Table 1 presents the distribution of characteristics according to the diagnosis. A total of 2,009 repeatedly hospitalized patients were enrolled in this study, of which 1,179 (58.7%) and 830 (41.3%) were diagnosed with BD and depression, respectively, at the time of initial hospitalization. The mean age of the study population was 37.41 ± 15.25 years, with the majority of patients being female (62.4%), residing in Beijing (60.0%), Han Chinese (95.9%), and married (54.6%). 71.2% of the study participants were hospitalized twice, 19.0% were hospitalized three times, and 9.8% were hospitalized four or more times. The mean duration of all patients was 5.26 years, and 4.87 years, 5.80 years in patients with BD and depression, respectively. 1174 (58.4%) patients accepted the treatment of MECT, of which 63.9% was BD patients.

Mean LOS was significantly higher for mood disorder patients as the number of hospitalizations increased. The mean LOS of patients with mood disorders was 32.12 ± 22.70 days, 32.57 ± 21.98 days, 36.60 ± 33.42days and 39.40 ± 65.06 days at the first, second, third, and fourth or above hospitalizations, respectively. Moreover, the mean LOS of patients with depression was consistently longer than that of patients with BD until the fourth or above hospitalization (Fig. 1).

The change in mean LOS among repeatedly hospitalized patients with mood disorders

As shown in Table 2, the comparison of baseline characteristics between the short LOS group and long LOS group of repeatedly hospitalized patients with mood disorders demonstrated statistically significant differences in various factors. These included occupation, monthly income, ethnicity, marital status, interpersonal relationship, whether have children, history of respiratory diseases, history of vaccinations, smoking, drinking, self-care ability, adverse effects of treatment, and precipitating factor of diseases, MECT, treatment effectiveness, whether use antipsychotics, whether use anxiolytics(P<0.050).

Using all hospitalization data of repeatedly hospitalized patients with mood disorders, a GEE model was employed to incorporate sociodemographic characteristics, medical history, disease-related characteristics and treatment-related characteristics. The short LOS group was taken as the reference group in the GEE model. Figure 2A presents the results of the GEE model on the influencing factors of LOS among total patients. In terms of demographic variables, the study revealed that ethnic minority patients tended to have short LOS (OR = 0.650, 95%CI: 0.468, 0.901). And those with good interpersonal relationships (OR = 0.778, 95%CI: 0.606, 0.999) were obviously less likely to develop long LOS compared to patients with poor interpersonal relationships.

GEE model of LOS among repeatedly hospitalized patients with different disease types

And in terms of medical history, history of vaccinations was strongly correlated with LOS. Patients with a history of vaccinations were more likely to have short LOS compared to unvaccinated patients, with an OR of 0.527 (95%CI: 0.420, 0.660). As for disease-related characteristics, patients who experienced adverse effects of treatment (OR = 1.713, 95%CI: 1.508, 1.945), those with continuous clinical characteristics (OR = 1.208, 95%CI: 1.031, 1.416), and those who had suicide attempt (OR = 1.352, 95%CI: 1.082, 1.689) showed a dramatically higher risk of long LOS compared to patients without adverse effects of treatment, presented intermittent clinical characteristics and never committed suicide. A negative correlation was found between disease precipitating factor (OR = 0.852, 95%CI: 0.743,0.976) and LOS. Additionally, comorbidity, treatment effectiveness and whether use antidepressants also affected the LOS among mood disorder patients. Patients who had comorbidity were more likely to stay longer in hospital (OR = 1.198, 95%CI: 1.028, 1.397).

Further analysis by different types of mood disorders indicated that there were discrepancies in the factors affecting LOS in patients with depression and patients with BD, as shown in Fig. 2B and C. Depressed patients who were elderly (OR = 1.016, 95%CI: 1.004,1.028), had low monthly income (OR = 1.488, 95%CI: 1.013,2.185), experienced adverse effects of treatment (OR = 1.822, 95%CI: 1.474,2.251), had continuous clinical characteristics (OR = 1.332, 95%CI: 1.044,1.699), attempted suicide (OR = 1.391, 95%CI: 1.039,1.861) and had comorbidity (OR = 1.328, 95%CI: 1.000,1.765) tended to have longer hospital stays. Conversely, depressed patients with a history of vaccinations (OR = 0.392, 95%CI: 0.288,0.534), and had the ability to care for themselves (OR = 0.226, 95%CI: 0.084,0.608) tended to have short LOS. As for treatment effectiveness, improvement (OR = 0.652, 95%CI: 0.466,0.912) and not healed (OR = 0.088, 95%CI: 0.037,0.210) was associated with short LOS among patients with depression. Slightly different from patients with depression, for BD patients, factors such as monthly income, residence, ethnicity, interpersonal relationship, food allergies history, vaccination history, adverse reaction in the treatment, hospitalization frequency, treatment effectiveness and whether use antidepressants were found to associated with LOS. BD patients who experienced adverse effects of treatment (OR = 1.677, 95%CI: 1.425,1.973), the number of repeated hospitalizations ≥ 4 (OR = 1.298, 95%CI: 1.004,1.678) and used antidepressants (OR = 1.554, 95%CI: 1.233,1.958) were more likely to stay longer in hospital.

Stratified analyses were performed based on different treatment effectiveness at the initial hospitalisation, but patients who were not healed were excluded due to the limited sample size and lack of validity of the analyses. As Fig. 3A shows, cured patients who were married, divorced or widowed stayed longer than those who were single. The risk factors of long LOS also included history of cardiovascular diseases, adverse effects of treatment and continuous clinical manifestation. History of vaccination was negatively related to the long LOS. Among the improved patients who were minority, not resident in Beijing, married, had the history of vaccination and precipitating factor tended to be stay shorter in hospital. The adverse effects of treatment, suicide attempt, comorbidity, use of moodstabilizers and antidepressants were positively associated with long LOS in hospital.

GEE model of LOS among patients with different treatment effectiveness