The focus of psychiatric services in Japan is being shifted from hospitalization to community care, and the Ministry of Health, Labour and Welfare aims for the prompt discharge of newly-admitted patients. Correspondingly, it set a goal to lower the "mean residual rate (MRR)", which indicates the discharge dynamics of newly-admitted patients, to 24%. As a measure to achieve this goal, the present situation should be investigated in each homogeneous patient group. In this study, we conducted a survey of newly-admitted patients to investigate discharge dynamics and related factors by the diagnosis and type of hospitalization. Out of 1,459 psychiatric hospitals to which we sent questionnaires, 183 (12.5%) replied. Each hospital completed questionnaires regarding a maximum of 5 patients for each type of hospitalization (voluntary hospitalization [VH], hospitalization for medical care and protection [HMCP], and involuntary hospitalization ordered by the prefectural governor [IHOPG]) between October 2005 and January 2006. We weighted the obtained patient data in proportion to the estimated total number of patients, and analyzed valid data on 1,784 patients. The MRR for the whole sample was 29.4%. By diagnosis, dementia showed the highest MRR (45.6%), followed by schizophrenia (34.9%); depression, bipolar disorder, and alcoholism showed the lowest MRRs (20-21%). We calculated MRRs by the type of hospitalization for dementia and the other diagnoses separately, considering confounding effect between the diagnosis and type of hospitalization (markedly high proportion of HMCP observed in dementia). In dementia, HMCP showed a higher MRR (46.8%) than VH (43.7%). In the other diagnoses, IHOPG showed the highest MRR (43.7%), followed by HMCP (34.5%) and VH (25.6%). Dementia differed from the other diagnoses in the distribution of residential settings before admission, with a higher proportion of residential care facilities (25.5%) and hospitalization in other departments (19.3%). In dementia, the residential setting after discharge showed a similar distribution, and death was also frequent (6.6%). Multivariate analyses revealed that a long stay (one year or longer) was significantly associated with a residential setting before admission, the type of ward at admission, a founder (a private hospital or public/university hospital), and symptom severity at admission in schizophrenia; and with the type of ward at admission and hospital founder in dementia. In schizophrenia, the risk of a long stay was higher on hospitalization in other psychiatric hospitals (odds ratio [OR] : 28) and other departments (OR: 18), and living alone (OR: 2.1) than in living with the family by residential setting. The risk was also higher in psychiatric long-term care wards than in general psychiatric wards by the type of ward (OR: 3.0), and in private hospitals than in public/university hospitals by hospital founder (OR: 3.0). Additionally, the higher risk was associated with higher symptom severity assessed using a 6-point scale (OR: 1.3 per point). In dementia, the risk was higher in senile dementia wards than in general psychiatric wards by the type of ward (OR: 2.9), and in private hospitals than in public/university hospitals by hospital founder (OR: 6.8). The most frequently reported direct causes of a long stay were problems regarding a family's acceptance (51.5%), poor improvement of symptoms (48.8%), and poor recovery of daily living abilities (44.0%). In dementia, physical diseases (20.8%) and undecidedness of residence after discharge (29.2%) were also frequent. Considering the elapsed time after survey, the low response rate, and the data analyses with sampling bias adjustment, the results should be interpreted carefully. Nevertheless, the discharge dynamics and related factors in newly-admitted patients varied with the diagnosis and type of hospitalization. Particularly, schizophrenia and dementia, as well as IHOPG and HMCP, showed high MRRs and frequent long stays. Additionally, a long stay was related to patients' demographic and social characteristics. Adopting measures suiting patients' characteristics and arranging treatment and casework for patients at high risk of a long stay are important to facilitate community care.