« Go back to issue 42(1) summary

Factors related to positive and negative outcomes in psychiatric inpatients in a General Hospital Psychiatric Unit: a proposal for an outcomes index

Hugo Karling Moreschi1, Gabriela Pavan1, Julia Almeida Godoy1, Rafael Mondrzak1, Mariana Ribeiro de Almeida2, Marco Antônio Pacheco1, Eduardo Lopes Nogueira1, Lucas Spanemberg1,2

1 Department of Psychiatry, Hospital São Lucas, Pontifícia Universidade Católica do Rio Grande do Sul (PUC-RS), Porto Alegre, RS, Brazil.

2 Department of Psychiatry, Hospital de Clínicas de Porto Alegre, Universidade Federal do Rio Grande do Sul (HCPA/UFRGS), Porto Alegre, RS, Brazil.

Received: 11/29/2014 – Accepted: 2/23/2015

DOI: 10.1590/0101-60830000000039

Address correspondence to: Lucas Spanemberg. Psychiatric Unit, 6º andar sul, Hospital São Lucas, Pontifícia Universidade Católica do Rio Grande do Sul – 90610-000 – Porto Alegre, RS, Brazil. Phone: +55 (51) 3320-3041. Email: lspanemberg@yahoo.com.br


Background: General Hospital Psychiatric Units have a fundamental importance in the mental health care systems. However, there is a lack of studies regarding the level of improvement of patients in this type of facility. Objective: To assess factors related to good and poor outcomes in psychiatric inpatients using an index composed by clinical parameters easily measured. Methods: Length of stay (LOS), Global Assessment of Functioning (variation and at discharge) and Clinical Global Impression (severity and improvement) were used to build a ten-point improvement index (I-Index). Records of psychiatric inpatients of a general hospital during an 18-month period were analyzed. Three groups (poor, intermediate and good outcomes) were compared by univariate and multivariate models according to clinical and sociodemographic variables. Results: Two hundred and fifty patients were included, with a percentage in the groups with poor, regular and good outcomes of 16.4%, 59,6% and 24.0% respectively. Poor outcome at the discharge was associated mainly with lower education, transient disability, antipsychotics use, chief complaint “behavioral change/aggressiveness” and psychotic features. Multivariate analysis found a higher OR for diagnoses of “psychotic disorders” and “personality disorders” and others variables in relation to protective categories in the poor outcome group compared to the good outcome group. Discussion: Our I-Index proved to be an indicator of that allows an easy and more comprehensive evaluation to assess outcomes of inpatients than just LOS. Different interventions addressed to conditions such as psychotic disorders and disruptive chief complaints are necessary.

Moreschi HK et al. / Arch Clin Psychiatry. 2015;42(1):6-12

Keywords: Psychiatry unit, inpatients, length of stay, clinical impression, global functioning, psychiatric diagnosis, outcomes.


Since the late 1970s, a new proposal for the composition of mental health care system was progressively implemented in some European countries, based on deinstitutionalization and replacement of asylums for community-based psychiatric services and beds in general hospitals1-3. Influenced by the European movement, the process of reformation of the psychiatric care in Brazil has led to a significant decrease in psychiatric beds in the past twenty years, even though replacement services have not expanded at the same pace4,5. In this context, psychiatric wards within general hospitals became the main facilities for treatment of acute cases, increasing the importance of General Hospital Psychiatric Units (GHPU)6.

Although GHPU have a fundamental importance in this new model of care, there is a lack of studies regarding the level of improvement of patients in this type of facility. Therefore, it is also not well established what are the best general parameters for evaluating outcomes in patients admitted to general hospitals. Shorter psychiatric length of stay (LOS) has been considered a strong indicator of good outcomes both in specialty and general hospitals7-9. Although the LOS of psychiatric inpatients has decreased in recent decades (from months to days), it is still longer than for patients with physical illnesses, increasing expenditures on health, generating stigma and delaying social reintegration of the patient7. Functional ratings as the Global Assessment of Functioning (GAF)10-12 and the Brief Psychiatric Rating Scale (BPRS)13 also have been used to measure outcomes in psychiatric inpatients acutely ill, as well as the Clinical Global Impression (CGI)14, a measure of disease severity.

In Brazil, only one study was conducted to assess outcomes of psychiatric inpatients in general hospitals. Dalgalarrondo et al.15 created a variable called “outcome of admission” on the basis of a combination of two other variables: LOS and condition at discharge, a non-standardized clinical assessment. These authors found three variables (poor social functioning before admission, advanced age and organic mental disorder) associated with the “worst outcomes”. Despite the merits of this study, the measure constructed to assess these “worst outcomes” used an unusual and subjective criterion for evaluating the condition at discharge, making it difficult to be reproduced.

The use and development of assessment outcomes parameters as routine outcome measurements (ROM) is particularly important in mental health. In addition to recent changes in model of care mentioned above, the evaluation of outcomes has a dual role: evaluating clinical results and generating data for the construction of a care policy and financing model in mental health. While countries like England already possess broader and pragmatically built outcome measures as the Health of the Nation Outcome Scales (HoNOS)16, the care reality in low- and middle-income countries is much more precarious. In Brazil, for example, the only variables available for assessing results in mental health public system are the psychiatric diagnosis and the LOS. The lack of funding and consequently the lack of professionals and technologies for the assessment of outcome parameters make it difficult to evaluate true reality of assistance. Thus, the proposal of measures of minimal clinical parameters of evaluation of outcomes in mental health is an urgent demand.

The present study aims: 1) to propose and test an index of evaluation outcomes for psychiatric inpatients, using usual and easily measured clinical variables to compose an outcome score; 2) to investigate clinical and sociodemographic factors related to positive and negative outcomes in psychiatric inpatients in a GHPU classified by this index.


Study design, data source and sampling design

All records of admission to a Psychiatric Unit of a General Hospital (Hospital São Lucas da PUCRS – HSL/PUCRS) in Porto Alegre, Brazil, were selected during 18 months (from February, 2013, to August, 2014). This unit has 18 psychiatric beds for public (six beds) and private (twelve beds) patients. We assessed data in two moments: 1) admission and; 2) discharge (last day of hospitalization). All patients admitted to the unit are evaluated on a routine protocol in the early hours of hospitalization. This protocol is part of routine care of the psychiatric unit and includes sociodemographic and clinical data, as well as tools to assess global functioning and severity of illness (described below). Some variables as “chief complaint” were categorized according to their distribution in the emergency room, in accordance with a classification already described in previous studies17. All patients who received medical discharge during these 18 months were included in the study. When the routine protocol is completed at discharge, some measures of improvement are collected to assess treatment response (as we describe below). Our initial sample consisted of 287 patients. We excluded from the analysis patients who did not have any data of the five outcome variables (CGI-I, CGI-S at discharge, GAF at discharge, Δ-GAF and LOS) and patients who discontinued treatment before medical discharge (n = 37). The final sample was composed of 250 patients.


Clinical Global Impression – Severity (CGI-S): this is one of the most widely used assessment tools in psychiatry, easy to apply and interpret, besides being in the public domain. The CGI is rated on a 7-point scale, with the severity of illness scale using a range of responses from 1 (normal) to 7 (amongst the most severely ill patients)18.

Clinical Global Impression – Improvement (CGI-I): as the instrument described above, the CGI-I is also in the public domain and assesses the degree of patient improvement or response to treatment. Scores range from 1 (very much improved) to 7 (very much worse)18.

Global Assessment of Functioning (GAF): This tool composes the so-called Axis V in the DSM-IV Multiaxial System19. It is used to report the clinician’s judgment of the overall level of functioning of the patient, rating subjectively the social, occupational, and psychological functioning of adults. The scale ranges from 0 (inadequate information) to 100 (higher functioning), with ten categories of functioning. Within each category, there is a range of 10 points, describing and exemplifying patterns of functioning in various environments. A number should be chosen as the most descriptive of the overall functioning of the patient.

Box 1. Score Index-I according to the score of each variable

Box 1. Score Index-I according to the score of each variable

Index to assess outcomes

In order to construct a measure that could consider several parameters of improvement commonly used in the literature, inexpensive, and easily collected in Brazilian care reality, we propose an Improvement Index (I-Index) with a score ranging from 0 to 10 points. This index takes into account five variables: length of stay, CGI-S at discharge, CGI-I, GAF at discharge and GAF variation (GAF at discharge – GAF at admission or Δ-GAF). These instruments were chosen due to four pragmatic criteria: 1) they are readily applicable and information be easily collected; 2) their application does not burden the assistant psychiatrist or the patient, that is, the clinical care is not modified or interfered with; 3) measures are usually assessed in clinical outcome studies of psychiatric inpatients; and 4) the psychiatrists of our institution are acquainted with the measures. Each variable might score from 0 to 2 points, according to the guidelines in box 1. The final Index-I score can achieve 10 points, generating three groups with the following cutoffs: from 0 to 4 points = poor outcome; from 5 to 7 points = regular outcome; and from 8 to 10 points = good outcome. The scores of GAF, Δ-GAF and LOS were data-driven defined, according to their mean and standard deviations in our sample.

Data analyses

Descriptive analyses were presented by means and standard deviations (SD) for continuous variables, and by numbers and percentages (%) for categorical variables. The initial scores for each variable of the
I-Index were calculated according to the distribution of each variable as illustrated in the box 1. Differences between groups in sociodemographic and clinical continuous variables were analyzed with analysis of variance (ANOVA) test, with Tukey’s multiple comparison test as post hoc analysis. Categorical variables were analyzed with Pearson chi-square tests and the analyses post hoc of the adjusted residuals were also performed to reveal the differences among the categories of each variable. In order to evaluate the correlations among variables used in I-Index, Pearson correlation was calculated, with the following parameters: very weak (from 0.00 to 0.19), weak (from 0.20 to 0.39), moderate (from 0.40 to 0.59), strong (0.60 to 0.79) and very strong (0.80 to 1.00) correlations20. To identify admission factors independently associated with discharge measurements, the polytomous multivariable logistic regression was used. The I-Index “good outcome” was chosen as a reference to estimate odds-ratios (OR) of “regular outcome” and “poor outcome”. The variables included at the multivariate analysis were those with p < 0.20 at uncontrolled analysis. Since the variables “diagnosis” and “chief complaint” show covariance two independent final models were calculated. The final models were established excluding variables with less interference one-by-one.

The p value for significance was set at 0.05. The statistical analyses were performed using the software SPSS 18.0 (IBM SPSS, Inc., 2009, Chicago, IL, www.spss.com).

Ethics considerations

This study was approved by the Research Ethics Committee of Pontifical Catholic University of Rio Grande do Sul (protocol number: 565.190).

Table 1. Sociodemographic and clinical data of the total sample and univariate differences among groups according to type of outcome

Table 1. Sociodemographic and clinical data of the total sample and univariate differences among groups according to type of outcome

Table 2. Clinical variables on admission and at discharge (n = 250)

Table 2. Clinical variables on admission and at discharge (n = 250)

Table 3. Polytomous logistic regression comparing different patterns of discharge outcomes

Table 3. Polytomous logistic regression comparing different patterns of discharge outcomes


Table 1 lists the sociodemographic and clinical data of the total sample (n = 250). Most of the patients were female (65.2%), with an average age of 41 (SD = 17.6). The majority of them was either single or separated (58.7%) and nonsmokers (70.2%); 38.4% were employed or active. Most of the patients had previous psychiatric hospitalizations (50.6%) and 48.4% had clinical comorbidities. The most frequent specific chief complaint of evaluation was change in behavior/aggressiveness (21.5%), followed by suicidal ideation (19.4%), suicide attempt (19%), psychotic symptoms (14%) and substance abuse (7.9%).

Table 2 compares clinical variables on admission and at discharge. The length of stay had a mean of 27.12 (±15.04) days. During hospitalization, there was an increase in the use of antipsychotics (+23.2%), with discrete changes in other classes of medication, such as a decrease in the use of benzodiazepines (-4.8%). Patients had an average increase in GAF of 31.92 points from admission to discharge. The CGI-S decreased 2.06 points in mean and the average of CGI-I was 5.8 points.

Box 2. Data-driven Index-I scores according to the score of each variable

Box 2. Data-driven Index-I scores according to the score of each variable

Box 3. Pearson correlations among the variables used in the I-Index

Box 3. Pearson correlations among the variables used in the I-Index

The box 2 presents the values used to generate the points of each variable according to the means and SD of the variables. Box 3 presents the results of correlations among variables of the Index-I. The LOS presented no significant (with CGI-I x Δ-GAF) or very weak (with CGI-S at discharge x GAF-D) correlations with others variables. The higher correlation was between GAF-D x Δ-GAF (strong) and GAF-D x CGI-S at discharge (moderate), while the other correlations were weak. In relation to the I-Index score, strong correlations with CGI-S (negative), GAF-D and Δ-GAF were found; moderate correlation with CGI-I; and weak correlation with LOS (negative).

In relation to the Index-I, 41 patients (16.4%) were classified in the poor outcome group, 140 (59.6%) in the regular outcome group and 60 (20%) in the good outcome group. The results of the univariate analyses comparing these groups are also presented in table 1.

The poor outcome group frequently had low education, higher percentage of transient disability and smaller percentage of active/employed in occupational status, more psychotic symptoms and change in behavior/aggressiveness and less suicidal ideations in chief complaints. This group also had more psychotic disorders and less depressive disorders as psychiatric diagnoses and more use of antipsychotics than the good outcome group (regular outcome group presented usually intermediate results).

The main results of the multivariate analyses (Table 3) show two different clinical admission factors related with a poorer outcome. Two regression models are present. With regards to diagnosis at admission, psychotic disorders (OR: 16.77; CI: 3.16 – 89.10), personality disorders (OR: 9.76; CI: 1.51 – 63.05) and “others” (OR: 8.19; CI: 1.52 – 44.15) were associated with poorer outcome at discharge compared with the reference variable depressive disorder. The admission chief complaints of “psychotic symptoms” (OR: 12.42; CI: 2.28 – 67.75) and “change in behavior/aggressiveness” (OR: 25.19; CI: 4.48 – 141.72) were also associated with poor outcome at discharge (in relation to suicide ideation). Transient disability was associated with poor outcome in both models.


This study proposed an index of evaluation of outcomes for psychiatric inpatients using variables easily measured and routinely collected in psychiatric units. Our strategy was able to identify clinical and sociodemographic risk factors associated with positive and negative outcomes regarding improvement at discharge. The results replicate and extend findings of the literature that used single variables as outcome, proposing more comprehensive and clinically useful criteria for evaluating results in psychiatric inpatients, what is particularly important for the mental care reality of low- and middle-income countries.

The first strategy of this study was composing an index of improvement that encompasses more than one dimension related to the outcome of inpatients. The most part of the literature has predominantly used isolated variables in this assessment, such as LOS10,12,21,22, as a single parameter of outcome. However, LOS might mistakenly evaluate a hospitalization for a long period as a poor outcome, despite a significant improvement in symptoms and functionality of a patient. Although LOS is recognized as an important parameter, it is often chosen because it relates directly to health expenditure7, underestimating clinical issues. Furthermore, LOS may even underestimate the degree of improvement. Prince et al. found very short admissions as predictors of readmission in patients with mood disorder in a US population-based study23, suggesting the insufficiency of this parameter. We found only a weak correlation between the I-index and the LOS. In addition, the weak correlation between LOS and other variables of the I-Index and the weak correlations of most variables with each other reinforce the independence and complementarity of them in our sample. Other authors have used subjective variables (such as “good clinical condition at discharge”)15 or indices constructed from more complex methodologies24, complicating replicability or limiting its application. Thus, the variables and the method used to measure the I-Index can be considered simple and practical for treatment environments with limited resources.

The sociodemographic variables associated with poor outcomes were low education and patients with transient disability. While several other variables have been usually related with poor outcomes (such as unemployment, being unmarried and public insurance)10,21, no other studies have evaluated transient disability, a very common condition in inpatients in Brazil; our results show increased OR for transient disability in relation to employed/active status on the groups with intermediate (OR vary from 3.96 to 4.36) and poor (OR vary from 9.43 to 15.01) outcomes. Disability pension due to mental disorders has been associated with increased suicide risk25 and heavy use of psychiatric inpatients services26, while this outcomes have not yet been evaluated in transient disability. On the other hand, being employed/active in occupational status was associated with a good outcome, which has been extensively replicated in the literature10,15.

Psychotic symptoms and change in behavior/aggressiveness were the most prevalent chief complaints and were both independently associated with the poor outcome group, while suicidal ideation was the most prevalent one in the good outcome group. Concerning psychiatric diagnosis, psychotic disorders were related to poor outcome and unipolar depressive disorders to good outcome. Psychotic disorders were the diagnostic group most associated with poor outcome when compared with depressive disorders. These results also possibly explain respectively the highest prevalence of psychotic symptoms and change in behavior/aggressiveness in the poor outcome group (related to psychotic disorders and more severe cases) and a higher prevalence of suicidal ideation in the good outcome group (related to depressive disorders). The relation between psychotic disorders and bad outcomes, including readmissions (“revolving door phenomenon”)27,28, longer hospitalizations7,10, 21 and higher mortality29, is well established. This association further explains the poor outcome for patients taking antipsychotics at admission. Psychotic disorders as schizophrenia are usually chronic disorders associated with long-lasting symptoms resistant or refractory to treatment30. It is also well established that patients with aggression issues are difficult to treat and keep in compliance on an outpatient basis and readmissions rates are high for them28; this corroborates that these patients have a poor outcome. On the other hand, unipolar depression was highly prevalent in those with good outcomes. While Masters et al.10 found depressive disorders with shorter LOS than schizophrenia and bipolar disorders, Green and Griffiths31 found a substantial decline in LOS over the last decades for patients with depressive disorders, with no changes in LOS for schizophrenia in England. Our results suggest that beyond the LOS, improving symptoms and level of functioning also differentiates these two diagnostic groups.

In the multivariate analysis, both models found high OR for specific variables associated with an increased risk for poor outcome in relation to those related to protective factors. Thus, in addition to psychotic disorders, transient disability and chief complaints of psychotic symptoms and change in behavior/aggressiveness, the poor outcome group presents a high chance (OR = 9.76) of having a diagnosis of personality disorder than the protective diagnosis of depressive disorder. Thus, although personality disorder showed shorter LOS, our more complex approach to examine outcomes could relativize the weight of this variable, valuing other clinical aspects. In this line, Leontieva and Gregory found shorter LOS in inpatients with borderline personality disorder compared with other diagnoses, but significantly more management problems, such as incidents of self-harm, episodes of restraint, stat administrations of medications and readmissions32. Thus, the chronic disruptive behavior of such patients can make the hospitalization insufficient to improve the functioning and the severity of their symptoms, regardless of LOS.

Although our index is composed of measures of easy extraction and availability, it is limited to assess other dimensions that are associated with hospitalization outcomes. Quality of care measures, satisfaction with care, quality of life, perception of improvement by the patients and their family and evaluation of specific symptoms of each diagnosis cluster (among others) can compose a more complex and dynamic evaluation of outcomes, but require the availability of human resources to conduct the process of applying them. Others more complex ROM approaches using broad standard instruments (as HoNOS) and more robust methodological designs (based on both anchor- or distribution-based approaches)33 must also be developed in more structured services. Thus, although limited, the measures of the I-Index can serve as an initial outcome indicator in mental health systems with limited resources and personnel.

This work has a number of limitations. First, we used data selected retrospectively from records, not being possible to test the inter-rater reliability of measuring instruments. Second, diagnoses were made by clinical evaluation without the use of standardized instruments and using only the primary diagnosis (comorbidities were not considered). Third, although our I-Index comprises five different parameters, we did not use any instrument evaluating different symptom dimensions. While we strongly suggest that LOS, CGI and GAF should be used, the BPRS may be another additional simple measure to be used in the evaluation of outcomes in clinical settings with limited time and personnel for the use of more complex tools. In addition, more complex parameters (such as evaluation of results by patients and use of more specific instruments assessing other dimensions of variable) can be useful, but require a more complex logistical organization that the current reality of mental health assistance in Brazil. The comparison of this index with other measures using more complex instruments can bring validation data for both. Forth, our sample size is limited and was selected in only one institution, limiting a generalization of the results and the interpretation of the regression analysis for some variables. Finally, shorter LOS is not necessary associated with a good outcome, just as the I-index indicates. However, as the LOS is easily measured and classically used as a measure of outcome, we prefer to ponder its weight and keep it in the index. The use of other parameters and the assignment of weights according to the specific objectives of each evaluation (measurement of clinical improvement or use of data to support the allocation of health funding) should be better tested in large clinical samples.

In conclusion, we suggest that an assessment composed of simple parameters can be useful for measuring outcomes in psychiatric inpatients. The identification of factors associated with poor outcomes may help build strategies to minimize or lessen the health, social and financial burden of mental disorders. Social support and health care programs directed to vulnerable groups can relieve the patients after hospitalization and prevent readmissions. The use of composed parameters to evaluate outcomes as the I-Index can be easily incorporated by managers of mental health policies in treatment environments to support funding of mental health service and evaluate its quality.


We thank our team at the Psychiatric Unit of Hospital São Lucas for their help in the development of this study.

Conflicts of interest

The authors have no conflicts of interest.


1. Piccinelli M, Politi P, Barale F. Focus on psychiatry in Italy. Br J Psychiatry. 2002;181:538-44.

2. Sayers J. The world health report 2001 – Mental health: new understanding, new hope. Bulletin of the World Health Organization. 2001;79:1085.

3. Botega NJ. Prática Psiquiátrica no Hospital Geral: Interconsulta e Emergência: Artmed Editora; 2012.

4. Duarte SL, Garcia MLT. Psychiatric reform: the path of psychiatric beds reduction in Brazil. Emancipação. 2013;13(1):39-54.

5. Weber CAT. Direction of mental health in Brazil after 1980. Revista Debates em Psiquiatria. 2013;3:14-22.

6. Botega NJ. Psychiatric units in Brazilian general hospitals: a growing philanthropic field. Int J Soc Psychiatry. 2002;48(2):97-102.

7. Tulloch AD, Fearon P, David AS. Length of stay of general psychiatric inpatients in the United States: systematic review. Administration and policy in mental health. 2011;38(3):155-68.

8. Glick ID, Sharfstein SS, Schwartz HI. Inpatient psychiatric care in the 21st century: the need for reform. Psychiatr Serv. 2011;62(2):206-9.

9. Baruch Y, Kotler M, Lerner Y, Benatov J, Strous R. Psychiatric admissions and hospitalization in Israel: an epidemiologic study of where we stand today and where we are going. Isr Med Assoc J. 2005;7(12):803-7.

10. Masters GA, Baldessarini RJ, Ongur D, Centorrino F. Factors associated with length of psychiatric hospitalization. Compr Psychiatry. 2014;55(3):681-7.

11. Affairs DoV. Patient treatment file (PTF) coding instructions. Washington, DC: Department of Veterans Affairs, 2008 February 4, 2008. Report No.

12. Douzenis A, Seretis D, Nika S, Nikolaidou P, Papadopoulou A, Rizos EN, et al. Factors affecting hospital stay in psychiatric patients: the role of active comorbidity. BMC Health Serv Res. 2012;12:166.

13. Rocca P, Mingrone C, Mongini T, Montemagni C, Pulvirenti L, Rocca G, et al. Outcome and length of stay in psychiatric hospitalization, the experience of the University Clinic of Turin. Soc Psychiatry Psychiatr Epidemiol. 2010;45(6):603-10.

14. Warnke I, Rössler W, Herwig U. Does psychopathology at admission predict the length of inpatient stay in psychiatry? Implications for financing psychiatric services. BMC Psychiatry. 2011;11(120):1-10.

15. Dalgalarrondo P, Botega NJ, Banzato CE. [Patients who benefit from psychiatric admission in the general hospital]. Rev Saúde Pública. 2003;37(5):629-34.

16. Macdonald AJ, Elphick M. Combining routine outcomes measurement and “Payment by Results”: will it work and is it worth it? Br J Psychiatry. 2011;199(3):178-9.

17. Spanemberg L, Nogueira EL, da Silva CT, Dargel AA, Menezes FS, Cataldo Neto A. High prevalence and prescription of benzodiazepines for elderly: data from psychiatric consultation to patients from an emergency room of a general hospital. Gen Hosp Psychiatry. 2011;33(1):45-50.

18. Guy W. Clinical Global Impression (CGI). In: Guy W, editor. ECDEU Assessment Manual for Psychopharmacology. Rockville: US Dept. of Health, Education and Welfare; 1976. p. 218-22.

19. American Psychiatric Association, American Psychiatric Association. Task Force on DSM-IV. Diagnostic and statistical manual of mental disorders: DSM-IV-TR. 4th ed. Washington, DC: American Psychiatric Association; 2000. xxxvii, 943 p.

20. Evans JD. Straightforward statistics for the behavioral sciences. Pacific Grove: Brooks/Cole Pub. Co.; 1996. xxii, 600 p.

21. Warnke I, Rossler W, Herwig U. Does psychopathology at admission predict the length of inpatient stay in psychiatry? Implications for financing psychiatric services. BMC Psychiatry. 2011;11(1):120.

22. Zhang J, Harvey C, Andrew C. Factors associated with length of stay and the risk of readmission in an acute psychiatric inpatient facility: a retrospective study. Aust N Z J Psychiatry. 2011;45(7):578-85.

23. Prince JD, Akincigil A, Kalay E, Walkup JT, Hoover DR, Lucas J, et al. Psychiatric rehospitalization among elderly persons in the United States. Psychiatric services. 2008;59(9):1038-45.

24. Barbato A, Parabiaghi A, Panicali F, Battino N, D’Avanzo B, de Girolamo G, et al. Do patients improve after short psychiatric admission?: a cohort study in Italy. Nord J Psychiatry. 2011;65(4):251-8.

25. Rahman S, Alexanderson K, Jokinen J, Mittendorfer-Rutz E. Risk factors for suicidal behaviour in individuals on disability pension due to common mental disorders – a nationwide register-based prospective cohort study in Sweden. PLoS One. 2014;9(5):e98497.

26. Morlino M, Calento A, Schiavone V, Santone G, Picardi A, de Girolamo G, et al. Use of psychiatric inpatient services by heavy users: findings from a national survey in Italy. Eur Psychiatry. 2011;26(4):252-9.

27. Gastal FL, Andreoli SB, Quintana MIS, Gameiro MA, Leite SO, McGrath J. Predicting the revolving door phenomenon among patients with schizophrenic, affective disorders and non-organic psychoses. Rev Saude Publica. 2000;34:280-5.

28. Nourse R, Reade C, Stoltzfus J, Mittal V. Demographics, clinical characteristics, and treatment of aggressive patients admitted to the acute behavioral unit of a community general hospital: a prospective observational study. Prim Care Companion CNS Disord. 2014;16(3).

29. Hoang U, Stewart R, Goldacre MJ. Mortality after hospital discharge for people with schizophrenia or bipolar disorder: retrospective study of linked English hospital episode statistics, 1999-2006. BMJ. 2011;343:d5422.

30. Knapp M, Mangalore R, Simon J. The global costs of schizophrenia. Schizophr Bull. 2004;30(2):279-93.

31. Green BH, Griffiths EC. Hospital admission and community treatment of mental disorders in England from 1998 to 2012. Gen Hosp Psychiatry. 2014;36(4):442-8.

32. Leontieva L, Gregory R. Characteristics of patients with borderline personality disorder in a state psychiatric hospital. J Pers Disord. 2013;27(2):222-32.

33. Parabiaghi A, Kortrijk HE, Mulder CL. Defining multiple criteria for meaningful outcome in routine outcome measurement using the Health of the Nation Outcome Scales. Soc Psychiatry Psychiatr Epidemiol. 2014;49(2):291-305.