Table 5.

Results: Main Findings.

Study	Complexity	Main Findings
Constantinou et al., 2015	Single MHSS Micro level	1. The decision support system “DSVM-MSS” predicted general violence (area under the curve scores = 0.691 (pre-discharge) and 0.730 (post-discharge); this difference is not statistically significant (p = 0.472)) and violent convictions (area under the curve scores = 0.845 (pre-discharge) and 0.774 (post-discharge); this difference was not statistically significant (p = 0.469)) in people with mental health problems living in medium secure services.
Delany et al., 1994	Single MHSS Micro level	1. The direct relationship between organizational structure (formalization, autonomy, specialization, routinization, knowledge complexity, and centralization) and service amenities (personal maintenance needs, case management services, and health substance abuse and mental health services) was not statistically significant (z = 0.363). 2. The direct relationship between organizational structure and organizational relations (diversity of funding, relationships, constraints, and independence) was statistically significant beyond the 0.01 level (z = 3.152). 3. The direct relationship between organizational relations and services amenities was not statistically significant (z = 1.482). 4. The organizational structure affected services and amenities (personal maintenance needs, case management services, and health-substance abused and MHS) through the organizational relations dimension, including funding, relationships, constrains, and independence. 5. The model showed a good reproduction of the observed covariance matrix for the following variables: specialization; diversity of funding; relationships; constrains; personal maintenance needs; case management services; and health, substance abuse, and mental health services: ξ2 (11) = 18.908, p = 0.06275; Bentler-Bonnet Fit Index = 0.84.
Kim and Oh, 2012	Single MHSS Micro level	1. Leadership positively and significantly (p = 0.000) impacted Measurement, Analysis, and Knowledge Management; Strategic Planning; Patient, Customer, and Market Orientation; and Human Resources Orientation. Leadership did not significantly impact Process Management (p = 0.574) or Hospital Performance (p = 0.190). 2. Strategic Planning positively and significantly (p = 0.000) affected Patient, Customer, and Market Orientation and Process Management(p = 0.004), and it did not impact significantly on Human Resources Orientation (p = 0.492). 3. Patient, Customer, and Market Orientation positively and significantly impacted Hospital Performance (p = 0.000) and Process Management (p = 0.017). 4. Human Resources Orientation impacted Process Management (p = 0.000) and Hospital Performance (p = 0.000). 5. Process Management positively influenced Hospital Performance (p = 0.000). 6. Measurement, Analysis, and Knowledge Management positively impacted Strategic Planning; Patient, Customer, and Market Orientation; Human Resources Orientation; and Process Management (p = 0.000). 7. The structural model showed the following results: χ2 = 14.034 (df = 3), p = 0.012, χ2/df = 4.678, Goodness-of-fit Index = 0.994, Root Mean Residual = 0.009, Normed Fit Index = 0.997, and Confirmatory Fit Index = 0.998.
Wolf, 1978	Single MHSS Micro and Meso levels	1. Mental hospital utilization had a weak and negative impact on long-term utilization of nursing homes (r = −0.071). 2. Catchment areas where there are more admissions of elderly people had a higher percentage of urban (β = −0.089), non-white (β = 0.074), aged persons (β = 0.105) and more persons unmarried and living alone (β = 0.160). The proportion of foreign-born people did not influence the model (β = 0.001). This model explained 9% of the variance in mental hospital utilization.
Green and Griffiths, 2014	Group of MHSSs Micro level	1. The reduction of beds availability entailed an annual inpatient admissions decrease in: depression (β = −1085; p < 0.01), dementia (β = −764; p < 0.01), schizophrenia (β = −468; p < 0.01), bipolar disorder (β = −159; p < 0.01), and OCD (β = −21; p < 0.01); and increase in use of alcohol (β = 1764; p < 0.01), eating disorders (β = 55; p < 0.01), and posttraumatic stress disorder (β = 17; p < 0.01). 2. The reduction of beds availability significantly decreased length of hospital stay in: use of alcohol (β = −0.29, p < 0.001), eating disorders (β = −0.52, p < 0.001), dementia (β = −0.55, p < 0.001), and depression (β = −0.96, p < 0.001). 3. The reduction of beds availability increased the number of detentions under Mental Health Act (β = 298, p < 0.01). 4. The number of mental health beds was negatively associated with the number of psychiatric severe admissions (coefficient = −0.683; p < 0.001, bootstrapped 95% CI: 0.37 to 1.06). 5. The number of beds was negatively associated with community team activity (coefficient = −0.521; p < 0.001, bootstrapped 95% CI: −0.71 to 0.25). 6. The community team activity was not associated with inpatient admissions (coefficient = −0.121, p < 0.001, bootstrapped 95% CI: −0.35 to 0.42). 7. The model (a path from community team activity to hospital beds and from hospital beds to hospital admissions) showed good fit: χ2 = 0.57; df = 1; p = 0.45; Tucker–Lewis Index = 1.07, root mean square error of approximation = 0.00.
Roux et al., 2016	Group of MHSSs Micro level	1. Patient needs (adaptation to stress, social exclusion, involvement in treatment decisions, and job integration) and outcomes (quality of life and personal recovery) were negatively associated (β = −0.60; p < 0.001). 2. Service performance (type and amount of support provided) and outcomes were positively associated (β = 0.40; p < 0.001). 3. Patient needs and service performance were negatively associated (β = −0.30; p < 0.001). 4. The model provided a good fit for the data, as suggested by the following statistics: non-significant goodness-of-fit based on the Bollen–Stine bootstrap distribution ((7) = 14.3, p = 0.107), TLI above 0.95 (TLI = 0.967) and RMSEA not statistically greater than 0.05 (RMSEA = 0.056, one-sided P = 0.358). The model explained 67% of the variance in outcomes. 5. Service performance had a partial mediation role between needs and outcome. A total of 16.4% of the impact of needs on outcomes was mediated by service performance (standard error: 0.05, z = 3.6, p < 0.001 with the Bollen–Stine bootstrap method after 2000 iterations).
Salvador-Carulla et al., 2013	Group of MHSSs Meso level	1. The treated prevalence of a small health area during a specific year was the result of combining service users that were in contact with the mental health service (during the year t-1) and the new services users who contacted the specialized mental health services within this year. Psychiatric morbidity was the root variable, which caused the treated incidence of new patients and the treated prevalence of patients who were in contact with mental health community services. Treated prevalence directly influenced workforce capacity, relative technical efficiency, and activities with patients. Another root variable is public health budget, directly related to workforce capacity. Accessibility was the third root variable that influenced the treated incidence of new patients.