Symptoms Predicting Inpatient Service Use Among Patients With Bipolar Affective Disorder Deborah A. Perlick, Ph.D. Robert A. Rosenheck, M.D. John F. Clarkin, Ph.D. JoAnne Sirey, Ph.D. Patrick Raue, Ph.D.
Objective: Symptoms that were risk factors for hospital readmission among psychiatric inpatients diagnosed as having bipolar affective disorder were evaluated. Methods: Subjects were 100 persons consecutively admitted to a psychiatric inpatient unit at a university-affiliated hospital who met Research Diagnostic Criteria for bipolar I or II disorder or schizoaffective disorder, manic type. Patients were assessed using the Schedule for Affective Disorders and Schizophrenia— Lifetime Version (SADS-L) and the Brief Psychiatric Rating Scale (BPRS) within one week of discharge, and their hospitalization status was documented by monthly phone contacts over a period of 15 months. Results: Twenty-four patients (24 percent) were rehospitalized within six months of discharge, and 44 (44 percent) were readmitted within 15 months. Survival analysis using the Cox proportional hazard regression model demonstrated that patients with high scores on a BPRS-derived mania factor were at significantly decreased risk of rehospitalization, whereas those scoring high on a factor consistent with neurovegetative depression were at significantly increased risk. A greater number of previous psychiatric admissions and younger age were also associated with significantly increased risk of rehospitalization. Conclusions: The findings suggest that patients with bipolar disorder presenting with a depressive episode characterized by prominent neurovegetative features should be treated more aggressively with both pharmacotherapy and intensive outpatient services to reduce the relatively high risk of rehospitalization that appears to be associated with this type of depression. (Psychiatric Services 50:806–812, 1999)
Dr. Perlick and Dr. Rosenheck are affiliated with the Northeast Program Evaluation Center of the West Haven Veterans Affairs Medical Center, 950 Campbell Avenue, West Haven, Connecticut 06516. Dr. Perlick is also affiliated with the department of psychiatry and Dr. Rosenheck with the departments of psychiatry and epidemiology and public health at Yale University School of Medicine in New Haven, Connecticut. Dr. Clarkin, Dr. Sirey, and Dr. Raue are with New York Hospital–Westchester Division and the department of psychiatry at Weill Medical College of Cornell University in White Plains, New York.
806
S
tudies aimed at identifying risk factors for early or frequent readmission of persons with chronic mental illness to psychiatric hospitals have focused on patients’ behavior, sociodemographic characteristics, and previous use of mental health services. Behaviorally, poorer treatment compliance (1–3) and aggressive behavior (4–6) increase the likelihood of readmission. Lower socioeconomic status and financial problems (7,8), older age (9–11), single marital status (8,10), and female gender (2,7,8) are sociodemographic factors that increase the risk of rehospitalization. A greater number of previous psychiatric admissions also increases the risk (1,9,12–14), as does a shorter length of stay (13,15). Recent investigations have also demonstrated that both the degree and type of symptoms that characterize the presenting episode of illness are associated with the likelihood of rehospitalization. For example, Lyons and associates (16) reported that patients readmitted within six months of discharge had higher global levels of symptom severity at the outset of the previous admission. Similarly, Swett (12) found that higher scores just before discharge on the thought disorder factor of the Brief Psychiatric Rating Scale (BPRS) and on the BPRS self-neglect item predicted readmission within 30 days of discharge. Studies have also found that comorbid disorders or problems, in-
PSYCHIATRIC SERVICES
♦ June 1999 Vol. 50 No. 6
cluding substance use disorders or substance-related problems (3,17,18) and sexual and impulse control problems (8), increase the likelihood of readmission. By contrast, primary psychiatric diagnosis has generally not been a significant predictor of readmission (17,18). Because identification of specific symptom constellations that place patients at risk for rehospitalization might better inform clinicians’choice of treatment and disposition, studies of symptoms that are predictors of rehospitalization are of particular relevance for mental health service providers. Most studies assessing the relationship of symptoms or diagnosis to risk of rehospitalization have been conducted in mixed-diagnosis samples. Although use of diagnostically heterogeneous samples sheds light on the chronic mentally ill population in general, use of such samples might obscure factors predictive of rehospitalization in a single-diagnosis group, resulting in a type II error. This limitation may be particularly applicable to the investigation of clinical or symptom predictors of readmission. Although some studies have focused on readmission of patients with schizophrenia (9), to our knowledge symptom risk factors for readmission in a sample of patients with researchdiagnosed bipolar affective disorder have not been investigated. Because studies of high users of inpatient psychiatric services have shown that bipolar affective disorder is overrepresented among recidivist samples (19), identification of symptom risk factors for rehospitalization of patients with this diagnosis could have important social and economic, as well as clinical, ramifications. Evidence from two different lines of research on the course of illness in bipolar affective disorder suggests that symptoms of the illness may influence rehospitalization. First, researchers have found a differential course of illness for diagnostic subgroups within the bipolar spectrum. Most relevant, in the Collaborative Study of Depression conducted by the National Institute of Mental Health, patients with bipolar I disorder were more likely to be rehospitalized durPSYCHIATRIC SERVICES
ing the follow-up period than patients with bipolar II disorder (20). Second, the different types of episodes— manic or depressed— have been associated with differing rates or degree of recovery, with most studies finding that patients recover more slowly or less fully from depression than from mania or hypomania (21,22). To study the impact of symptom presentation on subsequent psychiatric hospitalization, we assessed psychiatric inpatients with bipolar affective disorder that had been diagnosed based on Research Diagnostic Criteria, and we followed their hospitalization status prospectively for a period of 15 months after discharge. Based on the results of previous studies, we hypothesized that the nature of symptom presentation at hospital discharge would predict the likelihood of rehospitalization over the 15-month follow-up period.
Methods Subjects
The study sample consisted of 147 subjects consecutively admitted to an acute psychiatric inpatient service who were age 16 or older and who met Research Diagnostic Criteria (23) for several disorders. These disorders were bipolar depression with mania (bipolar I disorder), hypomania (bipolar II disorder), or schizoaffective disorder, manic type, based on the Schedule for Affective Disorders and Schizophrenia— Lifetime Version (SADS-L) (24). Because the study reported here was carried out as part of a study on family burden in bipolar illness (25), only patients whose family members consented to the large study were included. Patients were enrolled in the study between October 1993 and September 1995, and follow-up data were collected for 15 months after discharge (approximately through December 1996). A total of 100 persons completed the 15-month follow-up, and this group was the sample in the study reported here. The 32 percent attrition rate is equivalent to the 32 percent rate observed for the Collaborative Study for Depression (26). Comparison of the 100 subjects who completed the study and the 47 noncompleters did
♦ June 1999 Vol. 50 No. 6
Table 1
Baseline characteristics of 100 inpatients with bipolar disorder followed for 15 months after hospital discharge1 Characteristic Demographic Female Age (mean±SD years) Race or ethnicity White Black Hispanic Asian Other Socioeconomic status2 I II III IV V Living with family Clinical Diagnosis3 Bipolar I disorder Bipolar II disorder Schizoaffective disorder, manic type Length of illness (mean± SD years) Index episode characterized by mania4 N of lifetime psychiatric admissions (mean±SD) Total score on Brief Psychiatric Rating Scale (mean±SD) 5 1 2
3 4 5
N
%
60 56 38.2±14.6 79 13 5 2 1
79 13 5 2 1
2 13 31 21 16 56
2 16 37 25 19 52
49 10
48 10
44
43
16.8±11.1 47
60
7.9±11.7 39.9±13.0
Data on some items were missing for some patients. Based on Hollingshead and Redlich’s twopoint scale (32); level I indicates higher socioeconomic status. Based on the Schedule for Affective Disorders and Schizophrenia— Lifetime Version Combines mania, hypomania, and schizoaffective mania Possible scores range from 18 to 75, with higher scores indicating higher levels of psychopathology.
not reveal any significant differences in baseline sociodemographic and clinical variables. Measures
Within one week of discharge, patients were assessed on diagnostic measures using the SADS-L. In addition, the expanded version of the Brief Psychiatric Rating Scale (27), which was developed by Lukoff and colleagues (28) to incorporate the psychotic and affective symptoms associated with bipolar disorder, was 807
Table 2
Loadings on four factors resulting from a factor analysis of the expanded Brief Psychiatric Rating Scale Item
Mania
Psychosis
Cognitive depression
Vegetative depression
Depression Elevated mood Guilt Suicidality Anxiety Somatic concern Hostility Suspiciousness Unusual thought content Grandiosity Hallucinations Disorientation Self-neglect Motor retardation Motor hyperactivity Excitement Tension Speech coherence Distractibility Blunted affect Emotional withdrawal Uncooperativeness Mannerisms and posturing Bizarre behavior
–.167 .482 –.080 –.113 .110 .108 .411 .245 .270 .451 –.115 –.082 .180 .230 .761 .791 .280 .591 .315 –.198 –.123 .241 .350 .412
–.019 .415 .111 .240 .095 –.129 .026 .579 .833 .647 .725 .408 .137 .085 –.087 .037 .239 .224 .065 .104 .062 –.005 –.006 .341
.755 –.118 .681 .663 .785 .409 .356 .216 .096 –.025 .127 –.007 .339 .155 .054 –.063 .092 –.061 .161 .029 –.022 –.143 –.081 .096
.138 –.055 –.009 –.010 .039 –.072 –.174 .122 .050 –.085 .147 .036 .220 .712 –.036 –.089 .418 .124 .339 .774 .714 .385 .517 .105
used to rate symptom type and severity. The 24-item instrument uses a 7point scale and is administered by patient interview. An intraclass correlation coefficient was calculated for the four raters based on two videotaped interviews, using all 24 scales. Coefficients for tape 1 were .83, .82, .85,
Analyses
and .96; for tape 2, coefficients were .88, .87, 1, and .90. Internal consistency for the BPRS was acceptable (Cronbach’s alpha=.76). Data on the number and timing of previous psychiatric inpatient admissions were obtained from the SADSL interview. Data on months of com-
Table 3
Results of Cox regression analysis of time to rehospitalization among 100 patients with bipolar disorder Covariate Brief Psychiatric Rating Scale factor1 Mania Psychosis Cognitive depression Vegetative depression N of previous inpatient psychiatric admissions Index episode characterized by mania or major depression2 Lives with family caregiver Gender Age Socioeconomic status 1 2
Coefficient
SE
p
–5.809 1.065 –.022 3.716 .790
2.607 1.552 .177 1.727 .391
.026 .492 .903 .031 .043
1.138 .452 .132 –1.411 .017
.591 .377 .342 .550 .011
.054 .230 .700 .010 .126
Based on the expanded 24-item Brief Psychiatric Rating Scale (27) Compared with patients whose index episode was characterized by hypomania or minor depression
808
munity tenure and on psychiatric rehospitalization were collected over 15 months of follow-up by monthly phone or in-person contact with patients. When patients were not available, the contact was with a family member. The frequency and rate of subsequent psychiatric readmissions for the sample as a whole were calculated. A principal components analysis was performed on the 24-item BPRS data for purposes of data reduction. Although the BPRS has previously been factor analyzed, past studies either employed the original 18-item scale or used a general sample of severely mentally ill persons (29,30) and the results were thus not suited to the aims of this investigation. Survival analysis using the Cox proportional hazard regression model (31) was employed to describe the relationship between all predictor variables (the BPRS symptom factors and other covariates) and time to rehospitalization over the 15-month followup period. This method allowed us to use all available data, including censored observations (that is, data for patients who were not hospitalized during the 15 months). Time to rehospitalization was measured in number of months from the month of the index discharge until the month of the initial rehospitalization. In this study, a hospital readmission was the outcome variable of interest, and a community stay without readmission for the 15-month study period was a censored event. Hospital readmissions occurring before day 25 within the month of discharge were considered to be equal to a halfmonth. Preliminary bivariate analyses were conducted using the Pearson r product-moment correlation coefficient to help select variables for the model. To maximize power, we limited our model to include the four BPRS factors, sociodemographic variables (age, gender, socioeconomic status [32], and living or not living with the identified family caregiver [25]), the total number of previous psychiatric hospitalizations, and whether or not the index episode met SADS-L crite-
PSYCHIATRIC SERVICES
♦ June 1999 Vol. 50 No. 6
ria for major depression or mania versus minor depression or hypomania. The latter variable was included to control for the overall level of psychopathology— that is, to ensure that an observed effect of one or more of the specific BPRS factors was not simply a reflection of the patients’ overall severity of illness. Two additional SADS-L variables, a diagnostic variable that classified patients as schizoaffective, manic type, versus all others SADS-L diagnoses and a variable that classified the index episode as either mania or hypomania versus major or minor depression (pole of index episode), were not significantly associated with the dependent variable at the bivariate level and thus were not included in the model. Similarly, the BPRS total score was not significantly associated with the dependent variable and was not included in the model. Because we were interested in evaluating the contribution of the four BPRS factors while controlling for the effects of sociodemographic variables and overall level of psychopathology, all covariates were entered simultaneously. SPSS software programs (33) were used to perform the analyses.
psychosis, had high loadings on unusual thought content, hallucinations, grandiosity, and suspiciousness. Two depression factors were found. Factor 3, cognitive depression, had high loadings on anxiety, depressed mood, guilt, and suicidality, and factor 4, vegetative depression, had high loadings on blunted affect, motor retardation, and emotional withdrawal.
Results
Rehospitalization
Sample characteristics
Forty-four of the 100 patients in the sample had at least one psychiatric hospital admission during the 15 months of follow-up, and 56 stayed in the community without rehospitalization. Of the patients who were rehospitalized, 15, or 34 percent, were admitted within the first three months of the study; 24, or 55 percent, were admitted within the first six months; and 39, or 89 percent, were admitted within the first year. Overall, the variables in the model were highly effective at predicting rehospitalization (χ2=31.32, df=10, p< .001). Table 3 presents the regression coefficients for the predictor variables in the model and related statistics. Older patients and those with manic symptom profiles had a significantly higher probability of not being rehospitalized over the 15 months, while patients with symptom profiles consistent with neurovegetative depression and a greater number of previous psychiatric admissions were at
Characteristics of the sample are presented in Table 1. As a group, subjects were representative of the more chronic end of the bipolar spectrum, having been hospitalized on average almost twice a year during their 16.8 years of illness. Patients of both genders, those with manic as well as depressive index episode poles, and those with schizoaffective disorder as well as bipolar I or II subdiagnoses were well represented in the sample. Most patients were Caucasian and middle class and lived with one or more family members. Age was well distributed in the study sample. BPRS factor analysis
The principal components analysis yielded four factors accounting for a cumulative variance of 46 percent. As shown in Table 2, factor 1, mania, had high loadings (>.45) on excitement, motor hyperactivity, speech coherence, and elevated mood. Factor 2, PSYCHIATRIC SERVICES
Figure 1
Probability of remaining unhospitalized among low-, average-, and high-risk patients with bipolar disorder during 15 months after hospital discharge
♦ June 1999 Vol. 50 No. 6
significantly greater risk for rehospitalization. Patients whose index episode met criteria for a major affective episode of any type (major depression, mania, or schizoaffective disorder) were at marginally greater risk for rehospitalization (p