The present trial evaluated the influence of somatic and psychological factors on health care utilization in the course of Crohn's disease (CD) and the effect of a psychological intervention on high-utilizing CD patients. The somatic and psychosocial course of CD patients submitted to a psychological intervention has already been described [18, 19]. Eighty-seven of 488 consecutive patients in the four participating centers were included in the protocol of the sub study evaluated here. Due to an ambitious study protocol, 75% of the screened patients were excluded . High patient adherence was required in view of the many clinical and psychological examinations as well as the collection of health care data from German insurance companies. These patients were selected for showing high disease activity without requiring a surgical intervention. Moreover, there was a relatively high dropout rate in collecting the bio psychological and health care utilization data. Comparing the epidemiological data of the patients included with those of larger cohorts [34, 35] shows a slightly higher mean age. This can be explained by the fact that our study only included patients over the age of 18. There were more women than men in our study, while the literature generally reports a balanced gender ratio . The disease distribution along the intestinal tract in our study population is comparable to that reported in other studies [36, 37]. The study protocol excluded patients seeking psychological treatment for psychosocial problems as well as those exhibiting a severe somatic course with psychological distress. Psychological data thus showed few psychological symptoms in our study patients. A low baseline level of psychosocial distress in IBD patients was also confirmed by Drossman et al. . This is consistent with data indicating that, despite their symptoms, patients with IBD perceive their health-related quality of life as being quite good .
With the aim to present an appropriate outcome measure, it was very important to describe the somatic course of the disease over a two-year period. We used the ECCDS protocol  to develop a ranking scheme ranging from the best to the worst clinical course. This enabled very careful evaluation of the somatic outcome criterion in 87 patients.
After a six- and twelve-month follow-up, we also assessed the HRQL by a self-rating questionnaire and a well known psychological outcome measurement.
This study was the first to collect CD health care data from German insurance companies (with surprising results regarding the validity of clinical HCU data collection). We are thus able to present a validated third (HCU) outcome variable. We found a high rate of care-seeking behavior in our patient population. It seemed very helpful in this context that patients were also checked very carefully with regard to their somatic and psychological course. However, the longer collecting time and complex analyses of data are the cause of this study being completed much later than our first publication.
The predictor analyses according to the amount of health care utilization in the study population (combined measure of HD and SLD) detected steroid intake, sick leave status, depression, and gender as significant variables of the univariate analysis. This is in agreement with other studies [5, 7, 14, 38]. However, all variables except disease duration failed to reach statistical significance in the multiple regression analysis, and the statistical model with r2 = 0.15 was not significant. The findings in this analysis thus remain weak and unconvincing.
The HRQL in the course of disease could be predicted by psychological, anamnestic and HCU variables (anxiety, depression and coping skills as well as the disease duration and the number of sick leave days in the two years before examination). This is in agreement with other studies [11–15, 39]. Anxiety and coping skills could be included in a highly significant statistical model with r2 = 0.51.
Interestingly, the somatic course of disease could be predicted by variables of acute inflammation (CDAI, steroid intake), psychological indicators (depression), and variables of health care utilization (SLD). Regression analysis identified the CDAI as the only significant variable in a significant model (r2 = 0.28).
By combining the three outcome measurements of health care utilization and the HRLQ as well as the clinical course, we identified depression and steroid intake at randomization as highly significant predictors of the course of illness over two years in a significant statistical model (r2 = 0.22). Disease activity and depression are well-known predictors of the course in Crohn's disease patients [7, 11, 14].
The carefully selected outcome variables enabled us to examine the effectiveness of a psychological treatment in the high-utilizing sub sample. Since there is only a poor response concerning the somatic course of disease to psychological treatment in CD patients it seems important to analyze sub populations. The definition of high utilizers in this study were done by median split of the medical health care data. Whether this is useful in general future studies have to demonstrate. It seems interesting that health care utilization in terms of hospital days and sick leave days was much higher in this CD study group than in other studies [2, 5, 24, 25]. Here we have to take into consideration the health care situation in Germany, in the study cities and their surrounding regions. However, the fact that women had much higher HD and SLD values than men suggests that factors beside the somatic course of illness are important here.
Forty-one high-utilizing patients showed significant differences in the course of HD and SLD (difference between the value one year before randomization and at the one-year follow-up) with an advantage for the intervention group. Thus the present study was able to demonstrate the effectiveness of psychological treatment in reducing the health care utilization of these high-utilizing Crohn's disease patients. This seems interesting, since the randomized study failed to support significant somatic or psychological improvement following the psychological intervention [18, 19] but showed a significant decrease of HD in the treatment group . It is noteworthy that the results of the randomized study also hold true for high-utilizing Crohn's disease patients . We were thus able to show that a psychological treatment was benefitial for this CD subgroup of high and cost-intensive health care utilizers without worsening their adherence to the medical regimen. The psychological focus in treating a CD subgroup of patients to change their psychological status and health care utilization is new and has thus far only been used in this sample of CD patients. Moderating factors seem to be the patients' more effective disease self-management, better adherence and more security in illness behavior during disease crises. Depressive symptoms decreased in CD patients, who had formerly high levels of depression . A longer waiting time before (or neglect of) bowel surgery may be another reason. Greater skill in coping with the illness or with stressful life events may lead to fewer psychic symptoms [18–20] and a better HRQL [1, 14, 15] as well as risk factors like dietary habits or smoking may influence the possibility of going to the hospital or going to work . These aspects could only be partly detected by our psychological measurement (18), but we assume they are meaningful for the change in health care utilization in the psychological treatment group .
Several aspects and limitations should be taken into account when interpreting the present results. The many exclusion criteria in the original trial  led to an examination rate of 87 in this sub study with 488 screened CD patients from specialized GI university medical centers. 87 of these patients included in this health utilization study came from a randomized sample , an additional sample wanted psychotherapy  and a third group had a high rate of missing data  but could be accepted in this analysis. Patients with surgery or those without relapses within the two years before the trial were excluded. Few patients with psychic disturbances like depression, anxiety or a low HRQL were included in the study.
The treatment results relate to a high-level health care population without severe psychiatric co morbidity and were obtained over an observation period of four years. It is possible that the treatment effect on health care utilization would be greater if we examined CD populations with higher psychiatric co-morbidity.
The type and dose of applied psychological treatment in this study should also be taken into consideration: It seems helpful to spend a minimum amount of time communicating with CD patients  and to provide information and offer education programs [21, 22], but an intensive psychological treatment program [24, 25, 34] that focuses on changing coping skills and illness behavior  may intensify common treatment effects in high-utilizing CD patients.