogeneity was measured with the Q statistic yielding a χ2 and p value, and the I 2 statistic with scores above 50 and 75% indicating moderate and high heterogeneity, respectively (Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003). Statistical significance was considered to be at or below the 0.05 level.
Results
Search results and included participants
The flowchart of the article selection process is depicted in Fig. 1 and descriptive characteristics of each study are outlined in online Supplementary Table S1. The search yielded 1434 non-duplicated publications, which were considered at the title and abstract level; 382 of these were extracted for full-text review, of which 81 met the inclusion criteria with a total sample of 23 280 FEP patients (range = 20–12 071). The mean age at illness onset in these studies was 23.5 years (s.d. = 5.7), while mean age at first contact with mental health services was 27.3 year (s.d. = 64); 42.3% were female and 59.3% had a baseline diagnosis of first episode schizophrenia.
The flowchart of the article selection process in the meta-analysis of hospitalisation and length of hospital stay during follow-up in patients with first-episode psychosis (FEP).
Meta-analysis of hospitalisation
The proportion of people with FEP who were hospitalised at least once during the follow-up, together with heterogeneity and trim-and-fill analyses, is presented in Table 1. In total, 60 studies reported on the number of people with FEP who were hospitalised at least once during the follow-up period. Average length of follow-up across these studies was 7.6 years (s.d. = 6.1, interquartile range (IQR) = 2–11.8). The total sample at the end of the follow-up period was 19 675 FEP cases (range = 20–12 071, IQR = 47–149). The pooled proportion of hospitalised FEP patients during follow-up was 55.4% (95% CI 50.3–60.5, Q = 3575.1, I 2 = 98.5). The Begg-Mazumdar (Kendall's τ b = −0.005, p = 0.957) and Egger test (t = −2.53, df = 56, p = 0.014) indicated no publication bias.
Meta-analysis of the proportion of patients with first-episode psychosis who were hospitalised at least once during a follow-up period
n, number; FEP, first episode psychosis; FU, follow up period; FEAP, first episode affective psychosis; N/A, not appropriate; CI, confidence intervals. Bold values indicate statistically significant associations.
Subgroup analyses of hospitalisation
Stratified proportions of FEP patients who were hospitalised at least once during the follow-up period, together with heterogeneity and trim-and-fill analyses are presented in Table 1. The proportion of patients hospitalised during follow-up was significantly higher in studies from Australia and New Zealand (78.4%, 95% CI 59.2–97.5, I 2 = 98.4, Q = 203.7) compared to studies from Europe (58.1%, 95% CI 50.7–65.5, I 2 = 97.1, Q = 1212.1) and North America (48.0%, 95% CI 34.5–61.6, I 2 = 95.4, Q = 213.6); the lowest proportion of hospitalised patients was reported in studies from Asia (32.5%, 95% CI 25.3–41.4, I 2 = 81.4, Q = 39.7). The pooled proportion of hospitalised patients during follow-up was highest in studies which were conducted in high-income countries (57.9%, 95% CI 51.7–64.1, I 2 = 98.4, Q = 2833.6) compared with studies conducted in middle-income countries (34.8%, 95% CI 20.0–49.6, I 2 = 96.0, Q = 355.8). The trim-and-fill method demonstrated that the proportion of patients who required hospitalisation at least once during the follow-up period in the middle-income countries was 42.9% (95% CI 27.4–56.5) when adjusted for potentially missing studies. There were no studies from low-income countries.
Effect of moderator variables influencing hospitalisation
Full details of the moderators of hospitalisation during the follow-up period are presented in Table 2. A lower proportion of hospitalised patients during follow-up was associated with an older age of illness onset (β = −0.049, 95% CI −0.092 to −0.005, p = 0.028, R 2 = 0.07) and having a stable relationship at baseline (β = −0.011, 95% CI −0.018 to −0.004, p = 0.004, R 2 = 0.33). There was a trend association between Black ethnicity and increased hospitalisation (β = 0.004, 95% CI 0.000–0.009, p = 0.075, R 2 = 0.13), and between higher loss to attrition and reduced hospitalisation during follow-up (β = −0.003, 95% CI −0.007 to 0.000, p = 0.080, R 2 = 0.04).
Meta-regression of moderators of the proportion of patients with first-episode psychosis who were hospitalised at least once during a follow-up
DUP, duration of untreated psychosis; β, beta coefficient; CI, confidence intervals. Bold values indicate statistically significant associations.
Meta-analysis of LOS
Average LOS across the follow-up period with heterogeneity and trim-and-fill analyses is provided in Table 3. In total, 37 studies reported on LOS over the follow-up period. The average LOS was 176.8 days (s.d. = 186.7, median = 106 days, IQR = 76–204 days). Average length of follow-up across these studies was 7 years (mean = 6.6 years, s.d. = 6.4, IQR 2–8) with a cumulative sample of 4877 FEP cases (range = 20–720, IQR = 43.5–191.5). The pooled average LOS across the entire follow-up period was 116.7 days (95% CI 95.1–138.3, I 2 = 99.5, Q = 4435.1). The Begg-Mazumdar (Kendall's τ b = 0.18, p = 0.215) and Egger test (t = 4.31, df = 24, p < 0.001) indicated no publication bias.
Meta-analysis of the length of inpatient stays during a follow-up in patients with first-episode psychosis
n, number; FEP, first episode psychosis; FU, follow up period; FEAP, first episode affective psychosis; N/A, not appropriate; CI, confidence intervals.
Subgroup analyses of LOS
Stratified LOS during the follow-up period with heterogeneity and trim-and-fill analyses is provided in Table 3. The LOS was the longest in studies published from 1966–1995 (192.3 days, 95% CI 129.7–254.8, I 2 = 89.2, Q = 37.1). The trim-and-fill method demonstrated that the average LOS in these studies was 216.8 days (95% CI 126.3–307.3) when adjusted for missing studies. The mean LOS appeared to decrease in more recent studies from 1996–2002 (129.9 days, 95% CI 78.8–180.9, I 2 = 98.9, Q = 368.1) and 2003–2009 (97.7 days, 95% CI 55.3–139.9, I 2 = 99.8, Q = 3041.4). The shortest average LOS was recorded in studies from 2010–2017 (96.6 days, 95% CI 54.0–139.2, I 2 = 99.3, Q = 852.8).
Effect of moderator variables influencing LOS
Information on the moderators of LOS is presented in Table 4. The meta-regression analyses showed that a longer LOS was associated with Black ethnicity (β = 2.905, 95% CI 1.273–4.537, p < 0.001, R 2 = 0.14%) and longer DUP (mediandays) (β = 0.303, 95% CI 0.266–0.340, p < 0.001, R 2 = 0.11). Another significant moderator of a longer mean LOS was a longer length of follow-up (β = 11.707, 95% CI 6.577–16.838, p < 0.001, R 2 = 0.21). Several baseline factors associated with shorter average LOS were identified. A shorter average LOS was associated with White ethnicity (β = −0.181, 95% CI −0.219 to −0.143, p < 0.001, R 2 = 0.12), reduced severity of psychotic symptoms at baseline (β = −0.019, 95% CI −0.036 to −0.003, p = 0.018, R 2 = 0.08) and studies conducted in more recent years (β = −4.413, 95% CI −7.456 to −1.370, p = 0.004, R 2 = 0.15).
Meta-regression of moderators of an average length of inpatient stay during a follow-up during a follow up in patients with first-episode psychosis
DUP, duration of untreated psychosis; β, beta coefficient; CI, confidence intervals. Bold values indicate statistically significant associations.
Discussion
To our knowledge, this is the first systematic review and meta-analysis to investigate the proportion of FEP cases which required hospitalisation at least once after their first contact with mental health services and the average LOS in a hospital during follow-up. We found that more than half (55%) of all FEP patients required hospitalisation over an average follow-up of 7 years after FEP. This proportion may seem high, but is not surprising considering that only 38% of FEP patients recover during follow-up (Lally et al., Reference Lally, Ajnakina, Stubbs, Cullinane, Murphy, Gaughran and Murray2017), with 34% of FES patients meeting criteria for treatment resistance over a 5-year period (Lally et al., Reference Lally, Ajnakina, Di Forti, Trotta, Demjaha, Kolliakou, Mondelli, Reis Marques, Pariante, Dazzan, Shergil, Howes, David, MacCabe, Gaughran and Murray2016).
Hospitalisation and average LOS in FEP patients
While bed capacity in psychiatric services has decreased in many developed countries since the 1950s (Raftery, Reference Raftery1992) supported by intensive attempts to integrate and care for people in the community (Munk-jorgensen, Reference Munk-jorgensen1999), our findings demonstrate that the proportion of people with FEP who were admitted to the hospital after their first contact with mental health services has remained stable over time. Nonetheless, in accordance with previous research (Agius et al., Reference Agius, Hadjinicolaou, Ramkisson, Shah, Haq, Tomenson and Zaman2010; Hobbs et al., Reference Hobbs, Tennant, Rosen, Newton, Lapsley, Tribe and Brown2000; Leff and Trieman, Reference Leff and Trieman2000) we found that the average LOS in hospital for people FEP has decreased considerably over the past 20 years with the sharpest reduction observed in the last 7 years. This pattern was particularly pronounced in Australia and New Zealand. Our findings may indicate that while early intervention services for psychosis are successful in facilitating earlier discharge from hospital (Agius et al., Reference Agius, Hadjinicolaou, Ramkisson, Shah, Haq, Tomenson and Zaman2010); the sustained high proportion who require inpatient care over the illness course questions whether they are able to reduce the need for hospital admissions. Recent observational data indicate the benefits of antipsychotic long-acting injections and clozapine in reducing the need for hospitalisation in psychotic disorders (Tiihonen et al., Reference Tiihonen, Mittendorfer-Rutz, Majak, Mehtälä, Hoti, Jedenius, Enkusson, Leval, Sermon, Tanskanen and Taipale2017). Although we did not investigate the impact of antipsychotic long-acting injections and clozapine in reducing the need for hospitalisation in psychotic disorders in the present study, their wider use may be one route to reducing the sustained rates of hospitalisation identified in our study.
We found that the number of cases who were hospitalised at least once during follow-up did not differ significantly depending on the length of follow-up. Hospitalisation is considered an indicator of poor outcome in FEP (Lieberman et al., Reference Lieberman, Wiitala, Elliott, Mccormick and Goyette1998; Schoeler et al., Reference Schoeler, Petros, Di Forti, Klamerus, Foglia, Murray and Bhattacharyya2017) because it is costly and occurs when the illness becomes severe enough to warrant such an intervention (Pottick et al., Reference Pottick, Mcalpine and Andelman2000). Accordingly, it may be argued that the longitudinal illness trajectory of psychosis is not characterised by a deteriorating course for most patients (Zipursky and Agid, Reference Zipursky and Agid2015) as previously thought (Schmidt et al., Reference Schmidt, Blanz, Dippe, Koppe and Lay1995; Ropcke and Eggers, Reference Ropcke and Eggers2005). This is consistent with what was observed in relation to longitudinal recovery rates in patients with FEP where no evidence for worsening recovery rates with longer duration of follow up was found (Lally et al., Reference Lally, Ajnakina, Stubbs, Cullinane, Murphy, Gaughran and Murray2017).
We found that the proportion of patients hospitalised during follow-up was considerably higher in high-income compared to the middle-income countries. Although this might imply a more debilitating illness course in well-developed countries (Lin and Kleinman, Reference Lin and Kleinman1988), it could also be explained by differences in social support and family support structures and quality of mental health-care in middle-income countries where the burden of care and treatment costs tend to fall on families rather than hospitals (Saxena et al., Reference Saxena, Sharan, Garrido and Saraceno2006b; Patel et al., Reference Patel, Araya, Chatterjee, Chisholm, Cohen, De Silva, Hosman, McGuire, Rojas and van Ommeren2007).
Impact of moderator variables on hospitalisation and LOS in FEP patients
The reasons for hospitalisation are complex (Schoeler et al., Reference Schoeler, Petros, Di Forti, Klamerus, Foglia, Murray and Bhattacharyya2017) and likely to be explained by a range of clinical and social factors. Medication adherence was shown to be an important determinant for hospitalisation in patients with FEP during an 18-month follow-up study (Sfetcu et al., Reference Sfetcu, Musat, Haaramo, Ciutan, Scintee, Vladescu, Wahlbeck and Katschnig2017). However, this finding is not supported by studies with a longer follow-up period (Friis et al., Reference Friis, Melle, Johannessen, Røssberg, Barder, Evensen, Haahr, Ten Velden Hegelstad, Joa, Langeveld, Larsen, Opjordsmoen, Rund, Simonsen, Vaglum and McGlashan2016) including the present work. Comparable to previous reports (Immonen et al., Reference Immonen, Jaaskelainen, Korpela and Miettunen2017; Melle et al., Reference Melle, Friis, Hauff and Vaglum2000; Uggerby et al., Reference Uggerby, Nielsen, Correll and Nielsen2011), which identified an association between a younger age of illness onset and increased hospitalisation, we found that an older age of illness onset was associated with reduced hospitalisation, though it was not a significant moderator for LOS. Consistent with previous literature highlighting associations between DUP and poorer outcomes in patients with psychosis (Harrigan et al., Reference Harrigan, Mcgorry and Krstev2003; Schimmelmann et al., Reference Schimmelmann, Huber, Lambert, Cotton, Mcgorry and Conus2008), we found that longer DUP was significantly associated with longer LOS. It has been shown that the mode of onset of first psychotic symptoms is one of the strongest predictors of the duration of DUP (Compton et al., Reference Compton, Chien, Leiner, Goulding and Weiss2008), with an insidious onset of psychotic symptoms associated with a longer DUP (Morgan et al., Reference Morgan, Abdul-Al, Lappin, Jones, Fearon, Leese, Croudace, Morgan, Dazzan, Craig, Leff and Murray2006; Ajnakina et al., Reference Ajnakina, Lally, Di Forti, Kolliakou, Gardner-Sood, Lopez-Morinigo, Dazzan, Pariante, Mondelli, MacCabe, David, Gaughran, Murray and Vassos2017a, Reference Ajnakina, Morgan, Gayer-Anderson, Oduola, Bourque, Bramley, Williamson, MacCabe, Dazzan, Murray and David2017b). The type of clinical or non-clinical service with whom the first contact is made following the onset of psychosis has been shown to be another important factor in determining the length of DUP (Bechard-Evans et al., Reference Bechard-Evans, Schmitz, Abadi, Joober, King and Malla2007; Tanskanen et al., Reference Tanskanen, Moran, Hinton, Lloyd-Evans, Crosby, Killaspy, Raine, Pilling and Johnson2011; Ghali et al., Reference Ghali, Fisher, Joyce, Major, Hobbs, Soni, Chisholm, Rahaman, Papada, Lawrence, Bloy, Marlowe, Aitchison, Power and Johnson2013). Thus, multiple clinical and service level factors, as well as social factors, are related to DUP, these need to be examined in more detail to ascertain the best ways to reduce the length of DUP, and potentially LOS.
In the present study, Black ethnicity appeared to be an important factor influencing hospitalisation and was associated with longer inpatient care during follow-up. Black ethnicity has consistently been highlighted as a risk factor for psychosis onset (Lally et al., Reference Lally, Ajnakina, Di Forti, Trotta, Demjaha, Kolliakou, Mondelli, Reis Marques, Pariante, Dazzan, Shergil, Howes, David, MacCabe, Gaughran and Murray2016; Radua et al., Reference Radua, Ramella-Cravaro, Ioannidis, Reichenberg, Phiphopthatsanee, Amir, Yenn Thoo, Oliver, Davies, Morgan, McGuire, Murray and Fusar-Poli2018), and has been associated with the development of a treatment-resistant course of illness (Lally et al., Reference Lally, Ajnakina, Di Forti, Trotta, Demjaha, Kolliakou, Mondelli, Reis Marques, Pariante, Dazzan, Shergil, Howes, David, MacCabe, Gaughran and Murray2016). Evidence is emerging from the UK that the longitudinal trajectory of psychosis in patients of Black ethnicity is characterised by more extensive utilisation of psychiatric services compared with patients of White British ethnicity (Morgan et al., Reference Morgan, Lappin, Heslin, Donoghue, Lomas, Reininghaus, Onyejiaka, Croudace, Jones, Murray, Fearon, Doody and Dazzan2014; Ajnakina et al., Reference Ajnakina, Lally, Di Forti, Kolliakou, Gardner-Sood, Lopez-Morinigo, Dazzan, Pariante, Mondelli, MacCabe, David, Gaughran, Murray and Vassos2017a, Reference Ajnakina, Morgan, Gayer-Anderson, Oduola, Bourque, Bramley, Williamson, MacCabe, Dazzan, Murray and David2017b), results supported by the present study. Considering that patients of Black ethnicity are also at risk of becoming increasingly socially disadvantaged as their illness progresses (Morgan et al., Reference Morgan, Lappin, Heslin, Donoghue, Lomas, Reininghaus, Onyejiaka, Croudace, Jones, Murray, Fearon, Doody and Dazzan2014; Ajnakina et al., Reference Ajnakina, Lally, Di Forti, Kolliakou, Gardner-Sood, Lopez-Morinigo, Dazzan, Pariante, Mondelli, MacCabe, David, Gaughran, Murray and Vassos2017a, Reference Ajnakina, Morgan, Gayer-Anderson, Oduola, Bourque, Bramley, Williamson, MacCabe, Dazzan, Murray and David2017b), the need for hospitalisation and prolonged inpatient stays in those of Black ethnicity observed in the present study may be related to social adversity.
We found that being in a stable relationship at the time of the first contact with FEP was associated with a reduced proportion of hospitalisation during follow-up. Being in a stable relationship may constitute improved social integration and strong social networks that have been shown to be associated with improved outcomes in FEP (Erickson et al., Reference Erickson, Beiser, Iacono, Fleming and Lin1989). However, as it is common for individuals with psychosis to struggle to develop or maintain stable relationships (Sundermann et al., Reference Sundermann, Onwumere, Kane, Morgan and Kuipers2014), the protective effect of this factor may only be available to a minority of patients. Alternatively, being in a stable relationship may be indicative of a preserved premorbid level of functioning, improved outcomes and reduced hospitalisations during follow-up.
Methodological considerations
This is the first meta-analysis to investigate the proportion of FEP patients who required hospitalisation at least once after their first contact with mental health services and the average LOS in the hospital during the entire follow-up period. We examined the proportion of hospitalised patients and average LOS during follow-up separately for baseline diagnosis of FEP, first episode schizophrenia and first episode affective disorders. Stratification by diagnosis allowed us to capture the most representative trajectory of illness for these diagnostic categories. Focusing on the incident sample of patients with the first presentation to services for psychosis ensured that the findings are not biased by chronicity of illness.
Notwithstanding the strengths, there are several limitations to the data and meta-analysis that warrant discussion. While we identified studies from five regions of the world, there was marked variability in the number of studies from each region, with the majority conducted in Europe. We were unable to eliminate confounding variables relating to group differences in FEP cases that were enrolled in the different regions, and other service level confounds which may have existed between regions. This may include the variability in criteria employed that would warrant hospitalisation or prolonged hospital stays, bed availability, accessibility of community mental health services, treatment received in the community and in hospital, availability of community social supports, local mental health laws relating to involuntary hospital admission or other legal frameworks. Evidence suggests that all of these factors tend to vary between countries and regions (Burti, Reference Burti2001; Saxena et al., Reference Saxena, Paraje, Sharan, Karam and Sadana2006a, Reference Saxena, Sharan, Garrido and Saraceno2006b; Tulloch et al., Reference Tulloch, Fearon and David2012) and as such may have influenced hospitalisation and LOS across populations and studies included in the present meta-analysis. This should be taken into consideration when interpreting the study findings. Although it may be argued that studies utilising data from case notes may not have provided a reliable depiction of the clinical course of psychosis (Eaton et al., Reference Eaton, Mortensen, Herrman, Freeman, Bilker, Burgess and Wooff1992), in the present study we found that hospitalisation and average LOS did not significantly differ depending on sources of data ascertainment. In relation to the meta-regression analyses, some of the variables might have failed to achieve statistical significance because of a lack of power due to small sample sizes. Further, we did not obtain data on important confounders such as types of treatments received or services available, lifestyle factors such as substance use, and symptom profile over the course of follow-up precluding the meta-analytic assessment of these factors as moderating and/or mediating variables. By excluding hospitalisations that occurred at the first contact with mental health services for FEP we may have omitted a small proportion of severely ill patients who might have remained hospitalised for most of the follow-up period. Finally, we were unable to establish the reasons for hospitalisation, whether it was the result of psychotic relapse, antipsychotic intolerance or a comorbid mental disorder.
Conclusion
This meta-analysis indicates that one in two patients with FEP will require hospitalisation at least once during a 7-year follow-up with an average inpatient stay of 4 months during this period. While the proportion of those with FEP who were admitted to hospital in the years following FEP has remained stable over the years, the average time FEP cases spent in hospital during follow-up has decreased in the last 20 years. This suggests that patients are now discharged earlier compared to previous time periods. While most patients and clinicians may favour shorter LOS in hospital, the question as to whether patients are discharged prematurely needs further investigation.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291719000904.
Author ORCIDs
Olesya Ajnakina, 0000-0003-3987-1236.
Financial support
This paper represents independent research funded by the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London. O.A. is funded by the National Institute for Health Research (NIHR) (NIHR Post-Doctoral Fellowship - PDF-2018-11-ST2-020) for this project. R.M.M. and A.S.D. receive salary support from the NIHR Maudsley BRC. BS is supported by Health Education England and the National Institute for Health Research HEE/NIHR ICA Programme Clinical Lectureship (ICA-CL-2017-03-001). FG and BS are part funded by the National Institute for Health Research Collaboration for Leadership in Applied Health Research & Care Funding scheme with support from the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust. FG also receives support from the Stanley Medical Research Institute). The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health and Social Care.
Conflict of interest
R.M.M. has received honoraria from Janssen, Astra-Zeneca, Lilly, and BMS. A.S.D. has received honoraria from Janssen and Roche Pharmaceuticals. F.G. has received honoraria for advisory work and lectures from Roche, BMS, Lundbeck, Otsuka and Sunovion and has a family member with professional links to Lilly and GSK. The other authors (O.A., B.S., J.L., E.F.) have no conflict of interest to declare.
