Psychiatric hospitalization, the most restrictive form of health care for the patient and the most expensive for the payer, has come to be regarded as the behavioral health care of last resort, reserved only for persons who have decompensated to the extent that they are a danger to themselves or others. Research on psychiatric hospitalization often cites specific behavioral patterns that indicate a person is psychologically decompensating and heading for a psychiatric hospitalization (1,2,3,4,5).
Some people stay up all night, become socially withdrawn, and lose their appetites. Others hear voices, experience excessive anger, and become overtly hostile. Still others increase their use of alcohol and drugs. Reports from family members and friends suggest that the behavioral warning signs, also known as prodromal symptoms, are consistent across decompensation episodes and that decompensating people are often unaware of their unusual behavior or that they are heading for a hospitalization (6,7).
The goals of this study were to examine such warning signs among adults receiving outpatient treatment for severe mental illness, assess the prevalence of problems they had in being unable to recognize warning signs in themselves (poor warning-sign recognition), and explore the relationships between poor warning sign recognition, treatment for poor recognition, service use, and treatment costs.
This study is part of a larger project aimed at validating a new outcomes evaluation system called the Client Progress Assessment System (8,9). In this project, which began in April 1994, researchers are following the clinical and life progress and treatment experience of 735 adults who have been treated for severe mental illness at an urban community mental health center (CMHC). The center provides treatment to about 3,000 individuals a year. For the larger project, a total of 430 clients were randomly selected from the CMHC's outpatient providers' caseloads, 200 more were consecutive new admissions (individuals who had not received services from the facility during the preceding 12 months), and 105 were active outpatients in the facility's partial hospitalization program as part of a crisis intervention.
For the study reported here, a total of 370 of the 430 active outpatient clients—201 women and 169 men—who received treatment for at least eight months after the larger study began were rated by their case managers at baseline and at an eight-month follow-up. The primary diagnoses were schizophrenia for 51 percent, major depression for 30 percent, bipolar disorder for 13 percent, anxiety disorder for 5 percent, and adjustment disorder for 1 percent. Forty-seven percent had a diagnosis of a personality disorder, 29 percent had a coexisting substance use disorder, and 86 percent were considered severely mentally ill. Thirty-four percent were uninsured; 41 percent were insured by Medicaid, 23 percent by Medicare, and 2 percent by a health maintenance organization.
Forty-five case managers used clinician rating forms designed for the Client Progress Assessment System to record their perceptions of their clients' clinical and life status and treatment experiences at a baseline assessment and at an eight-month follow-up assessment. Results presented in this report are based on case managers' 4-point ratings of problems in several domains. They included symptoms, which were rated across 17 symptom categories derived from DSM-IV (10), such as appearance and thought content; quality of life, which was rated across 13 life areas, such as finances and living arrangements; treatment readiness, which was rated across nine characteristics describing a person's readiness to benefit from mental health treatments, such as motivation for treatment and medication compliance; and poor warning-sign recognition. The measures of total symptom severity, quality of life, and treatment readiness designed for the Client Progress Assessment System have reliabilities (Cronbach's alpha) of .77, .77, and .86 , respectively.
Case managers also rated the extent to which they were currently treating a client's poor warning-sign recognition. They reported the client's current level of functioning using the Global Assessment of Functioning (GAF) from DSM-IV (10).
Improvement scores for symptoms, quality of life, treatment readiness, warning-sign recognition, and the GAF were computed by summing the residual change scores produced by regression analyses involving the partialing of baseline ratings from the eight-month follow-up ratings. Improvement scores were computed with residual change scores rather than raw difference scores to control for differences in clients' initial levels of problem severity. At the follow-up assessment, case managers also rated whether a client typically exhibited each of 31 warning signs when he or she decompensated in the past (Cronbach's alpha=.82).
A total of 303 of the 370 subjects volunteered to be interviewed during a baseline assessment period, and 202 of these clients were located and agreed to be interviewed a second time at the eight-month follow-up assessment. During the follow-up interviews, clients were asked if they were ever admitted for inpatient psychiatric treatment. If so, they were asked to rate the extent to which they experienced the 31 warning signs before the hospitalization (Cronbach's alpha=.87) and whether they had been aware that they were decompensating at the time.
Client service transaction records for the eight-month period were also collected and coded. These records documented the types and amounts of services each client received and the charges and reimbursed revenues for these services. This report presents total service delivery costs and emergency service use per client for the eight months. Service delivery costs were computed using data presented in the CMHC's 1994 cost report, which is submitted annually to the Oklahoma Department of Mental Health and Substance Abuse Services as a requirement for state funding.
Tests for possible subject selection bias compared demographic and clinical characteristics of the 370 outpatients who remained in treatment for eight months with those of the 60 who did not. Clients who remained in treatment were more likely to be older (a mean of 44 years versus 39 years; t=-3.11, df=428, p<.01) and to have Medicaid or Medicare insurance (71 percent versus 32 percent; χ2=33.46, df=1, p<.001). They were more likely to be diagnosed as having schizophrenia (53 percent versus 33 percent; χ2= 33.46, df=1, p<.001) and to not have a coexisting substance use disorder (32 percent versus 58 percent; χ2= 13.45, df=1, p<.001).
Clinicians' ratings indicated that clients exhibited an average of 11.5 warning signs when they decompensated (range=0 to 27). Only 5 percent were rated as exhibiting no warning signs, whereas 67 percent were identified as exhibiting ten or more. The most common warning signs identified by clinicians were inability to cope with problems and daily activities, noted for 77 percent of the clients; excessive anxieties and worries, for 70 percent; and social withdrawal, for 70 percent. Suicidal ideation was identified as a warning sign for 40 percent, medication noncompliance for 34 percent, and increased drug and alcohol use for 22 percent.
Clinicians' ratings indicated that 45 percent of the clients had problems with poor recognition of warning signs at baseline. By eight-month follow-up, 45 percent demonstrated improved recognition, 41 percent stayed the same, and 14 percent had poorer recognition. According to clinicians, poor warning-sign recognition was either "not addressed at all" in 12 percent of patients with this problem, or "addressed once in a while" in 36 percent of cases. It was a "main focus of treatment" in 12 percent of cases and "addressed regularly" in 40 percent.
Pearson correlations showed that clients who typically exhibited more warning signs when they decompensated were more likely to have been rated as having poor warning-sign recognition at baseline assessment (r=.14, p<.05) and follow-up assessment (r=.29, p<.001). Warning signs most highly correlated with having poor warning-sign recognition were medication noncompliance (r=.41, p<.001), bizarre and inaccurate memories (r=.33, p<.001), failure to keep mental health appointments (r= .32, p<.001), and violent behavior (r=.29, p<.001).
Eighty-four percent of the clients said they had been previously hospitalized, and their average number of hospitalizations was three. Sixteen percent had been hospitalized once before, 29 percent two or three times, 30 percent four to nine times, and 9 percent ten or more times. Of those hospitalized, half (49 percent) reported their most recent hospitalization as occurring within the past two years. Half of those hospitalized (48 percent) also said they had not been aware they were heading toward a hospitalization at the time. Even though many had been unaware they were breaking down, 98 percent of those hospitalized were able to remember and report mental health problems they were experiencing before their most recent hospitalization.
Client reports showed they experienced an average of 15.6 warning signs (range=0 to 27) before their most recent admission. Only 3 percent reported no warning signs, whereas 80 percent experienced ten or more. The most commonly reported warning signs were feeling depressed and apathetic, 79 percent; feeling overwhelmed by problems and being unable to function, 76 percent; behaving like a completely different person, 76 percent; and feeling worried and fearful, 75 percent. In addition, 47 percent reported experiencing suicidal ideation during their most recent decompensation episode, 33 percent reported not taking their medications, and 19 percent reported increased drug and alcohol use.
The average agreement rate across the 31 warning signs between the client's report and his or her case manager's ratings of whether or not the client experienced the signs was 61 percent (range=16 to 87 percent).
In the eight-month period clients with poor warning-sign recognition at baseline were significantly more likely than those with no recognition problems to use both crisis intervention (22 percent versus 12 percent; χ2= 6.37, df=1, p<.01) and inpatient hospital services (16 percent versus 6 percent; χ2=9.10, df=1, p<.001). They also had higher treatment costs (mean cost of $4,629 versus $3,173; t=2.99, df=368, p<.01). Clients who received treatment for poor warning-sign recognition showed more improvement in their recognition than clients whose poor recognition was not treated (mean residual change scores of .09 and -.47, respectively; t=2.10, df=168, p<.05).
Finally, Pearson product-moment correlations showed that improved recognition during the eight months was significantly related to improvement in symptom severity (r=.35, p< .001), quality of life (r=.35, p<.001), treatment readiness (r=.71, p<.001), and daily functioning (r=.29, p< .001). In addition, improved recognition was related to less use of crisis intervention (r=-.19, p<.001) and inpatient services (r=-.24, p<.001). In fact, clients who no longer had problems with poor recognition at follow-up had treatment costs 42 percent lower than clients who continued to have poor recognition (mean cost of $3,047 versus $5,224; t=-4.37, df= 168, p<.001).
The mean eight-month treatment cost for clients who reported they were unaware that they were decompensating before the last hospitalization was 46 percent higher than the cost for clients who reported they had been aware of decompensating (mean cost of $4,908 versus $3,360; t=1.99, df=168, p<.05). Of the 52 percent of clients who had been aware that they were decompensating, 15 percent reported that they had been aware for less than 24 hours before admission, 21 percent for one to six days, and 54 percent for more than a week. Forty-six percent said they felt something could have been done to prevent their most recent hospitalization.
This study found that poor warning-sign recognition was common among people receiving outpatient treatment for severe mental illness. It was associated with poorer clinical outcomes and higher treatment costs and was more likely to improve when treated than when not treated. Consequently, to the extent these observations are generalizable, they suggest that treating poor recognition ought to produce more positive health care outcomes for those with a history of psychiatric inpatient hospitalization or excessive use of crisis intervention services.
However, as the results show, poor warning-sign recognition was treated regularly for only 50 percent of the clients who had this problem. These results suggest that warning-sign recognition may be a promising area in which behavioral health care organizations can both improve the quality of treatment and cut costs. Systematic interventions could involve structured assessments to develop warning-sign profiles for each at-risk client, psychoeducation focused on the self-management of warning-sign behavior, and continuous monitoring of warning signs by case managers and, when possible, by significant others and family members.
In thinking about these results, it is important to consider the selection bias associated with including only those clients who remained in treatment for at least eight months. The correlational nature of the data also preclude making any inferences about the causal role of poor warning-sign recognition in producing poorer clinical outcomes and higher treatment costs.
This research was supported by a grant from the Tulsa Psychiatric Center. The authors also thank David Bahlinger, M.A., and Linda Firth for assistance.
Dr. Novacek and Dr. Raskin are directors of the Tulsa Institute of Behavioral Sciences, 1620 East 12th Street, Tulsa, Oklahoma 74120, and adjunct professors in psychiatry at the University of Oklahoma Medical College in Tulsa. Dr. Novacek's e-mail address is firstname.lastname@example.org. This report is based on a paper presented at the National Conference on State Mental Health Services Research and Program Evaluation held February 11-13, 1996, in Arlington, Virginia.