Self- and observer-rated functioning and psychiatric readmission: findings from multicenter longitudinal data Authors: Mai Iwanaga, Sosei Yamaguchi, Momoka Igarashi, Junko Koike, Asami Matsunaga, Yasutaka Ojio, Makoto Ogawa, Hiroyuki Watanabe, Chiyo Fujii, Hiroshi Kimura Published: 2026 ABSTRACT AbstractWe examined self- and observer-rated functioning across four groups defined by psychiatric readmission status. Data were derived from Early Discharge and Prognostic Community Outcomes for Psychiatric Inpatients in Japan, including 483 patients discharged from acute psychiatric wards. Self-rated functioning (Sheehan Disability Scale, SDS) and observer-rated functioning (Personal and Social Performance Scale, PSP) were assessed at index admission, discharge, and 6- and 12-month follow-ups after discharge. Participants were divided into four groups: readmission at both follow-ups, readmission at 6-month follow-up, readmission at 12-month follow-up, and no readmission. Between-group and within-group changes were examined using generalized linear mixed models. For PSP, the no-readmission group showed significantly better scores than the readmission groups at 6-month (adjusted p = 0.002) and 12-month (adjusted p < 0.001) follow-ups. Excluding the group with readmission at both follow-ups, the remaining three groups demonstrated improvements across all time points through 12 months after discharge. For SDS, scores improved over time only in the no-readmission group (adjusted p < 0.001). These findings indicate that changes in functioning differed according to readmission status. Improvements were more pronounced in observer-rated functioning, whereas self-rated functioning showed limited change. This discrepancy may reflect differences between self- and observer-rated assessments.Clinical trial registration: This study was registered in the UMIN Clinical Trials Registry (UMIN000034220).Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-58321-4.Keywords: Social functioning, Readmission, Self-assessment, Observer-assessment, Longitudinal study, DischargeSubject terms: Diseases, Health care, Medical research, Psychology, Psychology Introduction Social functioning is crucial in the recovery of individuals with psychiatric disorders. It captures both the capacity to perform societal roles and actual social performance1 and has been highlighted as a key treatment target in clinical research and routine assessments2,3. Measures of social functioning vary and include both self-assessment from patients themselves and observer-assessment by staff or others. In particular, self-rated functioning has gained increasing attention in line with the broader movement toward patient-reported outcome measures and patient-centered care4. However, recent systematic reviews have reported a low correspondence between self-rated and observer-rated social functioning5,6. Self-reported assessments may capture aspects of functioning that are not apparent to clinicians in observer-assessments. Clarifying the differences and similarities between self-rated and observer-rated functioning across various contexts is highly useful for improving mental health services. Psychiatric readmission is one of the most important indicators in mental health services. It reflects the quality of community care and imposes a substantial economic burden7. Although previous studies have shown that readmission is associated with impaired social functioning, how self-rated and observer-rated functioning are differentially related to readmission remains unclear. For the observer-rated social functioning, one systematic review reported that lower Global Assessment of Functioning (GAF) scores at admission were associated with a higher readmission risk8, while another systematic review noted that the influence of GAF scores on readmission status varied between studies9. Additionally, while the GAF is a popular measurement, its validity has been questioned10,11. Consequently, the use of an alternative validated measurement, such as the Personal and Social Performance Scale (PSP), has been recommended12. The PSP is an observer-assessment evaluated by clinical staff, similar to the GAF but more structured, and is one of the most commonly used methods in both research and routine assessments3. Previous studies have suggested that lower PSP scores are associated with higher likelihoods of admission13 and the revolving-door phenomenon9. However, the number of studies is extremely limited, and their generalizability to other regions is insufficient. With regard to self-assessment, while social functioning and various mental health outcomes have been examined14,15, research findings on the association between social functioning and readmission also vary. One study demonstrated that the level of functional impairment significantly predicted 12-month hospitalization status, even after controlling for established risk factors16. In another study, readmission was associated with changes in social functioning scores in an unadjusted model, but no significant difference was observed in a model adjusted for all potential covariates17. These inconsistent findings leave unclear whether self-rated social functioning has an association with readmission. Moreover, it remains unclear whether self-rated and observer-rated assessments of social functioning differ in their associations on readmission. The following three research gaps must be addressed to examine the association between social functioning and readmission. First, social functioning and readmission may be interactive. Some studies have reported that low social functioning predicts readmission8, whereas others suggest that community living without admission contributes to better social functioning18–20. Alternatively, successful treatment during hospitalization may also lead to improvements in social functioning. Therefore, a one-way predictive model may be inadequate. Second, social functioning not only interacts with readmission but is also influenced by multiple factors2. Hence, social functioning can also be considered a dynamic construct rather than a static characteristic. Therefore, changes in social functioning at multiple points in time should be examined. Third, differences in assessment perspectives may influence how these longitudinal changes are captured. whether self-rated and observer-rated social functioning show the same associations with readmission remains unclear. Addressing these gaps can deepen our understanding of the relationships between these variables. This study conducted a secondary analysis to examine differences between self-rated and observer-rated social functioning in four groups based on psychiatric readmission status at four time points. To address the first research gap, an alternative analytical approach other than a one-way model for predicting readmission or social functioning was required. Therefore, we created four groups according to readmission status and timing, and performed between-group comparisons at each time point and within-group comparisons across time points. For the second research gap, assuming that social functioning changes over time, we used data from four time points: admission, discharge, 6 months post-discharge, and 12 months post-discharge. To address the third research gap, we obtained and compared scores from self-rated and observer-rated social functioning from the same participants. This approach provides a more comprehensive understanding of the functional dimensions associated with psychiatric readmission. Materials and methods Study design and participants We used longitudinal data from the Early Discharge and Prognostic Community Outcomes for Psychiatric Inpatients in Japan (ePOP-J study), conducted at 21 psychiatric hospitals providing acute inpatient psychiatric care21. The 21 participating institutions in this study were recruited via snowball sampling from facilities with sufficient staff capacity to conduct the survey. Participants were inpatients admitted between October 1, 2018, and September 30, 2019, who were discharged from the index admission within 12 months. The eligibility criterion was a diagnosis of a psychiatric disorder based on the International Classification of Diseases 10th Revision. Exclusion criteria were age < 20 or > 60 years, hospital admission solely for specific treatments (i.e., electroconvulsive therapy, psychiatric evaluation, and respite care), and immediate transfer to another medical department for long-term care. In this study, data were collected at the time of index admission (T1), discharge (T2), and 6- (T3), and 12-month (T4) follow-ups after discharge following index admission. Of the 4,603 patients admitted to the 21 psychiatric hospitals, 2,093 met the eligibility criteria. Of them, 609 completed the assessment at T117. The detailed protocol for this study has been published previously21. This study was registered in the UMIN Clinical Trials Registry (UMIN000034220), and the study protocol was approved by the Ethics Committee of the Graduate School of Medicine at Chiba University (no. 3154). Measures To comprehensively understand the functional dimensions associated with psychiatric readmission, we examined social functioning in four groups based on the psychiatric readmission status at four time points. We conducted between-group comparisons at each time point and within-group comparisons across time points. Using this method, we examined whether there were different or similar trends between self-rated and observer-rated functioning. Self-rated functioning Self-rated social function was performed using the Sheehan Disability Scale (SDS)14,15. This scale includes three items: (1) work/studies, (2) social life/leisure activities, and (3) family life/home responsibilities. Work includes paid and unpaid volunteer work and training. Each item is preceded by the question: “The symptoms have disrupted your.” Each item is scored on a discretized analog metric of 0 (not at all), 1–3 (mildly), 4–6 (moderately), 7–9 (markedly), or 10 (extremely). The three items are summed to obtain a total score (range: 0–30), with higher scores indicating greater impairments in functioning. The SDS is the one of the most commonly used self-rated functional assessment scale22. The validity of the original and Japanese version of the SDS has been confirmed14,15,23. Observer-rated functioning Observer-rated social functioning was assessed using the Japanese version of the Personal and Social Performance Scale (PSP)24 administered by staff at each medical institution. The PSP was developed as part of a package for planning and evaluating psychiatric rehabilitation. It uses a single-item rating scale ranging from 1 to 100 points subdivided into 10 equal intervals12. The ratings are based on the assessment of patient functioning in four main areas: (1) self-care, (2) socially useful activities, (3) personal and social relationships, and (4) disturbing and aggressive behaviors. Higher scores indicate better functioning and fewer difficulties over the 2 weeks prior to the assessment. A score of 71–100 reflects mild difficulties, 31–70 reflects varying degrees of impairment, and ≤ 30 reflects poor functioning that requires intensive support or management. The PSP has been validated for sensitivity to changes in acute psychiatric samples25 and is used by clinicians worldwide as a rating tool for social and general functioning2,3. Psychiatric readmission status groups Psychiatric readmission status was examined twice, at 6 (T3) and 12 months (T4) after discharge. The patients were divided into groups according to their readmission status at T3 and T4: readmission at both T3 and T4 (Group 1), readmission at T3 only (Group 2); readmission at T4 only (Group 3); and no readmission (Group 4). Demographic and clinical characteristics Age, sex, and primary diagnosis were used as demographic characteristics. Age and sex data were obtained from clinical records at T1. Primary diagnosis was classified according to the International Classification of Diseases 10th Revision (F0–F9)26. Data from T1 and T2 were used as the primary diagnosis because the primary diagnosis of some participants changed during their hospital stay. The updated diagnosis at discharge was recorded as the final diagnosis for the index admission. For statistical analysis, to prevent model overparameterization and ensure adequate statistical power, the variable “primary diagnosis” was consolidated into four categories based on diagnostic frequency: schizophrenia, major depression, bipolar disorder, and other diagnoses. Furthermore, data on the number of psychiatric services available in the participants’ residential areas, including home-visit nursing and welfare services, which may be related to psychiatric readmission, were also used in the analysis. Statistical analysis We included participants who completed all four survey time-points and had no missing data on social functioning or readmission. After grouping based on their psychiatric readmission status, descriptive statistics for participant characteristics and outcome variables were calculated for the entire sample and each group. Data normality was tested using the Shapiro–Wilk test, sphericity using Mauchly’s test, and homogeneity of variance using Levene’s test. However, these assumptions were not fully met. Additionally, intraclass correlation coefficients (ICCs) were calculated at each time point to assess the clustering effects of the medical institutions. ICCs exceeding 0.1 were observed at several time points, indicating data non-independence and suggesting that readmission risk may be influenced by site-specific treatment factors. The social functioning variables also exhibited right skewness. Based on these findings, we constructed generalized linear mixed models (GLMMs) using a gamma distribution with a log-link function, with medical institutions included as random effects to account for this hierarchical data structure. Using the GLMMs, we conducted between-group comparisons at each time point and within-group changes across time points. All analyses were conducted separately for self-rated and observer-rated assessments of social functioning. For between-group comparisons at each time point, we examined the differences between Group 1 (reference) and the other groups. The groups were treated as fixed effects, and medical institutions were included as random intercepts to account for cluster-level heterogeneity due to hospital-level nesting. For within-group changes across time points, the GLMMs were stratified into four groups based on readmission status. In these models, the time point was treated as a fixed effect and participant ID was included as a random intercept to account for repeated measures. To address zero values in SDS scores, we transformed the data by replacing any occurrences of zero with a small positive value of 0.01, since the gamma distribution cannot accommodate zero values. For multiple comparisons, we controlled the false discovery rate (FDR) using the Benjamini–Hochberg procedure27 for each outcome. For supplementary analyses, we adjusted the GLMMs for each time point, controlling for basic demographic and clinical variables such as sex, age, and diagnosis (schizophrenia, depression, bipolar disorder, or other diagnoses). Furthermore, as a supplementary analysis, we additionally adjusted for the number of psychiatric services available in the participants’ residential areas. Adjusted within-group GLMMs were not created because of the limited sample size. Additionally, to examine the association between SDS and PSP scores, we calculated Spearman’s correlations at each time point and repeated-measures correlation across all time points. All analyses were performed using R statistical software (version 4.4.2, The R Foundation for Statistical Computing, Vienna, Austria). The packages “readxl,” “dplyr,” “tidyr,” “tableone” were used for data preprocessing and descriptive statistics; the package “car” was used for Levene’s test; the packages “dplyr,” “tidyr,”, “tibble”, “boot”, “mrcorr” were used for correlation test between SDS and PSP; the packages “ICC,” “lme4,” “lmerTest,” and “broom.mixed” packages were used for GLMMs. Ethics approval and consent to participate The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and the Helsinki Declaration of 1975, as revised in 2008. The ethical considerations of the current study, including the informed consent process and patient privacy measures, were based on the ethics guidelines for medical research in Japan. The study protocol was approved by the Ethics Committee of the Graduate School of Medicine at Chiba University (No. 3154). Results In total, 484 patients participated at all time points. After excluding those who had missing social functioning variables, 483 were included in the analysis. The participant grouping according to readmission status was: Group 1 (n = 16, 3.31%), Group 2 (n = 37, 7.66%), Group 3 (n = 40, 8.28%), and Group 4 (n = 390, 80.74%). The characteristics of the participants in each group are shown in Table 1. The mean age was 42.20 years (standard deviation = 10.71). More than half of the participants were female (58.18%). The most common primary diagnosis was schizophrenia with an F2 code (59.01%). This was followed by major depression (13.87%) and bipolar disorder (13.46%). The prevalence of primary diagnoses varied among groups. The average number of psychiatric services available in the participants’ residential areas was 72.17 (Group 1: 83.38 ± 62.90; Group 2: 60.43 ± 58.31; Group 3: 81.47 ± 60.29; Group 4: 71.87 ± 63.05). Table 1Characteristics of participants per group based on readmission status (n = 483) a.All(n = 483)Group 1(n = 16)Group 2(n = 37)Group 3(n = 40)Group 4(n = 390) n % n % n % n % n %Age (years, mean ± SD)42.2010.7143.8112.0044.599.7844.409.5341.6810.82Female28158.18850.002464.862255.0022758.21Primary diagnosis based on the ICD-10 classification b F0: Mental disorders due to known physiological conditions81.6600.0012.7000.0071.79 F1: Mental and behavioral disorders due to psychoactive substance use132.6900.0012.7037.5092.31 F2: Schizophrenia, schizotypal, delusional, and other non-mood psychotic disorders28559.011275.002567.572357.5022557.69 F3: Mood [affective] disorders (except for F30/F31)6713.87212.5025.41820.005514.1 F30/F31: Mood [affective] disorders (Manic episode, Bipolar affective disorder)6513.4600.00410.81512.505614.36 F4: Anxiety, dissociative, stress-related, somatoform, and other nonpsychotic mental disorders204.1400.0012.7012.50184.62 F5: Behavioral syndromes associated with physiological disturbances and physical factors61.2416.2500.0000.0051.28 F6: Disorders of adult personality and behavior51.0400.0000.0000.0051.28 F7: Intellectual disabilities81.6616.2538.1100.0041.03 F8: Pervasive and specific developmental disorders51.0400.0000.0000.0051.28 F9: Behavioral and emotional disorder with onset usually occurring in childhood and adolescence10.2100.0000.0000.0010.26The number of psychiatric services available in the participants’ residential areas (mean ± SD)72.1762.4683.3862.9060.4358.3181.4760.2971.8763.05a The four groups were classified according to readmission status at six months after discharge (T3) and 12 months after discharge (T4): Group (1) readmission at both T3 and 4, Group (2) readmission at T3 only, Group (3) readmission at T4 only, and Group (4) no readmission.b ICD-10: The International Classification of Diseases 10th Revision. Characteristics of participants per group based on readmission status (n = 483) a. a The four groups were classified according to readmission status at six months after discharge (T3) and 12 months after discharge (T4): Group (1) readmission at both T3 and 4, Group (2) readmission at T3 only, Group (3) readmission at T4 only, and Group (4) no readmission. b ICD-10: The International Classification of Diseases 10th Revision. Figure 1 shows changes in SDS and PSP scores from T1 to T4; Supplementary Table S1 shows the descriptive statistics of the SDS and PSP at each time point for each group. At T1, SDS scores were highest for all groups (Group 1: 16.81 ± 7.88; Group 2: 16.43 ± 7.24; Group 3: 16.55 ± 7.73; Group 4: 16.13 ± 8.08) and PSP scores were lowest for all groups (Group 1: 46.88 ± 18.45; Group 2: 46.81 ± 15.15; Group 3: 43.70 ± 20.98; Group 4: 48.41 ± 18.68). The SDS and PSP scores showed different patterns of change depending on readmission status and time. Spearman correlation analyses revealed no significant association between SDS and PSP scores at T1 (ρ = −0.04, p = 0.39), whereas significant correlations were observed at T2–T4 (T2: ρ = −0.13, p = 0.004; T3: ρ = −0.34, p < 0.001; T4: ρ = −0.33, p < 0.001). A repeated-measures correlation analysis across all time points also demonstrated a significant association (ρ = −0.36, p < 0.001). Changes of self-rated and observer-rated functional assessment from index admission and discharge to 12 months after discharge. Table 2 shows the between-group comparisons of self-rated and observer-rated functional assessments at each time point. Two significant differences were observed in PSP. At T3, Group 4 demonstrated significantly higher PSP scores than Group 1 (b = 0.25, standard error [SE] = 0.07, adjusted p = 0.002). At T4, Group 4 showed significantly better functioning than Group 1 (b = 0.32, SE = 0.07, adjusted p < 0.001). No significant differences were observed in SDS scores between groups. These trends were similar in the models adjusted for sex, age, and diagnosis (Supplementary Table S2). Furthermore, when additionally adjusting for the number of psychiatric services available in the participants’ residential areas, the same pattern was observed. At T3, Group 4 demonstrated significantly higher PSP scores than Group 1 (b = 0.25, standard error [SE] = 0.07, p < 0.001). At T4, Group 4 showed significantly better functioning than Group 1 (b = 0.32, SE = 0.07, adjusted p < 0.001). Table 2Differences in self-rated and observer-rated functional assessment between groups based on psychiatric readmission status, by time point (n = 483).TimecGroupdSheehan disability scale (self-assessment) aPersonal and social performance scale (observer-assessment) bbSE95%CIP-valueAdjustedP-valueebSE95%CIP-valueAdjustedP-valueeT1Group 1refrefGroup 2-0.020.25-0.510.470.940.94-0.020.13-0.270.230.880.97Group 3-0.020.25-0.510.460.920.94-0.090.12-0.330.160.480.73Group 4-0.050.21-0.460.370.820.940.030.11-0.180.240.750.93T2Group 1refrefGroup 2-0.110.33-0.760.540.730.940.000.07-0.140.140.990.99Group 3-0.220.33-0.860.420.490.940.060.07-0.080.190.400.68Group 4-0.200.28-0.750.360.490.940.110.060.000.230.060.24T3Group 1refrefGroup 20.110.38-0.640.850.780.940.020.08-0.130.180.770.93Group 3-0.250.38-0.990.490.510.940.070.08-0.090.220.390.68Group 4-0.200.32-0.830.440.540.940.250.070.120.38< 0.001*0.002*T4Group 1refrefGroup 2-0.270.39-1.040.490.480.940.110.08-0.040.270.160.38Group 3-0.100.38-0.850.650.790.940.130.08-0.020.280.090.28Group 4-0.370.33-1.020.280.270.940.320.070.190.45< 0.001*< 0.001*SE: standard error, * Significant at the 0.05 level. Generalized linear mixed models were conducted by each time point as an indicator of group difference.a Sheehan disability scale is self-rated scale with the lower scores indicating better functioning and fewer difficulties (range = 0–30).b Personal and social performance scale is observer-assessment tool with the higher scores indicating better functioning and fewer difficulties (range = 1–100).c Assessments were conducted at index admission (T1), discharge from index admission (T2), and 6 months (T3) and 12 months (T4).d The four groups were classified according to readmission status at T3 and T4: Group (1) readmission at both T3 and 4, Group (2) readmission at T3 only, Group (3) readmission at T4 only, and Group (4) no readmission.e For multiple comparisons, p-values were adjusted using the Benjamini–Hochberg procedure to control the false discovery rate for each outcome. Differences in self-rated and observer-rated functional assessment between groups based on psychiatric readmission status, by time point (n = 483). SE: standard error, * Significant at the 0.05 level. Generalized linear mixed models were conducted by each time point as an indicator of group difference. a Sheehan disability scale is self-rated scale with the lower scores indicating better functioning and fewer difficulties (range = 0–30). b Personal and social performance scale is observer-assessment tool with the higher scores indicating better functioning and fewer difficulties (range = 1–100). c Assessments were conducted at index admission (T1), discharge from index admission (T2), and 6 months (T3) and 12 months (T4). d The four groups were classified according to readmission status at T3 and T4: Group (1) readmission at both T3 and 4, Group (2) readmission at T3 only, Group (3) readmission at T4 only, and Group (4) no readmission. e For multiple comparisons, p-values were adjusted using the Benjamini–Hochberg procedure to control the false discovery rate for each outcome. As shown in Table 3, within-group analyses across the time points revealed significant improvements in both self-rated and observer-rated functional outcomes. For PSP scores, Group 1 showed a significant increase from T1 to T2 (b = 0.26, SE = 0.12, adjusted p = 0.03). In Group 2, PSP scores were significantly higher at T2 (b = 0.25, SE = 0.07, adjusted p < 0.001), T3 (b = 0.14, SE = 0.07, adjusted p = 0.04) and T4 (b = 0.21, SE = 0.07, adjusted p = 0.002), compared with T1. Group 3 exhibited significant increases at T2 (b = 0.44, SE = 0.07), T3 (b = 0.29, SE = 0.07), and T4 (b = 0.32, SE = 0.07), with all with adjusted p-values < 0.001. Similarly, Group 4 demonstrated significant improvements at T2 (b = 0.37, SE = 0.02), T3 (b = 0.35, SE = 0.02), and T4 (b = 0.38, SE = 0.02), all with adjusted p-values < 0.001. For SDS scores, only Group 4 showed significant reductions at T2 (b = − 0.32, SE = 0.08), T3 (b = − 0.41, SE = 0.08), and T4 (b = − 0.49, SE = 0.08), all with adjusted p-values < 0.001, whereas no significant changes were observed in Groups 1 − 3. Table 3Changes over time in self-rated and observer-rated functional assessment by each group based on psychiatric readmission status.GroupcTimedSheehan disability scale (self-assessment) aPersonal and social performance scale (observer-assessment) bbSE95%CIP-valueAdjustedP-valueebSE95%CIP-valueAdjustedP-valueeGroup 1T1refref(n = 16)T2-0.170.22-0.600.270.450.500.260.120.040.490.02*0.03*T3-0.250.22-0.690.180.250.430.130.12-0.100.360.270.29T4-0.170.22-0.600.270.450.500.110.12-0.120.340.360.36Group 2T1refref(n = 37)T2-0.260.26-0.770.250.320.440.250.070.130.38< 0.001*< 0.001*T3-0.120.26-0.630.390.630.630.140.070.010.270.03*0.04*T4-0.410.26-0.920.100.110.280.210.070.080.340.002*0.002*Group 3T1refref(n = 40)T2-0.380.26-0.890.120.140.280.440.070.300.58< 0.001*< 0.001*T3-0.490.26-1.000.020.060.170.290.070.150.43< 0.001*< 0.001*T4-0.250.26-0.760.260.330.440.320.070.180.46< 0.001*< 0.001*Group 4T1refref(n = 390)T2-0.320.08-0.49-0.16< 0.001*< 0.001*0.370.020.330.40< 0.001*< 0.001*T3-0.410.08-0.57-0.25< 0.001*< 0.001*0.350.020.320.39< 0.001*< 0.001*T4-0.490.08-0.65-0.32< 0.001*< 0.001*0.380.020.350.42< 0.001*< 0.001*SE: standard error, * Significant at the 0.05 level. Generalized linear mixed models were conducted by each time point as an indicator of group difference.a Sheehan disability scale is self-rated scale with the lower scores indicating better functioning and fewer difficulties (range = 0–30).b Personal and social performance scale is observer-assessment tool with the higher scores indicating better functioning and fewer difficulties (range = 1–100).c The four groups were classified according to readmission status at T3 and T4: Group (1) readmission at both T3 and 4, Group (2) readmission at T3 only, Group (3) readmission at T4 only, and Group (4) no readmission.d Assessments were conducted at index admission (T1), discharge from index admission (T2), and 6 months (T3) and 12 months (T4).e For multiple comparisons, p-values were adjusted using the Benjamini–Hochberg procedure to control the false discovery rate for each outcome. Changes over time in self-rated and observer-rated functional assessment by each group based on psychiatric readmission status. SE: standard error, * Significant at the 0.05 level. Generalized linear mixed models were conducted by each time point as an indicator of group difference. a Sheehan disability scale is self-rated scale with the lower scores indicating better functioning and fewer difficulties (range = 0–30). b Personal and social performance scale is observer-assessment tool with the higher scores indicating better functioning and fewer difficulties (range = 1–100). c The four groups were classified according to readmission status at T3 and T4: Group (1) readmission at both T3 and 4, Group (2) readmission at T3 only, Group (3) readmission at T4 only, and Group (4) no readmission. d Assessments were conducted at index admission (T1), discharge from index admission (T2), and 6 months (T3) and 12 months (T4). e For multiple comparisons, p-values were adjusted using the Benjamini–Hochberg procedure to control the false discovery rate for each outcome. Figure 1 shows changes in self-rated (Sheehan Disability Scale, SDS) and observer-rated (Personal and Social Performance Scale, PSP) functioning scores from index admission to 12 months after discharge from index admission. Discussion This study examined social functioning in four groups based on psychiatric readmission status and four time points (i.e., between-group comparisons at each time point and within-group changes across time points) using 12-month longitudinal data. This allowed us to address three research gaps: (1) the possibility that the association between social functioning and readmission may be interactive, (2) how social functioning changes over time, and (3) the differences between self-rated and observer-rated social functioning. For both self-rated and observer-rated functioning at index admission, scores were similar across groups. However, at subsequent time points, the pattern of changes in social functioning differed between groups based on readmission status. These findings suggest that the association between social functioning and readmission may not be fully captured using simple predictive models. In addition, self-reported and observer-rated functioning were correlated at time points after discharge but not at index admission. However, the trajectories of functional change across readmission groups differed between self-reported and observer-rated assessments. A notable research gap is the difference between self-rated and observer-rated social functioning. The absence of a significant correlation between SDS and PSP scores at T1 suggests a substantial discrepancy between self-rated and observer-rated functioning of social functioning at emergency psychiatric admission. In contrast, the significant correlations observed from T2 onward may indicate that self-rated and observer-rated assessments become more closely aligned following discharge. However, the present study further demonstrated that self-rated and observer-rated evaluations of social functioning show distinct characteristics in relation to readmission risk. Specifically, we found that, overall, functional improvements were more pronounced in observer-rated functioning, whereas self-rated functioning showed limited change. For self-rated functioning, a significant difference was observed between admission and post-discharge time points, but only in the non-readmission group. In other words, self-rated functioning showed different trajectories across groups defined by readmission status, with greater improvement observed in the non-readmission group than in the readmission group. Regarding the observer-assessment, scores improved between index admission and discharge in all groups. This might suggest that clinical staff recognized improvements in social functioning resulting from successful inpatient treatment. At post-discharge follow-ups, scores showed significant improvement at all time points in all groups except for the group readmitted at both follow-ups. Clinical staff who administered the PSP may have been able to grasp only limited aspects of patients’ daily lives through questioning. Moreover, in this study, clinical staff who rated the PSP might have been aware of patients’ readmission status, which could have influenced PSP evaluations. Prior research has reported that agreement on the perception of social service needs is low between clients and staff28, and that there may also be differences in the aspects emphasized during assessments of social functioning. Therefore, our findings indicate that, even when clinical staff perceive an improvement in social functioning, individuals might not feel that their social functioning has improved to the same extent and may have unexpected difficulties in their daily life. Social functioning should be assessed from the perspectives of both patients and clinical staff. The findings of this study also suggest that the association between social functioning and readmission may be interactive and may not be fully captured by simple predictive models. The non-readmission group (Group 4) showed the most pronounced changes in social functioning in both self-assessment and observer-assessments. Compared with baseline (at admission), social functioning improved at all time points following discharge. In other words, higher social functioning at discharge and its maintenance during community living were observed in the non-readmission group. According to previous studies, low social functioning is a risk factor for readmission8,13, and community living (avoiding hospitalization) leads to improved social functioning18–20. Participants in the non-readmission group might have maintained high levels of social functioning after discharge through these mechanisms. However, at discharge, no significant differences in social functioning scores were observed between the groups. In other words, it was difficult to predict the occurrence of readmission over the subsequent 12 months based on social functioning at discharge. Social functioning may or may not change frequently, depending on the situation. In Groups 2 and 3, which were readmitted only once at six months or 12 months follow-ups, social functioning scores fluctuated. This may be due to an interactive influence of functioning and readmission, as previously mentioned, or other factors. Meanwhile, in Group 1 (patients readmitted at both follow-ups), although the PSP score at discharge was significantly higher than at admission, subsequent scores showed no significant changes at follow-ups after discharge. This supports the findings of a previous systematic review that reported an association between low social functioning and the revolving door phenomenon9. One possible reason for this is that low social functioning makes it difficult to continue treatment in communities without intensive care. According to a large-scale epidemiological study, even among people with a mental disorder lasting more than 12 months, less than 25% continue treatment, indicating that barriers to treatment continuation and healthcare facility visits persist29. Furthermore, low social functioning leads to noncompliance with medication or follow-ups9. Previous studies also have shown that timely follow-up visits reduce the risk of readmission7,29. In Japan, intensive mental health outreach services delivered by multidisciplinary teams have gradually expanded30–32, and these services can potentially improve social functioning32,33. Despite this development, the majority of individuals still do not have access to such services in their communities34. Our data also showed that the readmission group had the highest number of community resources, and the results remained unchanged even after adjustment for the number of community resources. These results may reflect this situation. If intensive care is well developed in a community, the assessment of social functioning and the role of readmission may differ. The stability of social functioning scores may be influenced not only by readmission itself but also by the impact of such community support systems. Strengths and limitations This study had two strengths. First, both self-rated and observer-rated functioning were evaluated across four time points from admission to 12 months after discharge, enabling an understanding of longitudinal functional trajectories. Second, we created groups based on readmission status and conducted between-group comparisons at each time point and within-group comparisons across time points using GLMMs. This enables a more comprehensive and multifaceted understanding of social functioning and readmission. This study also has limitations. First, the sample size was relatively small, particularly in Groups 1, 2, and 3. As a result, we adjusted only for basic demographic and clinical characteristics, and potential confounders could not be fully accounted for. Furthermore, due to the limited sample size, the adjusted models were not assessed for within-group GLMMs. Second, although the SDS and PSP are validated and widely used tools, they may still have limitations in capturing the full complexity of functional impairments across diverse psychiatric populations. In addition, because of differences in the items assessed by the SDS and PSP, direct comparisons between the two measures are not appropriate. Third, PSP ratings may have been influenced by information bias because clinical staff serving as raters were aware of patients’ readmission status. Future studies would benefit from assessments conducted by independent third-party evaluators. Fourth, the participating institutions in this study were recruited through snowball sampling and consisted of facilities with sufficient staff capacity to conduct the survey. This recruitment approach may have influenced the generalizability of the findings. Fifth, the readmission status between time points other than T3 and T4, specifically between T2 and T3, and between T3 and T4, was not considered, and the readmission duration was also not considered. Future studies should include more detailed information on readmissions. Sixth, this study was a secondary analysis in which groups defined by readmission status were classified post hoc. Therefore, causal interpretations cannot be made. Future prospective studies that longitudinally examine both readmission status and social functioning would be valuable. Conclusions This study examined self-rated and observer-rated social functioning in four groups based on psychiatric readmission status using 12-month longitudinal data. Social functioning scores were similar across all groups at index admission. However, at subsequent time points, differences in changes in social functioning were observed according to readmission status. Further, self-rated and observer-rated functional assessments did not show the same trends across time. In the non-readmission group, both self-rated and observer-rated social functioning improved after discharge compared with the admission period. In contrast, in groups that experienced at least one readmission, self-assessment did not show improvement even when observer-rated assessments by medical staff indicated improvement. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors thank the participants for their cooperation. Author contributions M.Iw led the conceptualization, methodology development, formal analysis, and drafting of the original manuscript. S.Y. and M.Ig contributed to the conceptualization and methodology in supporting roles. S.Y., J.K., A.M., Y.O., M.O., H.W., C.F., and K.H. supported the methodology and were responsible for investigation and data curation. S.Y. acquired funding for the study. All authors contributed to validation, interpretation of the results, and manuscript revision and editing, supported visualization efforts, and approved the final version of the manuscript. Funding This work was funded by a grant from the Japan Ministry of Health, Labour, and Welfare [grant no. 25GC1002 to SY]. Data availability The datasets generated and/or analyzed during the current study are not publicly available in accordance with the Japanese government’s Ethical Guidelines for Medical and Health Research Involving Human Subjects and the ethics committee approval for this study but are available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. Ethical approval The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and the Helsinki Declaration of 1975, as revised in 2008. The ethical considerations of the current study, including the informed consent process and patient privacy measures, were based on the ethics guidelines for medical research in Japan. The study protocol was approved by the Ethics Committee of the Graduate School of Medicine at Chiba University (No. 3154). Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets generated and/or analyzed during the current study are not publicly available in accordance with the Japanese government’s Ethical Guidelines for Medical and Health Research Involving Human Subjects and the ethics committee approval for this study but are available from the corresponding author upon reasonable request.