A Pragmatic Randomised Controlled Trial of Dialectical Behaviour Therapy: Effects on Hospitalisation and Post-Treatment Follow-Up

Abstract

p = 0.54). However, removing the hospitalisation days for the pa-tient who began treatment as a long-term inpatient, the difference was statistically significant (incidence rate ratio 9.08, 95% CI 1.01–81.6, p = 0.05). Six-month post-intervention follow-up data was obtained for 14 of the 19 patients allocated to DBT who completed the full 12 months of treatment. A standardised self-harm interview was used to assess self-harm frequency during the follow-up period. The mean number of days with self-harm in the last 2 months of treat-ment for DBT completers was 1.79 (SD 3.68) whilst the mean num-ber of days with self-harm during the 6 months after treatment was 1 (SD 1.80), i.e. a rate of 0.33 days per 2-month period. A Wilcox-on signed-rank test showed that this was not a significant differ-ence (z = 1.42, p = 0.16). No DBT completers had any inpatient hospitalisations during the 6-month follow-up period. For treat-ment dropouts, the rate of follow-up was too low (8 of 21 partici-pants) to render statistical comparison valid. Our findings on hospitalisation concur with international RCTs that have shown DBT can reduce hospitalisation [3, 4] , but are in contrast with another UK RCT which found hospitalisation days did not differ between DBT and TAU [5]. Treatments which reduce the use of inpatient resources are particularly important, given that patients with BPD have been found in several studies to make greater use of inpatient psychiatric services than patients with other personality disorders [6] and than patients with major depressive disorder [7]. The high healthcare costs (and presumably patient distress) resulting from such frequent hospitalisation ren-der the implementation of interventions that can reduce hospi-talisation an important priority for this patient group. In 2012 in this journal we published the results of a pragmatic randomised controlled trial (RCT) of dialectical behaviour therapy (DBT) versus treatment as usual (TAU) in the United Kingdom National Health Service for patients with borderline personality disorder (BPD) and frequent self-harm [1] . This was a sample of 80 patients, 40 allocated to DBT and 40 to TAU. In this publication we reported on the primary outcome, self-harm, and showed that patients in the DBT condition achieved a significantly greater re-duction in self-harm frequency over time than patients in the TAU condition. We report here the effect of DBT compared to TAU on inpatient service use, and a follow-up 6 months after the end of treatment. The sample and treatment characteristics are reported in full in the original RCT publication [1] .Data on psychiatric hospitalisation were collected by interview- ing patients at two monthly intervals using the Client Service Re-ceipt Inventory [2] , which was then triangulated with data from electronic patient records. Figure 1 shows the percentage of pa-tients admitted to hospital in each condition in the year before and the year during treatment. In the year prior to treatment, 24 pa-tients had been hospitalised with the number of inpatient days ranging from 0 to 365 (mean 20.5, SD 63.1). The number of inpa-tient days in the year prior to treatment did not differ between conditions. During the 12-month intervention period, 2 patients allocated to DBT and 11 allocated to TAU were hospitalised. For the 2 patients hospitalised in the DBT condition, 1 was hospitalised following dropping out of DBT, whilst the other was a long-term inpatient when beginning DBT, and remained so for the first 3 months of treatment. A logistic regression showed that the odds of hospitalisation during the intervention period were significantly higher in patients allocated to the TAU condition (odds ratio 4.68, 95% CI 1.20–18.3, p = 0.03). This difference remained significant after adjusting for whether patients had been hospitalised in the year prior to treatment (adjusted odds ratio 10.77, 95% CI 1.96–59.2, p < 0.01). The total number of inpatient days per person over the year was lower in the DBT condition (mean 4.0 days, SD 20.0) than in the TAU condition (mean 8.4 days, SD 17.6), but a negative binomial regression showed that this was not a statistically sig-nificant difference (incidence rate ratio 2.06, 95% CI 0.21–20.5,