Abstract Major depressive disorder is a common psychiatric condition. Hospitalization is usually indicated for patients with more severe symptoms and severe functional impairment. Rehospitalization is known as the re-emergence of significant depressive symptoms. The purpose of the present study was to investigate the risk factors affecting time to rehospitalization. Rehospitalization status was monitored for all patients with major depressive disorder discharged from Kai-Suan Psychiatric Hospital between 1 January 2002 and 31 December 2003. Patients were followed up with respect to rehospitalization until 31 December 2004. The Kaplan–Meier method was used to calculate the median time to rehospitalization. Risk factors associated with rehospitalization were examined on Cox proportional hazards regression. Three hundred patients were recruited. Median time to readmission was 174 days (SD = 37). Comorbid alcohol abuse/dependence (hazard ratio [HR] = 1.841, 95% confidence interval [CI] = 1.229–2.758, P < 0.01), comorbid personality disorders (HR = 1.530, 95%CI = 1.053–2.223, P < 0.05), and the number of previous hospitalizations (HR = 1.121, 95%CI = 1.056–1.190, P < 0.001) were found to be predictors of the shorter time to rehospitalization over the 360-day study. Further research should be carried out to test risk factors in a prospective study, and to study the cost-effectiveness of interventions to reduce risk factors and rehospitalizations.
Major depressive disorder is a common psychiatric condition. For many, this is a recurrent mental disorder with lifetime rates as high as 17–19%.1 Studies suggest that ≥50% patients with an initial depressive episode may have recurrence.2
Most of the patients with major depressive disorder accept treatment in an outpatient setting, with only a small percentage of patients requiring hospitalization.3 Hospitalization is usually indicated for patients who have a greater number of, or more severe, symptoms and severe functional impairment, are more likely to have psychotic features, and are more likely to be seriously suicidal. Hospitalization also represents the category of care with the greatest direct cost.
Rehospitalization is known as the re-emergence of significant depressive symptoms or dysfunction that do not respond adequately to outpatient treatment following a discharge. Compared to schizophrenic disorders, fewer studies about rehospitalization for major depressive disorder have been published. Haywood et al. found that recidivist patients with unipolar depression were admitted as often as patients with schizophrenia, and that substance abuse and medication non-compliance was associated with rehospitalization.4 Other studies have shown that frequently admitted patients were more likely to be older, female, married, suffer from recurrent depression, have the number of previous hospitalizations, have comorbid personality disorders, comorbid alcohol abuse/dependence, be non-compliant with medication, have received electroconvulsivetherapy or have a medical condition contributing to their admission.3,5–8 Social factors, such as poor education and unemployment have also been reported to be linked to rehospitalizaion.9,10 The combined serotonin–norepinephrine re-uptake inhibitor (SNRI), venlafaxine, has demonstrated better short-term efficacy than selective serotonin re-uptake inhibitors (SSRI).11–13 It is possible that venlafaxine is associated with a greater likelihood of efficacy than SSRI, which could translate to improved long-term outcomes.
In the present study time to rehospitalization was treated as the long-term outcome measure. The purpose of the present study was to investigate the clinical predictors affecting the time to rehospitalization for patients in Taiwan with major depression.
The subjects consisted of all patients, aged ≥18, with major depressive disorder (DSM-IV) diagnosed by psychiatric specialists, who had been successfully discharged from Kai-Suan Psychiatric Hospital between 1 January 2002 and 31 December 2003. Kai-Suan Psychiatric Hospital is a public mental hospital with 820 beds. It is located in Kaohsiung City, the second largest city in Taiwan, and possesses approximately 70% of the psychiatric beds in Kaohsiung City. These study subjects represent most of the major depressive inpatients in this city.
Subject rehospitalization status was examined over a 1-year follow-up period, ending 31 December 2004. A successful discharge was defined as a patient discharged because symptoms improved. After discharge, all patients were followed up in the outpatient department and received adequate treatment with SSRI or venlafaxine. Adequate treatment was defined as at least the minimum usual dose of the SSRI or venlafaxine recommended by the American Psychiatric Association Practice Guidelines.14 No special psychosocial interventions (e.g. individual psychotherapy, group psychotherapy, cognitive behavior therapy, or family therapy) or community services (e.g. home visit, halfway house, community rehabilitation center, or day hospital) were provided for any of the patients. Rehospitalization was defined as admission to Kai-Suan Psychiatric Hospital for a psychiatric reason. Chart reviews were performed by three psychiatric specialists to verify the data. The present study was approved by the human subjects committee of Kai-Suan Psychiatric Hospital as being exempt from the requirement for written informed consent.
The 13 predictor variables, including demographic data and factors on the likelihood of rehospitalization, were sex, marriage, employment, family history of depressive disorders, comorbid alcohol abuse/dependence, comorbid personality disorders (cluster A personality disorder, cluster B personality disorder, and cluster C personality disorder),15 comorbid anxiety disorders, classification of antidepressants (SSRI vs venlafaxine), age, age at onset, years of education, length of stay in hospital, and the number of previous hospitalizations. Age at onset was regarded as age at the first major depressive episode. Each subject's family history was taken, which included first-degree, second-degree, or third-degree blood relatives who had depressive disorders.
The Kaplan–Meier method was used to calculate the time to rehospitalization and follow-up time for all patients. The univariate Cox proportional hazards regression model was used to analyze the aforementioned variables. If multiple significant variables were identified, the forward multivariate Cox proportional hazards regression model was used to further analyze predictive factors for a rehospitalization. All tests were two-tailed, and significance of tests was defined as α < 0.05. Data were analyzed with SPSS version 10.0 for Windows (SPSS, Chicago, IL, USA).
Three hundred patients were recruited for the study (94 male, 206 female; mean age, 43.8 ± 13.7 years). Figure 1 presents the time to rehospitalization. Median time to rehospitalization was 174 days (SD = 37, 95% confidence interval [CI] = 101–247). Event rate was 51.3%. Median follow-up time was 276 days (SD = 9, 95%CI = 259–293). Seventy-two patients (24.0%) were lost to follow up during the 1-year period.
Table 1 shows that on univariate Cox proportional hazards regression, multiple significant variables were unemployment, comorbid alcohol abuse/dependence, comorbid personality disorders (especially cluster B personality disorder), and the number of previous hospitalizations. Table 2 shows the various SSRI or venlafaxine prescribed to the patients. Table 3 shows that after further analysis by forward multivariate Cox proportional hazards regression, comorbid alcohol abuse/dependence, comorbid personality disorders, and the higher number of previous hospitalizations were found to be the significant predictors associated with the shoter time to rehospitalization over the 360-day study period.
Table 3. Factors affecting time to rehospitalization
Forward multivariate Cox proportional hazards regression, over a 1-year follow-up period.
CI, confidence interval.
Comorbid alcohol abuse/dependence
Comorbid personality disorders
No. previous hospitalizations
According to the median time to rehospitalization (174 ± 37 days) and event rate (51.3%), it has been suggested that rehospitalization for depressive disorders is common.3 We also conclude that non-adherence to treatment is common due to the short median follow-up time (276 ± 9 days) and the high percentage of patients (24.0%) lost to follow up over 1 year.16
Except for employment status, no other demographic variables correlated with the time to rehospitalization in the present study. Similar to the results in an earlier study by Lee and Murray, neither age nor sex had a clear correlation with rehospitalization.17 Why did the present study not find a significant difference in time to rehospitalization for patients taking venlafaxine rather than SSRI? The possible explanation is that venlafaxine's serotonin re-uptake inhibition is approximately threefold higher than for noradrenergic re-uptake inhibition.18 Harvey et al. suggested that venlafaxine is a potent serotonin re-uptake inhibitor, and that by increasing the daily dose to >150 mg, it is also a potent inhibitor of norepinephrine re-uptake.19 However, most of the patients in the present study received lower doses of venlafaxine (117.2 ± 43.6 mg/day), with possible weak noradrenergic effects. It seems that they do not have the advantage of dual action, which possibly lengthens the time to rehospitalization. This result is in accordance with the theory that venlafaxine has been characterized as functioning as an SSRI at lower doses.20
Furthermore, in the present study the length of stay in hospital has no influence on the time to rehospitalization. This result was the same as that of Thompson et al.21 Unemployment, however, was likely to contribute to the association of low socioeconomic status, yielding more chronic social stressors and resulting in poor outcome.22 However after further analysis on forward multivariate Cox proportional hazards regression, it did not remain significant. The more important predictors for shorter time to rehospitalization were comorbid alcohol abuse/dependence, comorbid personality disorders, and a higher number of previous hospitalizations.
The association between comorbid alcohol abuse and rehospitalization is a relationship that has been discussed in previous studies.4,6,23 This relationship between major depressive disorder and alcohol abuse/dependence is complex. They tend to coexist.24 That is, alcohol can cause many depressive symptoms, such as low mood, agitation, apathy, suicidal ideation, loss of libido, early morning waking, loss of appetite, and weight loss.25 Patients with comorbid alcohol abuse/dependence may be attempting self-medication for underlying depressive symptoms, but comorbid alcohol abuse/dependence has a wide range of adverse impacts on the course of depression and psychosocial functioning, resulting in poor compliance with treatment, disruptive behaviors, symptom exacerbation, decreased social function, and poor prognosis. These consequences lead to relapse and rehospitalization.26
Patients with comorbid personality disorders also demonstrated shorter times to rehospitalization in the present study, perhaps because depression and comorbid personality disorders are related to one another in several different ways. Patients with personality disorders are conceptualized as having interpersonal difficulties and social skill deficits. These factors may contribute to the development and persistence of stressful and adverse life events associated with family, social, and work environments. Comorbid personality disorders may therefore modify the presentation, treatment response, or longitudinal course of depressive illness. There is much evidence that personality comorbidity adversely affects the outcome of such patients' psychotherapy27,28 and pharmacotherapy.29,30 These adverse effects have included lower rates of response or remission, slower response to treatment, increased likelihood of chronicity, and increased likelihood of relapse as well as rehospitalization.
Finally the present study, like previous studies, also found that a higher number of previous admissions is associated with a higher likelihood of rehospitalization.21,31 This finding suggests that patients who frequently seek psychiatric services often present with a range of complex, recurring problems that are not easily ameliorated, and which leave these individuals vulnerable to further crises and hospitalizations.7
The findings in the present study must be seen within the limits of the method. One limitation was that the present study was a naturalistic, retrospective study with a relatively small sample size, and with all subjects coming from the same hospital. Some of the patients lost to follow up may have been admitted to other hospitals. Such attrition of the study population due to losses through follow up might have been minimized if the register had included all facilities in a sufficiently large catchment area.
Another limitation was that the diagnoses were based on clinical impressions and not using structured instruments for Axis I or II diagnosis. Because clinicians infrequently use structured instruments in making diagnoses of personality disorders, tending instead to rely on observations of patients' attitudes, behavior, and descriptions of their interpersonal interactions,32 subjects' personality disorders or other comorbidities may be underdiagnosed based on the clinical impression.
Because of incomplete medical records or the patients' recall bias, some factors affecting rehospitalization, such as duration of episodes, numbers of episodes, psychotic depression, non-compliance with medication, adverse life events, and lack of social support, were not addressed.4,33–36 This was the other limitation of the present study.
Despite the limitations described here, the present analysis has demonstrated that the predictors were conceptually consistent with findings of previous studies. Patients with repeat admissions clearly require assistance in coping with the routine demands of living, as well as with managing their illness.7,8 Because clinical characteristics such as comorbid alcohol abuse/dependence and comorbid personality disorder are associated with rehospitalization, these clinical characteristics should be closely monitored. If these comorbid disorders are diagnosed early and treated adequately, their contribution to rehospitalization might diminish. Further research should test risk factors in a prospective study, and study cost-effective strategies to reduce such risk factors and the resulting rehospitalizations.
This study was supported by a grant from Kai-Suan Psychiatry Hospital in 2006.