An initial survey was conducted on January 10, 2014, targeting patients hospitalized for 1.5 years or longer in 32 designated inpatient facilities nationwide, established under the MTS framework designated by the MHLW. Of the 791 patients hospitalized in MTS Act wards on that date, 563 had been hospitalized for less than 1.5 years and were excluded. Among the remaining 228, two declined participation after being informed about the study, resulting in an inclusion rate of 99.1%. After excluding 10 patients with unknown outcomes and six who were transferred to other hospitals, a final total of 210 patients were enrolled.

This prospective cohort study examined patients hospitalized under the MTS Act for 1.5 years or longer, using data from 32 designated institutions nationwide. Cox regression and descriptive statistics were used for analysis. In 2015, the MHLW Research Group established criteria for identifying “severe and chronic” psychiatric patients (Figure 1). These criteria include four components: “Psychiatric symptoms,” “Behavioral disorders,” “Life disorders,” and “Physical complications (18)”. To distinguish “Psychiatric symptoms” as a specific criterion for “severe and chronic” status — rather than general psychiatric symptoms — we enclosed the term in quotation marks (18).

Cox regression analysis was conducted to examine the association between clinical variables and treatment outcomes (discharged vs. inpatient) at 1.5, 2.5, and 3.5 years. The dependent variable was outcome status, while independent variables included age, sex, duration of hospitalization at the time of the survey, and the components of the proposed “severe and chronic” criteria (“Psychiatric symptoms,” “Behavioral disorders,” “Life disorders”) as well as “Physical complications”. Given the previously established association between psychiatric hospitalization duration and both age and sex in MTS Act wards, these variables were included as covariates in the analysis (20).

Additionally, multiple linear regression analysis was performed using the actual length of stay (in days) as the dependent variable. Independent variables included age, sex, psychiatric diagnosis, and the three primary components of the proposed “severe and chronic” criteria. This analysis was conducted to capture a more nuanced understanding of hospitalization duration, independent of policy-defined time points. However, despite repeated attempts at multiple regression analysis, there were missing values for 1.5-year timepoint.

To identify factors contributing to long-term hospitalization, we conducted Cox proportional hazards regression analysis using discharge outcome as the dependent variable. Independent variables included the components of the proposed criteria — “Psychiatric symptoms,” “Behavioral disorders,” and “Life disorders” — as well as “Physical complications,” their sub-items, and hospitalization duration. A hazard ratio significantly less than 1 was interpreted as indicative of a lower likelihood of discharge. For exploratory purposes, no Bonferroni correction was applied. Statistical significance was defined as p < 0.05.

Statistical analyses were conducted using SPSS for Windows, version 20.0 (IBM Corp., Armonk, NY).

Ethical approval was obtained from the Ethics Committee of the National Center of Neurology and Psychiatry (Approval No. A2014-004; study title: Investigation of the Criteria for “Severe and Chronic” in Medical Observation Law Wards). The study was conducted in accordance with the latest revision of the Declaration of Helsinki. Written informed consent was obtained from all participants prior to enrollment.

The 210 enrolled subjects had a mean hospitalization duration of 1,026 days (95% confidence interval [CI]: 974–1,079; median: 943 days). By July 15, 2015, 126 patients (60.0%) had been discharged; by July 15, 2016, 166 (79.0%); and by July 15, 2017, 185 (88.1%). As of the final survey, 25 patients (11.9%) remained hospitalized in MTS Act wards (Table 1). The median length of stay was 943 days at baseline, increasing to 1,516 days at 1.5 years, 1,950.5 days at 2.5 years, and 2,246 days at 3.5 years. Median lengths of stay were consistently longer for men than for women. Of the total sample, 171 patients (81.4%) were male, and 186 (88.6%) were diagnosed with ICD-10 F2 disorders (schizophrenia, schizophreniform disorder, or paranoid disorder). Among male patients, 11 (6.4%) were diagnosed with F1 disorders (mental and behavioral disorders due to psychoactive substance use), three (1.8%) with F0 disorders (organic, including symptomatic mental disorders), two (1.2%) with F3 disorders (mood disorders), one (0.6%) with an F4 disorder (neurotic, stress-related, and somatoform disorders), two (1.2%) with F8 disorders (developmental disorders), and one (0.6%) with a G40 condition (epilepsy). Among female patients, four (10.3%) were diagnosed with F3 disorders (Table 2). The mean age of the sample was 44 years (95% CI: 42–46 years). Regarding index offenses, 63 male cases (36.8%) involved injury, while 13 female cases (33.3%) involved attempted murder. The most common offenses were injury (34.8%), attempted murder (19.5%), arson (18.1%), and homicide (17.6%), followed by robbery (3.3%), indecent assault (3.3%), attempted arson (1.9%), and attempted robbery (1.4%). Robbery, indecent assault, and attempted arson were reported only among male patients (Table 3). In terms of physical complications, 15 patients had water intoxication, eight had recurrent ileus, four had recurrent pneumonia, and 15 had other physical complications. A total of 28 patients presented with at least one physical complication.

Seventy-nine subjects met the proposed “severe and chronic” criteria (Table 1). Under the proposed framework, “Physical complications” were treated as a subgroup of the “severe and chronic” classification and therefore were not included in the primary category. Among the 28 subjects with “Physical complications,” 16 also met the criteria for “Psychiatric symptoms”. Notably, no subject met criteria for both “Psychiatric symptoms” and “Physical complications” without also exhibiting “Behavioral disorders” or “Life disorders”. As a result, the total number of subjects classified as “severe and chronic” remained unchanged when transitioning from the provisional standard to the proposed criteria.

Table 4 presents the associations between discharge status (discharged vs. hospitalized) at 1.5, 2.5, and 3.5 years and the following variables: age, male sex, “Psychiatric symptoms,” “Behavioral disorders,” “Life disorders,” “Physical complications,” and hospitalization duration.

“Psychiatric symptoms” and “Life disorders” were significantly associated with continued hospitalization at 1.5 and 3.5 years (p < 0.05). At 2.5 years, only “Life disorders” remained significantly associated with ongoing hospitalization. In contrast, “Behavioral disorders,” Age, Sex, “Physical complications,” and length of hospitalization at baseline were not significantly associated with hospitalization status at any timepoint.

Tables 5–8 and Supplementary Tables 1 and 2 present the results of the Cox regression analysis for exploratory purposes, including estimated regression coefficients, hazard ratios, 95% CIs, p-values, and significance levels for discharge and continued hospitalization at 1.5, 2.5, and 3.5 years. Analyses were conducted for the sub-items of “Psychiatric symptoms,” “Behavioral disorders,” “Life disorders,” and “Physical complications”. Notably, among the “Behavioral disorders,” only suicidal ideation had a hazard ratio greater than 1 and was significantly associated with discharge. When comparing outcomes at 3.5 years with those at 1.5 and 2.5 years, the following sub-items were consistently associated with prolonged hospitalization across all timepoints (p < 0.05).

Psychiatric symptoms: positive symptoms, conceptual disorganization, suspiciousness, hallucinatory behavior, excitement, uncooperativeness, unusual thought content, and total BPRS score.

Additionally, certain items—such as comprehensive psychopathology (within “Psychiatric symptoms”) and behaviors including habits, posture, and undressing (within B-items of “Behavioral disorders”)—were significantly associated with hospitalization at 1.5 or 2.5 years but not at 3.5 years. No sub-items within the “Physical complications” category were significantly associated with hospitalization status at any timepoint. In addition to the Cox regression analysis, we conducted multiple regression analysis using the total number of hospitalization days as the dependent variable to provide a continuous perspective on hospitalization duration. At 3.5 years, male sex (B = 199.57, p = 0.0408), psychiatric symptoms (B = 202.17, p = 0.0288), and life disorders (B = 255.88, p = 0.0072) were significantly associated with longer hospital stays. At 2.5 years, male sex was also significantly associated with prolonged hospitalization (B = 203.08, p = 0.0076). These findings reinforce the contribution of these variables to long-term inpatient care in forensic psychiatric settings.

Our analysis indicated that key components of “Psychiatric symptoms” and “Life disorders” were significantly associated with the length of hospitalization, consistent with previous findings (Table 4; 21, 22). In contrast, the major components of “Behavioral disorders” and “Physical complications” — the latter ultimately excluded from the proposed criteria — did not show a strong association with hospitalization duration. However, a limited number of sub-items within “Behavioral disorders” were associated with long-term hospitalization. This approach revealed that male sex, psychiatric symptoms, and life disorders were significantly associated with longer hospitalization at 3.5 years, while male sex was also a significant factor at 2.5 years. These findings underscore the relevance of demographic and clinical predictors in explaining long-term institutionalization and suggest that certain associations, particularly those involving sex, may emerge more clearly when considering cumulative hospital stay rather than timepoint-specific discharge.

Importantly, analysis of “Psychiatric symptoms,” “Behavioral disorders,” “Life disorders,” and “Physical complications” (Table 5; Supplementary Tables 1, 2) revealed that positive symptoms and total scores within “Psychiatric symptoms,” as well as community life disorder and acts interfering with social adaptation under “Life disorders,” were significantly associated with prolonged hospitalization. Previous studies have also identified associations between hospitalization duration and positive symptoms measured by the BPRS (23), anxiety, depressive mood, financial difficulties (within “Life disorders”), as well as age and sex (24, 25). With respect to “Behavioral disorders,” B items—particularly resistance to stress—were associated with longer hospitalization. Although prior research identified aggressive behavior (26) and substance use (27) as contributors to extended stays, these associations were not confirmed in the present analysis. Suicidal ideation, notably, was found to be significantly associated with shorter hospitalization in this study, similar findings have been found in previous studies, such as a Canadian study of patients with schizophrenia (28, 29). For instance, a recent study reported that patients admitted for serious overdose remained hospitalized longer due to heightened medical and psychiatric risk, underscoring contextual and diagnostic variability across samples (30). Regarding “Physical complications,” no significant association was observed in this cohort, likely due to the low prevalence (approximately 10%) of such conditions. This finding differs from earlier studies linking diabetes mellitus (30), electrolyte imbalance and hypotension (31), and cardiovascular and metabolic diseases (32) with longer hospital stays (33). However, water intoxication, ileus, and pneumonia are common physical complications in psychiatric hospitalized patients, and may be factors for poor prognosis. Polydipsia may identify a subset of schizophrenia patients whose enhanced stress reactivity contributes to their mental illness (34). Constipation is a common problem during clozapine treatment which can progress to full-blown ileus which can be fatal (35). Advanced age, underweight, smoking habit, use of atypical antipsychotics, and large doses of antipsychotics were risk factors for pneumonia that is a major cause of death in patients with schizophrenia (36). In the “severe and chronic” criteria, it is thought that there was no significant difference in “Physical complications” because there were few people who met the criteria, but if “Physical complications” are expanded to include milder cases, it is possible that they could become a factor in long-term hospitalization. Additionally, no significant associations with age or sex were observed in this analysis, which diverges from prior findings in general psychiatry settings. For instance, a study found that older age and male sex were independently linked to longer psychiatric stays, suggesting that demographic influences may operate differently in forensic versus civil psychiatric population (37). This may reflect differences in demographic composition between patients in Mental Health and Welfare Law wards and those in MTS Act wards.

This study has several limitations that should be considered when interpreting the findings. First, the number of participants in MTS Act wards was limited compared to those in Mental Health and Welfare Law wards, which imposed a physical constraint on sample size. Second, the duration of hospitalization prior to the survey varied among participants, and this variability was not incorporated into the analysis. Third, although most participants were diagnosed with schizophrenia, individuals with other psychiatric conditions were also included. Fourth, only the primary psychiatric diagnosis was recorded; comorbidities such as personality disorders and alcohol or drug addiction were not captured.

Fifth, post-discharge factors—including family relationships, residential environment, and community reintegration—were not examined, limiting the scope of interpretation regarding real-world outcomes. Prior studies have highlighted the impact of family involvement, institutional characteristics, and system-level predictors of length of stay and readmission, suggesting that these unmeasured variables may have influenced the results. Sixth, multiple comparisons were not adjusted for, given the exploratory nature of the study and the large number of variables assessed. This decision was intended to facilitate hypothesis generation for future research. Seventh, socioeconomic status and educational attainment — both known to influence psychiatric hospitalization duration — were not included due to constraints imposed by the predefined parameters of the study, which were determined by a research group affiliated with the MHLW. Eighth, the findings may not be generalizable to patients under the Mental Health and Welfare Law due to differences in legal frameworks and ward structures. In addition, various sociodemographic and clinical differences between the analyzed groups also may limit generalization. This study was limitedly carried in forensic psychiatric facilities in Japan, and because healthcare systems and forensic psychiatric care systems vary from country to country, there may be limitations to the generalizability of these results to other countries and healthcare systems. Ninth, data collection and analysis were both conducted by entities under the jurisdiction of the MHLW, raising the potential for institutional bias. Tenth, the follow-up intervals of 1.5, 2.5, and 3.5 years reflect time elapsed since the study’s initiation and do not directly represent individual patient length of stay. Additionally, the multiple regression analysis could not be performed for the 1.5-year timepoint due to insufficient data. As a result, findings from this method primarily reflect outcomes at 2.5 and 3.5 years, and interpretations should be made within that context. However, tracking patients hospitalized for more than 1.5 years over an additional 3.5-year period enabled identification of individuals at risk for long-term hospitalization. Finally, this study did not assess facility-level variation in resources or clinical practices, which may significantly affect hospitalization duration; future research should employ multilevel models to account for such institutional factors.

In conclusion, by applying the proposed “severe and chronic” criteria to a forensic psychiatric population, this study identified significant associations between prolonged hospitalization and major components of “Psychiatric symptoms,” “Life disorders,” and selected sub-items of “Behavioral disorders”. These findings offer important implications for discharge planning, clinical prioritization, and policy development within forensic mental health services.