Best Practices: Relationship Management Therapy for Patients With Borderline Personality Disorder

The institutional review board at the University of Western Ontario approved the retrospective study presented here. Data were examined for patients who were enrolled in the relationship management therapy program from its inception in September 1998 to December 2000. The relationship management therapy team accepted patients aged 18 to 64 years. As part of the intake screening, the team met with both the referred individual and the referring clinicians or agency to establish that the diagnosis was borderline personality disorder, and all patients in relationship management therapy met DSM-IV criteria.

Patient chart review yielded data on demographic characteristics; incidents of self-harm; hours of seclusion, restraint, and constant nursing observation; number of hospital admissions; and inpatient length of stay. Each participant had at most six years of data (five years before enrollment in relationship management therapy and one year of receiving relationship management therapy). All participants had at least 12 months of data that were collected before they entered the relationship management therapy program.

A retrospective pre-post analysis was selected because the team had initially been established on the basis of clinical demands and an evaluation of the relationship management therapy team had not been planned. Because patient data were charted over different lengths of time, annual mean pre- and post-program rates were constructed and compared. Because of the team's focus on reducing chaos and conflict through the promotion of patient choice, the primary outcomes were incidents of self-harm and number of hours of seclusion, restraint, and constant nursing observation. Additional outcomes included number of hospital admissions and inpatient length of stay (days).

Data were analyzed by using paired statistical tests (both paired t tests and Wilcoxon's signed-rank tests).

Forty-one patients were consecutively referred to the relationship management therapy program from September 1998 to December 2000. Fourteen were not admitted to the program: eight did not have borderline personality disorder, four refused service, and two chose a different caregiver. Of the 27 patients who were admitted to the program, 25 (93 percent) were female, and 14 (52 percent) were single. The mean±SD age of the participants was 38±9 years. Fifteen (56 percent) had completed or had some secondary education, and seven (26 percent) had completed or had some university education.

A high level of comorbid conditions was present. Seven patients (26 percent) met DSM-IV criteria for an additional personality disorder (the most prevalent was an antisocial personality disorder), and 16 (59 percent) met DSM-IV criteria for an additional axis I disorder (the most prevalent was a substance-related disorder).

All 27 patients had data available for at least one year before they enrolled in relationship management therapy. Two participants (7 percent) had only three months of data after enrollment. Fifteen (56 percent) had data available for one year after enrollment. During the period before enrollment in the relationship management therapy program, data were available for a mean of 4.25 years. During the period of enrollment, data were available for a mean of .83 years.

Differences in pre-post outcomes were tested statistically by using paired t tests and Wilcoxon's signed-rank tests. No differences were seen in the mean number of hours in seclusion (from 12±34 hours in the year before enrollment to 9±36 hours in the year during enrollment). Sizeable and statistically significant differences were seen in the mean annual number of hours in restraints (from 33±135 hours to 1±3 hours; p<.01 for Wilcoxon signed-rank test) and constant nursing observation (from 95±342 hours to 4±14 hours; p<.001 for Wilcoxon's signed-rank tests). Even though relationship management therapy prohibits seclusion and restraint, our study found that some incidents still occurred. Although the relationship management therapy team was available during regular working hours, duty physicians provided night and weekend coverage and sometimes ordered seclusion or restraint. Overall, self-harm incidents declined significantly, from 2±3 per year before enrollment in the relationship management therapy program to none during the time that patients were enrolled in this program (p<.001 for Wilcoxon's signed-rank tests). Hospital admissions rose from a patient average of 1±1 per year before admission to the program to an average of 2±4 per year while in the program. However, this increase was not statistically significant. At the same time, the average number of inpatient days per year declined by 60 days (from 101±74 days to 41±53 days per year; p<.001 for Wilcoxon's signed-rank tests).

The use of seclusion and restraint was significantly reduced after the introduction of relationship management therapy. Some may view these variables as inappropriate outcomes because seclusion and restraint are prohibited under relationship management therapy's protocol and thus would be expected to decrease while patients received this type of therapy. However, this study reports the findings of the first implementation of relationship management therapy in a psychiatric hospital setting, and it is important to know whether this type of therapy works in real-world settings. This study indicates that it is possible to essentially stop using seclusion and restraint to "treat" patients with borderline personality disorder. This more permissive approach was not associated with an increase in deliberate self-harm. Indeed, the opposite was true; over the study period, no patients had any serious episodes of self-harm or completed suicide.

Although not the primary goal of relationship management therapy, the decreased hospitalization was associated with significant fiscal savings. By using a Canadian per diem rate of $462.50, the annual hospital savings during the relationship management therapy period would be nearly $750,000 ($462.50 per hospital day × 60 fewer hospital days per patient × 27 patients) through the reduction in inpatient care. By using a .80 conversion rate, this savings is about U.S. $600,000. The reduction of 1,620 bed-days (60 fewer hospital days per patient × 27 patients) may have allowed other patients to benefit from inpatient treatment.

It is possible that changes in the availability of beds during the study period could have reduced admission opportunities. However, no patient who was enrolled in the relationship management therapy program was refused admission because of lack of a bed. Although relationship management therapy was associated with an average reduction in inpatient length of stay of 59 percent (101 to 41 days), it is possible that the decline is associated with a temporal trend in bed reduction. However, the total reduction in beds at London (Ontario) Psychiatric Hospital over the study period was small; there were 367 beds at the beginning of 1993 and 332 at the end of 2000 (a 10 percent reduction compared with the study's observed 59 percent). Because of the shortage of psychiatric beds (7), this finding has important policy implications.

It is possible that the patients who were discharged from the hospital for 24 hours or who used the opportunity to discharge themselves from the hospital remained disturbed in the community. Thus the financial savings from reduced hospitalization and constant nursing observation may have been offset by intervention by police or community mental health agencies. Although no malpractice suits resulted from our study, the threat of a lawsuit related to discharging a disturbed patient may be a costly barrier to the adoption of relationship management therapy.

Our findings must be viewed as a naturalistic pilot study. The sample was small, and the study was not a randomized controlled trial. Furthermore, a number of patients who were referred to relationship management therapy were high users of hospitalization, so some improvement may be because of regression toward the mean. Finally, the United States does not have universal health care like Canada.

Although it appears that relationship management therapy has the potential to reduce restrictive practices, self-harm incidents, and inpatient length of stay, further study is needed to refine patient outcome measures in both the short and the long term. A societal perspective involving costs and benefits to the patients, hospital staff, and community would also greatly enhance future assessments of relationship management therapy. Because this type of therapy has never been previously formally implemented in a hospital setting (perhaps because of fear of lawsuits), our results clearly fill a gap in the literature about a treatment model that one day could be considered a best practice.

Dr. Hoch received financial support from a Career Scientist Award from the Ontario Ministry of Health and Long Term Care.