The Value of Treatment-Goal Themes for Treatment Planning and Outcome Evaluation of Psychiatric Inpatients

Background: The taxonomy of treatment-goal themes of the Bern Inventory of Treatment Goals (BIT-T) has shown to be a comprehensive coding system for outpatient use. Aim: The current study examines the value of the BIT-T for treatment planning and outcome evaluation of psychiatric inpatients. Method: The 1991 treatment goals of 675 predominantly non-psychotic psychiatric inpatients were coded using the BIT-T. Ratings of goal-related changes by patients and length of stay were used as outcome criteria. Results: The BIT-T showed to be reliable and exhaustive. Distributions of goal themes are partially associated with diagnoses. Goal-related improvements as well as length of hospital stay differed depending on the themes of patients’ treatment goals. Conclusions: Coding of treatment-goal themes with the BIT-T provides researchers as well as practitioners with valuable information that goes beyond psychopathological diagnoses. This information can be used clinically for treatment planning as well as outcome evaluation.


INTRODUCTION
Patients' treatment goals serve both clinical as well as research purposes. Clinically, treatment-goal definition can be used for treatment planning and outcome evaluation. It was found to have beneficial effects on patient motivation as well as on outcome (Driessen et al., 2001;Tryon & Winograd, 2001). In research as well as in quality assurance, treatment goals can serve as reference points for an individualised outcome assessment. Conventionally some form of the Goal-Attainment-Scaling procedure (GAS) (Kiresuk & Sherman, 1968) is used, and the different goal-attainment ratings are averaged for each patient as a single value.

Development of the BIT-T
By averaging GAS ratings, the researcher misses a possibly important source of variation in the outcome data. Therefore Kazdin called for the development of a typology of treatment goals in order to 'call forth those types of clinical significance that are most relevant' (1999, p. 337). The taxonomy of treatment goal themes of the Bern Inventory of Treatment Goals (BIT-T) was empirically developed and successfully tested with the treatment goals of psychotherapy outpatients. In this setting, the BIT-T proved to have a good exhaustivity and inter-rater reliability, identified differences in the treatment goals of anxiety and depression patients and showed meaningful relations to standardised intake measures (Grosse Holtforth & Grawe, 2002). Table 1 shows the categories .3) on the first two of three levels of abstraction.
For inpatient psychiatry, Driessen and colleagues (2001) diagnosed a lack of studies in psychiatry about methods as well as instruments for the assessment of patients' treatment goals. With the BIT-T now, the assessment of patients' treatment goals is possible also in these patients. Table 1 The taxonomy of treatment goal themes of the Bern Inventory of Treatment Goals v. 3.

3)
Coping with specific problems and symptoms (P

Studies on treatment-goal themes
Several authors have studied the themes of inpatients' treatment goals (Dimsdale et al., 1979;Go¨llner, 1983;Ruff & Werner, 1987;Riehl-Emde & Vogler, 1990;Harris et al., 1993;Finke et al., 1995;Heuft et al., 1998;Sack et al., 1999;Berking et al., 2001;Driessen et al., 2001). Based on these studies, two distinctions seem necessary. First, treatment goals formulated by patients themselves can differ considerably from those defined by their therapists (Dimsdale et al., 1979). Second, different methods of goal definition can be used, including the use of goal checklists/questionnaires (Dimsdale et al., 1979;Finke et al., 1995;Driessen et al., 2001), free formulation of treatment goals (Ruff & Werner, 1987;Riehl-Emde & Vogler, 1990;Heuft et al., 1998;Sack et al., 1999;Berking et al., 2001) or a combination of these (Grosse Holtforth, 2001). Therefore when different studies are being compared, different data sources and methods should be taken into account. In addition, some studies use taxonomies of treatment-goal themes of a rather limited range (Ruff & Werner, 1987;Riehl-Emde & Vogler, 1990;Sack et al., 1999), others are empirically constructed and more elaborated (Heuft et al., 1998;Berking et al., 2001). Most relevant for this study are the results by Berking and colleagues (2001) who used the BIT-T (v. 3.0) to categorise the single most important treatment goals of 1695 inpatients in psychosomatic rehabilitation with the diagnoses of predominantly depressive, anxiety or somatoform disorders. While patients receiving treatment for somatoform pain disorders concentrated heavily on pain reduction, and anxiety patients also strongly on anxiety reduction, the treatment goals of depressive patients were much more thematically diverse, supporting the outpatient results by Grosse Holtforth and Grawe (2002). Additionally, treatment goals concerning well-being/functioning had the best goal-attainment ratings, while treatment goals concerning symptom relief turned out the worst.

Aim
To our knowledge, there have been no studies conducted on the themes of the treatment goals of psychiatric inpatients being formulated by patients themselves or in conjunction with their therapists. The current study examines the value of the BIT-T for treatment planning and outcome evaluation in this patient population. Quality criteria are: (1) a good inter-rater reliability as well as an infrequent use of thematic residual categories (applicability); (2) provision of additional information beyond diagnoses; and (3) prediction of outcome.

Subjects
Treatment goals of 675 patients of a private psychiatric hospital in Switzerland being consecutively admitted in 25 months between 1997 and 1999 were analysed. The patients had a mean age of 48.3 years (SD ¼ 14:4, 17-88), 65.0% were female, and 50.5% were married. The highest level of education was high school for 55.4%, 1.2% had a university degree and 11.3% had no completed education; 39.6% were employed, 10.2% self-employed, 7.0% unemployed, 2.7% on welfare, 9.2% were retired, 22.8% worked in the household, 8.6% were in training or had another employment status. Mean length of stay was 50.5 days (SD ¼ 31:9, 1-187). Patients were grouped in six diagnostic categories according to their principal diagnoses (ICD-10; Dilling et al., 1991) (Table 2).

Material
After being admitted, patients were asked to list on a worksheet up to three areas of change as their individual treatment goals. Afterwards, these were discussed with the principal therapist and eventually modified. As an additional restriction, therapists urged substance use/ addiction patients to include at least one corresponding treatment goal into their goal list. One thousand nine hundred and ninety-one treatment goals on 786 worksheets entered the analysis. The figure is higher than the patient number since some patients were admitted more than once. The chief of staff of a ward (MD) did the ICD-10-categorization together with the primary therapist, who could be a psychiatrist or a psychologist. At the end of the hospital stay, the patients rated the individual changes concerning the problems being defined on the treatment goal worksheet on six-point Likert scales between 'very much worse' and 'very much better'.

Procedure
Data were taken from archived patient files and personal information was recoded by hospital staff to disguise patients' identities. The two raters were Masters students who had participated in the whole process of taxonomy revision after the construction of BIT-T 1.0 (Grosse Holtforth & Grawe, 2002). Therefore, they were considered well trained. First, the raters independently assigned each treatment goal to a single category. Then, inter-rater reliabilities were computed, disagreements were identified, discussed with the principal author and a consensus rating was performed.

RESULTS
On average, patients defined 2.53 (SD ¼ :65) treatment goals. The two raters reached an inter-rater reliability of Cohen's ¼ :73 at the goal type level (82% agreement), ¼ :73 for goal categories (73%) and ¼ :62 for subcategories (63%). Due to small numbers in Table 2 Grouping of patients according to principal diagnoses (ICD-10) the subcategories, only treatment-goal types and categories will be examined in the following analyses. Figure 1 shows the percentage of psychiatric inpatients with specific treatment-goal categories and treatment-goal types of the BIT-T (v. 3.3). The majority of patients defined treatment goals concerning symptom relief (P; 77.3%), being followed by interpersonal goals (I; 36%), goals concerning personal growth (G; 26.1%), goals concerning wellness (W; 18.8%) and goals concerning existential issues (E; 6.2%). Of all treatment goals (N ¼ 1991), 15.2% were categorised in the residual goal type; 8.6% belonged to the residual category of non-categorisable goals, 0.9% to the residual category of regeneration, 2.6% psychosocial rehabilitation, 2.1% were somatic goals and 1.1% was not considered to be treatment goals. Additional themes in the residual categories occurred rarely: P 2.3%, I 0.1%, W 0.1%, E 0% and G 0.1%. Many of the additional goals in the P goal type concerned changes of psychotic experiences. The distribution of patients defining goals in the other thematic categories can be seen in Figure 1. A large proportion of patients defined symptom relief as the most important goal type (58.4%), followed by interpersonal goals (10.2%), wellness and personal growth (9.3% each) and existential issues (1.5%); 11.3% of most important goals belonged to the residual category.
To examine the association between treatment-goal themes and diagnoses, for each treatment-goal type and category, separate tables of frequencies were constructed containing the number of patients in the diagnostic groups having a treatment goal in the goal type/category or not. Chi 2 -statistics (d:f: ¼ 5) were computed to test the strength of the association. Because the distribution of treatment-goal themes is biased for the alcohol group in favour of P-goals concerning alcohol, the analyses were performed for the sample with or without the alcohol patients. Taking a significance level of p < :05 as a criterion, results differed in strength, but not statistical significance. Thus only the results for all diagnostic groups are reported. Table 3 (p. 86) shows the distribution of patients within the six diagnostic groups defining a treatment goal within the statistically differing goals types and categories of the BIT-T.
Treatment outcome was assessed by improvement ratings by patients and by the number of days spent in the hospital. One thousand five hundred and seventy-four change ratings were available. The distribution of change ratings was: 18.2% 'very much better', 23.7% 'much better', 40.6% 'better', 11.6% 'unchanged', 0.8% 'worse', 0.3% 'much worse', 0.3% 'very much worse', 4.6% 'cannot say'. One-way analyses of variance tested the differences in mean improvement ratings for the different goal types and categories. Treatment goals being coded in the residual categories were excluded from the analysis, leaving 1288 treatment goals on the goal-type level, and 1207 goals at the category level, for which also an improvement rating was available.
Mean improvement ratings differed significantly among goal types (F ¼ 5:00, d:f: ¼ 4, p < :001), and goal categories (F ¼ 4:43, d:f: ¼ 24, p < :001). Figure 2 (p. 87) shows the mean improvement ratings of goal types. In post-hoc Scheffe´tests, interpersonal goals were significantly less improved than goals concerning symptoms ( p < :002), or wellness ( p < :025). Other differences were not statistically significant. Goal categories that were associated with the highest improvement ratings were substance use/addiction and eating behaviours. The categories of obsessive-compulsive disorder (OCD) symptoms and current relationship showed the lowest improvement ratings. For reasons of space, post-hoc tests of differences among goal categories are not reported here and in the following analysis.
One-way analyses of variance tested the differences in length of stay for the different goal types and categories. Treatment goals being coded in the residual categories were excluded from the analysis. To avoid effects of outliers, the sample was limited to patients staying in the hospital between one and 100 days, leaving 1108 treatment goals of 551 patients. Treatment-goal types (d:f: ¼ 4, F ¼ 4:14, p < :002) as well as categories (d:f: ¼ 24, F ¼ 3:04, p < :001) differed significantly in the associated durations of stay. Figure 3 (p. 87) shows the mean durations of stay of patients with specific goal types. In post-hoc Scheffe´tests, the goal type concerning personal growth differed significantly from the symptom-related Table 3 Percentage of patients in the diagnostic groups defining a treatment goal in specific categories of the BIT-T Note. Only treatment-goal types/categories are displayed that show significant differences in frequencies * p < :05, ** p < :01, *** p < :001 ( p < :007) as well as the wellness ( p < :027) goal types. Goal categories that were associated with the longest duration of stay were OCD symptoms and assertiveness; those with the shortest durations of stay were substance abuse/addiction and other (interpersonal) relationships.
To test whether goal-related improvements are associated with a longer length of stay, analyses of covariance were performed using goal types or goal categories as independent variable, improvement ratings as the dependent variable, and length of stay as a covariate. Length of stay did not correlate significantly with change ratings and clearly missed statistical significance as a covariate for goal types (d:f: ¼ 1, F ¼ :92, p > :338) as well as categories (d:f: ¼ 1, F ¼ 1:23, p < :268).

DISCUSSION
Our results show that the BIT-T can be profitably used for categorising treatment goals in an inpatient psychiatric setting: raters are able to categorise treatment goal themes with satisfying inter-rater reliability, and thematic residual categories are used rather infrequently. The definition of certain treatment goal themes is only partially determined by patients' diagnoses. Treatment-goal themes are associated with goal-related change ratings, and to the length of hospital stay.

Applicability and distribution of goals
Our results support the exhaustivity and applicability of the BIT-T in inpatient settings. The most frequent additional themes in the current inpatient psychiatric sample concerned changes of psychotic experiences. For research in this setting, an additional category of psychotic experiences should be included in the BIT-T. Psychiatric inpatients in our sample more strongly focus on symptom relief and improvement of well-being and less on interpersonal goals and goals concerning personal growth than outpatients do (Grosse Holtforth & Grawe, 2002). Berking and colleagues (2001) reported similar findings for psychosomatic inpatients. Thus, a stronger focus on symptom relief seems to be a characteristic of inpatients' treatment goals. It is up to future studies to clarify the generalisability of these results for other inpatient settings.

Treatment goal themes and diagnoses
Our findings confirm and extend outpatient findings concerning only partial determination of treatment goal themes by diagnoses. In this psychiatric inpatient sample, the six diagnostic groups (alcohol, psychotic, depressive, anxiety, adjustment and bipolar patients) differ in the frequencies with which treatment goals of specific categories are defined. However, diagnoses only partially determine the treatment goals. Thus, the kind of psychopathological symptoms experienced by a patient only increases the likelihood that a patient defines certain treatment goals; with this influence seemingly being strongest in anxiety disorders. The results concerning substance abuse/addiction in our sample should be interpreted with caution, since they are biased by hospital regulations. Other factors (personal goals, learning history, cultural biases, etc.) obviously also exert a strong influence on the kind of treatment goals a patient defines. Future studies should examine patients' intrapsychic processes in forming certain goals as well as the interpersonal processes of reaching goal agreement.

Treatment-goal themes and treatment outcome
Problem-change ratings show substantial improvements for the vast majority of patients. However, the less than perfect ratings leave room for further improvements after termination of inpatient treatment. Problems related to interpersonal goals show less improvement than goals related to symptoms, and wellness. In contrast, Berking and colleagues (2001) found problem/symptom-related goals to be rated the worst in their sample of psychosomatic inpatients. As their results on the category and subcategory levels show, this might be attributable to the high number of patients with pain-and sleep-related problems, who rated goal attainment the worst. Our exploratory results also point to differences in improvements between the categories of the same goal-type. Concerning interpersonal goals, for most, a hospital stay seems to make only a start. Given these results, former inpatients should also show a heightened prevalence of interpersonal goals when entering outpatient psychotherapy, as found for unselected outpatients (Grosse Holtforth & Grawe, 2002). Patients with treatment goals concerning personal growth stayed the longest, but the related problems do not receive the lowest mean change ratings. Future studies should try to replicate the finding of a longer duration of stay for patients with treatment goals concerning personal growth and investigate the reasons for this association. To allow for more detailed analyses and to examine the generality of our results, future studies with psychiatric patients should study problem-change or goal-attainment ratings at the more detailed levels of treatment-goal categories and subcategories, calling for an even higher number of inpatients being studied.
In our sample, the absolute differences in improvements between single BIT-T goal types or categories and the remaining types/categories are rather small. Thus, the kinds of problems for which inpatients define treatment goals predict problem change to a certain, but small, degree. This is not surprising, given the total ignorance of the psycho-and pharmacotherapeutic treatments received. Our results should encourage researchers to integrate information on patients' treatment goals into the outcome measures of clinical studies. Thus, future studies might lead to a more adequate differential prediction of treatment response that goes beyond diagnostic and sociodemographic information.

Clinical implications
As Tryon and Winograd (2001) have shown in their review of the literature, goal agreement between patient and therapist is an essential part of the working alliance and is associated with better outcomes in psychotherapy. Although psychiatric inpatients emphasise symptomatic goals more strongly than outpatients do, the full range of treatment goals of the BIT-T was encountered. This result thus suggests that future therapists should receive a broad training beyond solely providing symptomatic relief in order best to meet their patients' concerns. Additionally, the finding of only partial determination of treatment goals by diagnoses indicates against standardised treatments for patients with the same diagnoses. Obviously, patients having the same diagnoses can have very different treatment goals, asking for individualised therapy contracts as well as treatment plans. Also, information on average success ratings for different treatment-goal themes can help therapists to inform their patients better about what they can expect from their treatment. To justify such individual predictions better, more studies are needed examining the relationship between treatment-goal themes and goal-attainment in different settings, by different interventions and in different time ranges.

Limitations
The current study has several limitations. First, it covers only treatment goals of patients in one hospital. Further studies in different psychiatric hospitals in different countries are needed to be able to generalise the results. Second, goal definition was biased for addiction/substance abuse patients in this particular clinic. Thus, examination of unbiased treatment goals of patients in this diagnostic group is needed. Third, no reliable information was available about the kind of psychotherapeutic treatment patients received. Future studies should examine the interactions of treatment-goal themes and classes of therapeutic interventions. The best way to examine these differential treatment effects on goal attainment would be to use the BIT-T to classify the treatment goals of patients in randomised-controlled trials. If treatment-goal data are available, this can be done retrospectively.

Future development of the BIT-T
The BIT-T has been developed further into two directions. First, a standardised checklist of treatment goals (BIT-C) was developed by transforming the subcategories of the BIT-T into questionnaire items. In a first application, the BIT-C provided efficient assessment of the treatment goals of inpatient psychotherapy patients (Grosse Holtforth, 2001). Second, Hasler and colleagues (2002) devised a direct change measure using the BIT-T categories and found different amounts of change for different thematic domains.

Final remark
In 1977, Minsel described the topic of treatment goals as a 'dark chapter' in psychotherapy research. With the BIT-T, now researchers as well as practitioners have a practical and empirically founded 'torchlight' at hand to enlighten further the field of treatment goals in various mental-health settings.