Since the 1950's mental health services have undergone important transformations (1, 2). These included the development of community mental health care and the closure of psychiatric asylums and hospitals, which in most cases were built in the previous century. Reforms encouraged both reductions of psychiatric bed numbers and length of stay for psychiatric hospitalization, and promoted long-stay housing facilities in the community (3). At present, the reforms are still ongoing in many places. Substantial parts of inpatient psychiatric services have been removed and services have shifted toward community care (4–6). Reforms have not established a minimum or optimal number of psychiatric beds in order to assure balanced mental health systems, and it remains under debate whether to continue to remove psychiatric beds (7–10). Furthermore, research is still scarce on how many beds have actually been removed in the context of reforms and how these changes relate to other types of institutionalization (8).

The rates of psychiatric beds differ between countries and geographical regions (11), especially between different income groups (12). According to the WHO Mental Health Atlas (2017), the median number of psychiatric beds per 100,000 population is around 50 in high income countries (HIC), opposed to a rate of 7 in low and middle-income countries (LMIC). Residential care beds are almost inexistent in LMIC, whereas a median of 23 residential care beds per 100,000 population were reported for HIC (11).

The definition of psychiatric beds used by the Mental Health Atlas Project (11) incorporates short-stay and long-stay beds in psychiatric hospitals, beds in general hospital psychiatric units (GHPU), inpatient psychiatric services based in community settings and forensic inpatient units. This includes public and private facilities, psychiatric beds only for children and adolescents and other specific groups such as older adults. The definition excludes beds, which are exclusively used for the treatment of individuals with intellectual disability or substance use disorders, as well as facilities that exclusively provide rehabilitation and recovery services. The exact definition of beds in long-stay facilities causes difficulties (10, 13). The concept of psychiatric beds has undergone substantial changes over time (14, 15). This global variety of definitions limits international comparisons that aim to develop and optimize services (10, 16–18). A review assessing mental health plans of five English-speaking HIC showed that even in regions with the same official language, recommendations used variable nomenclature and mostly promoted a mix of inpatient and community services (16). It showed the need to define core mental health service components including specific resource targets in order to deliver more strategic clinical care.

Three different approaches have been proposed for estimating the required number of psychiatric beds (18). First, an empirical population health approach that estimates or calculates current and future psychiatric bed requirements for a specific catchment area based on epidemiological data considering current provision and quality of care (19). Secondly, expert consensus has been conducted in developed countries (20–22). Thirdly, a normative approach that assumes that different catchment areas with similar mental health and demographic profiles may require a similar number of psychiatric beds, so a well-functioning mental health system can be used as a model for other similar areas (18, 19).

We conducted a systematic review on expert opinions, arguments and recommendations for trends of psychiatric bed numbers.

The following seven databases were searched from their inception until December 27, 2020: PubMed, Embase Classic and Embase, PsycINFO and PsycIndex, Open Gray, Google Scholar, Global Health EBSCO and Proquest Dissertations. The search term used was “psychiatric AND hospital^* AND bed^*” with no filters based on study types. We did not use any language restrictions. Since Google Scholar produces very high numbers of hits (>500,000) and sorts them by relevance, the search was limited to the first 561 hits. We restricted searches in PubMed, PsycINFO and PsycIndex to title and abstract. This helped us to assure that the psychiatric beds were a central topic of the articles rather than tangentially discussed. References and citations of articles retained in this study were reviewed for additional unidentified studies.

We included studies that presented arguments, opinions, recommendations and suggestions for trends of psychiatric bed numbers. We excluded studies that only reported specific numbers of psychiatric beds. Studies providing purely numerical recommendations were excluded from this paper. Publications referring to beds or places in other mental health facilities such as residential facilities or day hospitals were also excluded.

Literature screening was conducted by SD, Dr. ERS, and Dr. MS. Data were extracted independently by SD, Dr. ERS, and Dr. MS.

The following variables were extracted: year of publication, whether the expert recommendation referred to a local, national, regional or global area, the income group of the country for which the recommendation was made according to the World Bank classification (12), author's hypothesis or argument for the proposed change and recommendations. Multiple arguments were extracted, if present in the same publications.

The respective recommendations were a priori classified whether they dealt with acute care and short-stay on the one hand and long-stay on the other hand. When short- and medium-stay beds were reported in an aggregated format, they were classified as short-stay and when medium to long-stay beds were reported in an aggregated way, they were classified as long-stay. Furthermore, recommendations were grouped by suggestion to either reduce and maintain or increase psychiatric bed numbers. The arguments were analyzed using a thematic analysis approach with six steps (23), which is a qualitative analytic method used to identify, analyze and report codes, subthemes and themes for qualitative data. We used a theoretical rather than inductive thematic analyses approach (23), since the themes were sorted into the overarching a priori defined groups “reduce” or “increase- or maintain” psychiatric bed numbers. The themes were identified on a latent rather than semantic level (23). Findings from different publications were integrated in order to achieve an interpretation of the arguments that resulted in a higher level of scientific evidence, consistent with meta-synthesis (24). After familiarization with the texts, an open coding was developed and initial codes were determined. Then initial codes were grouped into different categories according to their similarities. In the third step, these categories were organized into themes. This implied the combination of codes into wider and overarching themes that adequately reflect the extracted information. Three authors subsequently examined this classification reorganizing the data and recoding the categories in a reiterative process between researchers and ensuring that categories were understood in the same way by all researchers. In the fifth step, a comprehensive analysis of the different themes was conducted ensuring that each one contributed to a better understanding of the data. For each category, a summary of the main ideas was derived from all included quotes. In the sixth and final step, the report was written allowing for a detailed description of the results. To ensure robustness, quotes were included in the description of the results to illustrate the descriptions.

The number of quoted recommendations per country and category were analyzed.

Critical appraisal tools typically aim to assess the quality of a study in particular. As quoted recommendations and expert arguments were not necessarily identical to the main objective of the respective studies, the methodology and overall quality of the publication were not suitable to assess the quality of the opinion. Thus, by consensus between the authors, we did not rank or differentiate the quality of the included recommendations or arguments. Nonetheless, in order to ensure validity of the results several quality criteria were included. First, triangulation was implemented meaning three complementary researchers from different backgrounds (psychology and psychiatry) participated in the data analysis. Secondly, theoretical validation was used to compare results with the scientific literature. Lastly, an iterative process was conducted. When a new code was added, the codification was read again in order to ensure that the initial classification was accurate and the extracted data were complete.

We distinguished two main categories: (a) arguments to reduce psychiatric bed numbers, and (b) arguments to maintain or increase psychiatric bed numbers. In both categories, we identified themes and subthemes supporting each trend (Annex 1). We identified 36 quotes arguing for a decrease and 123 quotes supporting to increase or maintain psychiatric bed numbers.

Illustrative verbatim are given in Table 3. A thematic map was built in order to show the main themes that emerged from the data (Figures 2, 3), which are described in further detail below.

This systematic review on expert arguments regarding trends of psychiatric bed numbers was based on 106 publications from 25 countries, including 14 studies from LMICs. First, there was not any general agreement on the direction of trends that should be pursued in policies, as there were arguments for reductions as well as for increases of psychiatric bed rates. Secondly, a complex matrix of partially contradicting arguments emerged, i.e., with respect to costs and quality of care. In the absence of hard evidence, those arguments remain relevant and have to be carefully considered in specific contexts for local policies.

Mental health service planners need orientation on how to further develop the psychiatric inpatient sector (7–10). Despite the diversity of services and nomenclatures used across regions (10, 16–18), there was agreement on several themes, such as the need to develop integrated mental health systems that assure coherence and continuity of care, along with the need to aim for specific populations and contexts (127). In all, arguments expressing concern about further bed reductions prevailed as there is a high demand on inpatient services, especially on short-stay and acute beds, evidenced in high occupancy and increasing admission rates (10, 99). In the US and Australia, overcrowding and long waiting times in emergency departments showed to be particularly relevant (25, 27, 30, 43, 52, 60, 66, 119). The relevance of psychiatric beds for problems that arise outside of the health system, such as violent crime of people with mental illnesses, increased detention rates and discharge to homelessness due to a lack of timely treatment and admission, was considered in several HICs (36, 61, 85, 128). Criminalization of mentally ill consistently appeared as a relevant theme in the US (10, 70, 93, 119) and was also mentioned for Latin America (8, 105). Inadequately short lengths of stay emerged as an issue in HICs (73), and in South Africa among LMICs (62). Mental health professionals are forced to prematurely discharge patients in order to free already scarce beds risking short readmission intervals, referred to as the revolving door effect (99). Several authors have expressed their concern that there are gaps in the mental health system due to major reductions of inpatient bed capacities that have been implemented without the appropriate development of community care (52, 129). There was agreement for HICs and LMICs that insufficient provision of outpatient services may not compensate for more bed reductions, resulting in overcrowding of inpatient facilities, which subsequently compromises the quality care (99). Several authors therefore call to halt current trends to further remove beds. Within this context, limited post discharge support in the community showed to be an important point in the UK (14, 38, 50, 87), and the lack of specialized psychiatric beds for children and adolescents emerged as particularly relevant in the UK and the US. Other subspecialty beds, such as mother baby units, may also be lacking, especially in LMICs (130). This overall lack of beds compromises patient safety and quality of care, carrying hardships for patients and families, and may provoke serious incidents (97). Some arguments emerged for specific contexts, such as increasing suicide rates related to major bed reductions, especially in the US (93, 123), and the argument that financial pressures on the mental health system have resulted in too many bed removals in Greece and the UK (45, 49).

In regard to arguments supporting the further reduction of psychiatric beds, most authors referred to long-stay beds and to inappropriately long lengths of stay (71, 78, 79, 81, 83), emphasizing that patients could be transferred to the community if more timely and adequate after-care could be found (33, 39, 40, 48, 64). The paradigm emerged that successful development of effective community services, including housing, and a better integrated mental health system would allow to continue psychiatric bed removals without negative outcomes (4–6), especially considering that inpatient services can be restrictive environments (48). We observed that a further reduction of psychiatric beds was usually not recommended for LMICs, except for Moldova (88). Several authors identified a need to follow trends of psychiatric bed reductions, which were enforced in most developed countries. This was mentioned for Japan (83) and Korea (80). For New Zealand authors argued that inpatient psychiatric bed capacity and availability generates utilization and coercive treatments (57). Economic incentives for inadequately long inpatient bed use have been observed in Japan (116). In the UK, it was suggested that bed reductions lead to better use and development of existing community care, along with reduced reliance on such beds (100). Lastly, for the US it is suggested that hospital bed numbers should be reduced to serve the most severely ill patients (90).

Opinions were divided on themes, such as the cost-effectiveness of psychiatric bed reductions. On the one hand, the lower cost of community services compared to inpatient care allows for greater concentration of inpatient resources on the most severely ill. However, in 1998 an US-American study showed increased costs in the entire health system after bed reductions due to a higher use of acute care in general hospitals as a result of severely mentally ill patients queuing up in these facilities while waiting for transfers to intermediate care in the community (115). In regard to quality of care opinions were also divided. Lack of beds can lead to reduced quality of inpatient treatments and overcrowding. However, hospitals in the US have shown to maintain or increase their quality of care by removing beds while maintaining the staff (82).

Our study shows that an empirical approach to argue for trends of psychiatric bed numbers was most frequent, while the normative approach was least frequently used with no differences between HICs and LMICs. It has been argued that need of psychiatric beds may vary between regions (131), and even changes between seasons (132). The provision of psychiatric beds has to respond to local requirements and conditions (127). The arguments presented here may help to tailor policies after evaluation of fit and context-dependent applicability. One-dimensional models that simply focus on the numbers to address possible bed shortages may need to consider incorporating more complex aspects of the system, including costs, quality and pathways of care (133). At the level of catchment areas, the need of beds has to be evaluated assuring a continuum of care with quality treatment before, during and after an acute episode of mental illness (134).

To our knowledge this is the first systematic review of expert arguments on psychiatric bed numbers. We provide evidence and examined arguments from 25 countries worldwide, including 14 LMICs. This research has several limitations. The arguments in the present review were not necessarily identical to the main objective of the respective studies but were extracted from the discussions during our screening process. Secondly, the usage of variable nomenclatures referring to psychiatric beds is a further limitation. Thirdly, study quality was not assessed. Another limitation was that this review did not assess the opinions of patients, caregivers, families and other members of the community.

Several implications arise from our findings. First, there are relatively few arguments that are repeated in the literature and should be considered by discussants on the required number of psychiatric beds. This synthesis of arguments can help to focus future debate and to guide policymakers who need to define targets for the number of psychiatric beds in specific countries and catchment areas. Secondly, further research is needed to guide which arguments are best suited for specific contexts. Thirdly, there is need to establish targets for more specific populations (e.g., juvenile, older adult, forensic, acute and long-stay populations) and for specific types of facilities (public, private, mental or general hospital and residential). The low number of publications from LMICs highlights the need for further evidence from these countries, especially from regions underrepresented in research, such as Central and East Asia, Africa and Central America.