Cost-utility of a specific collaborative group intervention for patients with functional somatic syndromes

Journal of Psychosomatic Research 90 (2016) 43–50

Contents lists available at ScienceDirect

Journal of Psychosomatic Research

Cost-utility of a specific collaborative group intervention for patients with functional somatic syndromes Alexander Konnopka Dr. a,⁎, Hans-Helmut König Prof. a, Claudia Kaufmann Dr. b, Nina Egger Dr. a, Beate Wild Prof. b, Joachim Szecsenyi Prof. c, Wolfgang Herzog Prof. b, Dieter Schellberg Dipl.-Psych. b, Rainer Schaefert PD b a b c

Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center Hamburg-Eppendorf, Germany Department of General Internal Medicine and Psychosomatics, University Hospital Heidelberg, Germany Department of General Practice and Health Services Research, University Hospital Heidelberg, Germany

a r t i c l e

i n f o

Article history: Received 29 February 2016 Received in revised form 29 August 2016 Accepted 1 September 2016 Keywords: Functional somatic syndromes Collaborative care Cost-effectiveness

a b s t r a c t Background: Collaborative group intervention (CGI) in patients with functional somatic syndromes (FSS) has been shown to improve mental quality of life. Objective: To analyse incremental cost-utility of CGI compared to enhanced medical care in patients with FSS. Methods: An economic evaluation alongside a cluster-randomised controlled trial was performed. 35 general practitioners (GPs) recruited 300 FSS patients. Patients in the CGI arm were offered 10 group sessions within 3 months and 2 booster sessions 6 and 12 months after baseline. Costs were assessed via questionnaire. Quality adjusted life years (QALYs) were calculated using the SF-6D index, derived from the 36-item short-form health survey (SF-36). We calculated patients' net-monetary-benefit (NMB), estimated the treatment effect via regression, and generated cost-effectiveness acceptability curves. Results: Using intention-to-treat analysis, total costs during the 12-month study period were 5777EUR in the intervention, and 6858EUR in the control group. Controlling for possible confounders, we found a small, but significant positive intervention effect on QALYs (+0.017; p = 0.019) and an insignificant cost saving resulting from a cost-increase in the control group (−10.5%; p = 0.278). NMB regression showed that the probability of CGI to be cost-effective was 69% for a willingness to pay (WTP) of 0EUR/QALY, increased to 92% for a WTP of 50,000EUR/ QALY and reached the level of 95% at a WTP of 70,375EUR/QALY. Subgroup analyses yielded that CGI was only cost-effective in severe somatic symptom severity (PHQ-15 ≥ 15). Conclusion: CGI has a high probability to be a cost-effective treatment for FSS, in particular for patients with severe somatic symptom severity. © 2016 Elsevier Inc. All rights reserved.

1. Introduction Functional somatic syndromes (FSS) are characterised by persistent bodily complaints for which even intensive diagnostic efforts do not reveal sufficient explanatory peripheral organ pathology [1]. Patients with severe FSS have disproportionate impairment of daily functioning and quality of life, are very prevalent in primary and hospital care, and are difficult to treat [2]. They show high health care utilisation, substantial ⁎ Corresponding author at: Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center HamburgEppendorf, Martinistr. 52, 20246 Hamburg, Germany. E-mail addresses: [email protected] (A. Konnopka), [email protected] (H.-H. König), [email protected] (C. Kaufmann), [email protected] (N. Egger), [email protected] (B. Wild), [email protected] (J. Szecsenyi), [email protected] (W. Herzog), [email protected] (D. Schellberg), [email protected] (R. Schaefert).

http://dx.doi.org/10.1016/j.jpsychores.2016.09.001 0022-3999/© 2016 Elsevier Inc. All rights reserved.

losses of productivity, and high societal costs [3]. Guidelines recommend a stepwise therapy for FSS, comprising treatment by the general practitioner (GP) in mild cases, and additional psychotherapy in more severe cases [4]. Although several cost-analyses have been conducted for FSS [3], only few cost-effectiveness analysis concerning cognitive behavioural therapies have been conducted in this spectrum of disorders so far [3,5–12]. In a recent cluster randomised controlled trial (RCT) (speciAL) a combination of GP training in diagnosis and management of FSS plus a specific collaborative group intervention (CGI) was compared to enhanced medical care (EMC) based on the same GP training in diagnosis and management of FSS [13]. In terms of the primary outcome, the physical composite score of the SF-36, intention to treat analyses (ITT) showed no difference between the study arms at 12 months. In terms of the secondary outcome, the mental composite score of the SF-36, CGI was associated with a significant and clinically meaningful improvement in ITT after 12 months (effect size 0.31). In the per protocol

44

A. Konnopka et al. / Journal of Psychosomatic Research 90 (2016) 43–50

analysis, CGI was further associated with a significant reduction in health anxiety and psychosocial distress. Alongside this clinical trial data on health care utilisation and loss of productivity were collected, which we used to analyse the incremental cost-utility of CGI compared to EMC from a societal perspective.

2. Methods 2.1. Study population This study was conducted alongside a clustered RCT (ISRCTN 55280791) of a specific collaborative group intervention for FSS patients in generAL practice (speciAL). The study was approved by the ethics committee of the University of Heidelberg. Design, treatment manual and results of the trial have been reported elsewhere [13,14]. Patients were recruited in two steps (Fig. 1): First, 1112 GPs located within a 45-minute drive from Heidelberg were invited to participate. Of these, 39 GPs were finally included and randomised by a statistician under independent management. After randomisation 4 GPs left the study due to workload. Second, the 35 GPs contacted 569 patients that seemed eligible for study participation between November 2007 and December 2009. Patients were eligible, if they fulfilled the following criteria: (1) presence of persistent (≥6 months) bodily complaints without explanatory organ pathology following; (2) presence of at least one of the two conditions: (I) at least mild somatic symptom severity (SSS) on the Patient Health Questionnaire (PHQ-15) represented by a cut-off

of ≥5 and/or (II) relevant health anxiety on the Whiteley-7 (WI-7) represented by a sum score of ≥4 on a WI-7-scale dichotomized (0 = not at all/a little, 1 = moderately/quite a bit/extremely) according to Christensen et al. [15]; (3) age between 18 and 70 years; (4) speaking German sufficiently; (5) informed consent. By this procedure the patient groups were formed by the GPs according to respective instructions in the GP training and in close consultation with the collaborating psychosomatic specialist when necessary. The number of patients per group was 6 to 12. All data were assessed at baseline (T0), after 6 months (T1; post-treatment) and after 12 months (T2; follow up).

2.2. Interventions 2.2.1. GP training All GPs received a guideline-based [16] training in diagnosis and management of FSS [17]. The training encompassed 15.5 h in total and emphasised an interpersonal perspective on somatisation in order to build a sustainable working alliance through an accepting and patientcentered communication. In detail, the training addressed the following skills: (1) how to ask for a patient's subjective illness theory; (2) how to help the patients in developing a bio-psycho-social illness model; (3) use of specific treatment tools, like symptom diaries; (4) how to support the patient in an ongoing use of active coping skills. The GPs were recommended to establish a structured treatment setting with regularly scheduled and time-limited consultations.

Fig. 1. Consort chart of patients in the speciAL-trial.

A. Konnopka et al. / Journal of Psychosomatic Research 90 (2016) 43–50

2.2.2. GP training plus collaborative group intervention (CGI) Additionally to the GP training described above, GPs in the intervention group received a training of 12 h in total in conducting a manualised disorder-specific collaborative group intervention. Our treatment manual as well as the guideline follow a comprehensive approach focusing on what the single, broadly overlapping functional somatic syndromes have in common [18]. After this training, each GP together with a psychosomatic specialist (physician or psychologist with a formal specialization to treat patients with psychosomatic disorders) offered the group intervention in his own practice. All patients randomised to the intervention group were offered 10 group sessions lasting 90 min per session over a period of 3 months. The patients were also offered 2 booster sessions 3 and 9 months after the group phase. During the 10 sessions phase, GPs and psychosomatic specialists met 3 times for supervision (4.5 h in total), which was conducted by CK and RS. The group intervention applied was based on an interpersonal approach of a psychodynamic-based therapy extended by elements of cognitive-behavioural therapy (CBT) [14]. In each session, more than half of the time was used for process-oriented, theme-centered and interactive group work, and the rest of the time was used for structured but still interactive psychoeducation and progressive muscle relaxation training. 2.3. Instruments 2.3.1. Socio-demographic and illness-related variables Patients were asked about gender, age, education and employment status, marital status, the number of children, duration of somatic symptoms, and psychotherapeutic pre-treatments. As a measure of physical comorbidity the Charlson Comorbidity Index (CCI) was calculated [19]. The main clinical outcome was health related quality of life, assessed by the 36-item short-form health survey (SF-36) [20]. Based on the SF-36, the SF-6D utility index was estimated and used to calculate quality adjusted life years (QALYs) [21]. SSS was quantified using the German version of the PHQ-15 [22]. 2.3.2. Resource utilisation and costs We measured resource utilisation and loss of productivity using the German version of the “Client Sociodemographic and Service Receipt Inventory (CSSRI)” adapted to the purposes of our study [23]. All variables were assessed for retrospective periods of six months at T0, T1 and T2. We measured utilisation of hospital treatments and rehabilitation, outpatient physician and non-physician services, home-visits, phone contacts, pharmaceuticals, and transports to medical treatment. We also assessed sickness absence days and time spent for treatments (absenteeism) as well as productivity reduction at work (presenteeism) [24]. Inpatient treatments were monetarily valued by costs per diem, outpatient physician and non-physician visits by costs per contact [25]. For home visits and phone calls we used prices from the German medical fee schedule [26]. Pharmaceuticals were valued using prices from the German catalogue of drugs (“Red list”) [27]. Costs for transportation were asked from the patients, except for private car use which was valued with 0.30EUR per kilometre according to the German income tax act. To value lost productivity, we used average gross earnings, adjusted by additional payments of employers (e.g. for social insurances) [28]. Costs were calculated in EUR, from a societal perspective and refer to the year 2007. Unit costs from years before 2007 were inflated to 2007 values using the German consumer price index for direct costs and the German earning price index for indirect costs. Intervention costs were divided into fix and variable costs. Fix costs per patient covered GP trainings, supervision of GPs and psychosomatic specialists, and were 4.17EUR in the control group and 15.92EUR in the intervention group (for details please see Appendix Table 1). Variable

45

costs covered the therapeutic sessions in the intervention group. According to the reimbursement codes for outpatient care the sessions were valued with 36.72EUR per patient per session (reimbursement code 22,222 “psychosomatic therapy; group modus”) [29]. 2.4. Analyses We used the multiple imputations by chained equations method to impute missing values using sociodemographic variables, morbidityrelated variables and available utilisation data as predictors [30]. Health effects and costs were analysed using multiple regression analysis. QALYs were distributed normally and analysed via linear regression. Because cost were strongly skewed, they were analysed using a generalized linear regression with gamma distribution and log-link function [31]. Compared to OLS-regression and bootstrapping, this approach provides more precise estimates, because the gamma-distribution shows a better fit to the underlying distribution. To analyse the cost-utility of CGI, we calculated the incremental cost-effectiveness ratio (ICER). To analyse the uncertainty of the ICER, the net-monetary-benefit (NMB) was calculated and analysed via linear regression using the same independent variables as for costs and QALYs [32]. Based on the results, cost-effectiveness acceptability curves were drawn that show the probability of the intervention to be cost-effective plotted against the willingness to pay for a health effect. Level of significance was set at p b 0.05. All analyses were carried out by SAS (version 9.2 for windows; copyright © 2002–2008 SAS Institute Inc., Cary, NC, USA). 2.4.1. Base case For base case analysis, we used an ITT approach, QALYs as measure of effectiveness and direct plus absenteeism costs. 2.4.2. Scenario analysis To test the robustness of our results, we analysed several scenarios. Firstly, we analysed the cost-utility for three subgroups defined by SSS at baseline according to the PHQ-15 as follows: mild (PHQ15 b 10), moderate (9 b PHQ-15 b 15), severe (14 b PHQ-15). Secondly, we used a per-protocol approach. According to the main study, the per-protocol sample was defined by the exclusion of 15 patients that did not finish the interventional therapy; i.e. patients who dropped out at T1 were excluded, whereas patients who missed some sessions but finished the therapy were included. Thirdly, we varied costs, by using a third-party payer's perspective (direct costs only), or by additionally adding presenteeism costs. Fourthly, we varied the measure of effectiveness by using the reduction by at least one SSS level. 3. Results 3.1. Sample characteristics Baseline sample characteristics are shown in Table 1. Patients were on average 49 years old and 75% were female, 64% were either married or in a relationship and 82% had an upper secondary educational level or more. The average PHQ-15 sum score was 12.6 and duration of illness was 36 months on average. We found statistically significant differences between the intervention and the control group regarding age, marital status, living situation, number of children and illness duration. 3.2. Health outcomes and costs Unadjusted SF-6D utility indices increased over time in both groups, but more pronounced in the intervention group (Table 2). QALYs were 0.666 in the intervention and 0.645 in the control group, corresponding to an unadjusted QALY gain of 0.021. During the 12-month study period, total costs were 5777EUR in the intervention and 6858EUR in the

46

A. Konnopka et al. / Journal of Psychosomatic Research 90 (2016) 43–50

Table 1 Sociodemographic and clinical sample characteristics at baseline.

Table 2 Unadjusted 6-month costs (in 2007 EUR) and SF-6D utility indices. p-Valuea

Variable

Total sample (N = 300)

Age, M (SD) Gender, N (%) Men Women Marital status, N (%) Single Married/partnership Divorced Widowed Unknown Living situation, N (%) Alone With partner Others (e.g. parents, friends) Unknown Number of children Education, N (%) ISCED level 2 ISCED level 3 ISCED level 4 Unknown Employed, N (%) Health insurance, N (%) Statutory health insurance Private health insurance PHQ-15 score, mean (SD) PHQ-15 somatic symptom severity levels, N (%) Minimal (PHQ-15 b 5) Mild (4 b PHQ-15 b 10) Moderate (9 b PHQ-15 b 15) Severe (14 b PHQ-15) Illness duration (months), M (SD) Psychotherapeutic pre-treatment, N (%) Yes No Unknown Charlson comorbidity index, M (SD) Panic disorder, N (%) Generalized anxiety disorder, N (%) Major depression, N (%) Minor depression, N (%) SF-36 physical component score, M (SD) SF-36 mental component score, M (SD)

48.8 (12.7) 50.8 (12.9) 46.1 (13.1) 0.001 0.986 74 (25) 42 (25) 32 (25) 226 (75) 128 (75) 98 (75) 0.001 61 (20) 24 (14) 37 (28) 192 (64) 126 (74) 66 (51) 29 (10) 14 (8) 15 (12) 13 (4) 4 (2) 9 (7) 5 (2) 2 (1) 3 (2) b0.001 66 (22) 27 (16) 39 (30) 203 (68) 132 (78) 71 (55) 22 (7) 9 (5) 13 (10)

Intervention Control group group (N = 170) (N = 130)

9 1.3

(3) 2 (1) 7 (5) (1.13) 1.48 (1.15) 1.02 (1.05) b0.001 0.663 50 (17) 26 (15) 24 (18) 209 (70) 123 (72) 86 (66) 36 (12) 19 (11) 17 (13) 5 (2) 2 (1) 3 (2) 165 (55) 86 (51) 79 (61) 0.079 0.387 290 (97) 163 (96) 127 (98) 10 (3) 7 (4) 3 (2) 12.6 (4.77) 12.6 (4.73) 12.7 (4.86) 0.832 0.868 3 84 113 100 35.6

(1) (28) (38) (33) (30.9)

1 47 65 57 40.4

(1) (28) (38) (34) (32.5)

2 37 48 43 29.3

(2) (28) (37) (33) (27.6) 0.002 0.953

82 183 35 0.34

(27) (61) (12) (0.71)

47 104 19 0.38

(28) (61) (11) (0.79)

35 79 16 0.30

(27) (61) (12) (0.58) 0.354

52 54

(17) (18)

31 27

(18) (16)

21 27

(16) (21)

0.637 0.340

32 (25) 24 (18) 42.0 (8.9)

0.339 0.244 0.276

66 (22) 47 (16) 42.7 (9.0)

34 (20) 23 (14) 43.2 (9.1)

41.1 (10.7) 41.5 (10.2) 40.7 (11.4) 0.485

Total sample

Intervention group (CGI)

Control group (EMC)

Difference CGI - EMC

Outpatient physician contacts, M (SE) T0 608 (66) 598 (78) T1 511 (50) 490 (51) T2 472 (43) 456 (52)

622 (63) 537 (69) 493 (55)

−24 −47 −36

Outpatient non-physician contacts, M (SE) T0 179 (18) 153 (20) T1 158 (16) 135 (18) T2 141 (16) 129 (19)

213 (31) 189 (27) 157 (28)

−60 −54 −28

Pharmaceuticals, M (SE) T0 187 (22) T1 173 (22) T2 176 (22)

225 (46) 214 (46) 205 (46)

−66 −72 −51

Inpatient hospital stays, M (SE) T0 301 (116) 377 (189) T1 223 (307) 122 (220) T2 236 (238) 207 (247)

201 (102) 357 (456) 275 (262)

176 −235 −67

Inpatient rehabilitation stays, M (SE) T0 71 (47) 82 (73) T1 713 (495) 524 (382) T2 578 (404) 581 (396)

57 (48) 961 (673) 574 (447)

25 −438 8

Other direct costs, M (SE)a T0 64 (8) T1 98 (18) T2 85 (13)

71 (12) 79 (11) 92 (21)

55 (9) 124 (40) 75 (14)

16 −46 17

Absenteeism costs, M (SE) T0 1191 (157) T1 1304 (214) T2 1171 (189)

1090 (206) 1172 (296) 1136 (260)

1323 (241) 1476 (276) 1218 (264)

−233 −304 −82

Presenteeism costs, M (SE) T0 4562 (341) T1 3936 (367) T2 4066 (355)

4253 (458) 3457 (417) 3471 (425)

4966 (496) 4562 (554) 4844 (562)

−713 −1105 −1373

Intervention costs, M (SE) T1 206 (11)

360 (7)

159 (16) 141 (15) 154 (16)

4 (0)

356

All costs except presenteeism, M (SE) T0 2602 (223) 2530 (309) T1 3386 (612) 3022 (564) T2 2859 (486) 2755 (523)

2696 (304) 3862 (821) 2996 (584)

−166 −840 −241

SF-6D utility index, M (SE) T0 0.635 (0.007) T1 0.662 (0.012) T2 0.669 (0.011)

0.630 (0.010) 0.649 (0.017) 0.651 (0.013)

0.008 0.023 0.031

0.638 (0.009) 0.672 (0.012) 0.682 (0.012)

M: mean; SE: standard error; T0: baseline; T1: 6 months (post-treatment); T2: 12 months (follow-up). a Includes home visits, phone contacts and travel costs.

Bold values indicate significance at p b 0.05. ISCE: international standard classification of education, M: mean, SD: standard deviation. a t-Test, chi-square-test.

3.3. Cost-effectiveness control group corresponding to cost savings of 1081EUR by the intervention. This difference was strongly driven by a noticeable increase of direct costs in the control group at T1. This increase was primarily driven by an increase of costs for rehabilitations, which occurred in both groups, but was more pronounced in the control group. Connected to the increase in rehabilitations, we also found a slight increase in absenteeism costs. On the other hand, outpatient physician and non-physician costs showed a slight decrease over time. Controlling for possible confounders (in particular baseline values), we found a small, but significant positive effect of the intervention on QALYs (+0.017; p = 0.019), but no significant effect of the intervention on costs (Table 3).

Being more effective and less costly, CGI dominated EMC in the base case point estimate. Uncertainty analysis showed the probability of the intervention to be cost-effective was 69% given a willingness to pay off 0EUR/QALY, increased to 92% for a willingness to pay of 50,000EUR/ QALY and reached the level of 95% at a willingness to pay of 70,375EUR/QALY (Fig. 2). Differentiating by baseline SSS levels revealed clear differences between patients starting with severe SSS, and patients starting with moderate or mild SSS. For patients starting with severe SSS, the probability of the intervention to be cost-effective was 81% for a willingness to pay off 0EUR/QALY and reached 95% given a willingness to pay of 73,665 EUR/QALY. Cost-acceptability curves for patients starting with mild or moderate SSS did not reach the 95%-level even at a willingness to pay of 150,000EUR/QALY.

A. Konnopka et al. / Journal of Psychosomatic Research 90 (2016) 43–50 Table 3 Regression analysis of costs and QALYs during 12-month study period. Variable

Constant CGI (ref: EMC) Baseline costs Baseline SF-6D utility index Age Female (ref: male) Employed (ref: not employed) Marital status (ref: single) Married Divorced Widowed Unknown

Costsa (GLM with gamma distribution and log-link function)

QALYsb (Linear regression)

p-Value Coefficient p-Value

8075 0.895 1.0001 0.374

0.000 0.278 0.000 0.143

0.356 0.017 b0.001 0.552

0.000 0.019 0.355 0.000

1.006 0.902 1.620

0.394 0.535 0.012

b0.001 0.009 0.007

0.513 0.249 0.378

1.129 0.834 1.206 1.721

0.654 0.462 0.650 0.228

−0.018 0.009 −0.009 −0.018

0.238 0.540 0.681 0.561

0.294 0.650

0.022 −0.010

0.133 0.532

0.478 0.856

−0.008 −0.001

0.762 0.722

0.404 0.005 0.446

−0.012 −0.034 −0.003

0.125 0.390 0.000

0.592 0.114

b0.001 −0.002

0.191 0.751

0.135 0.333

−0.018 0.002

0.087 0.862

Psychotherapeutic pre-treatment (ref: no) Yes 1.104 Unknown 3.288 Baseline PHQ-15 sum 1.010 score Illness duration 0.999 Charlson comorbidity 1.157 index Any depression disorder 1.259 Any anxiety disorder 0.874

reduction in SSS (intervention group: 49%, control group: 41%) as measure of effectiveness, the probability of the intervention to be costeffective did not reach the 95%-level at all. 4. Discussion

Exponentiated coefficientc

Living situation (ref: alone) With partner 0.788 Others (e.g. parents, 1.151 friends) Unknown 1.316 Number of children 0.989

47

4.1. Summary

Bold values indicate significance at p b 0.05. a Costs include direct as well as absenteeism costs. b QALYs are based on SF-6D utility values. c Except for the constant (absolute value), the exponentiated coefficient indicates the relative change associated with the variable.

Fig. 3 shows cost-effectiveness acceptability curves for different scenario analyses. Compared to the base case, all scenarios using QALYs but varying costs showed lower values of willingness to pay needed for a 95%-probability of the intervention to be cost-effective. Using the

We analysed the incremental cost-utility of a combination of a GP training in diagnosis and management of FSS plus a specific collaborative group intervention compared to only EMC based on the GP training, and found the combined intervention to be more effective and less costly. When estimates were controlled for potential confounders, a willingness to pay of 70,375EUR/QALY was needed for a probability of 95% for the intervention to be cost-effective. 4.2. Study population By using the ICD-10 concept of persistent and distressing somatic symptoms not sufficiently explained by peripheral organ pathology as inclusion criterion, our patient sample does only partly correspond to the SSD concept used in DSM-5, as in DSM-5 the contentious distinction between medically explained and medically unexplained symptoms has been given up, and positive psychobehavioral characteristics (Bcriteria) have been introduced instead. The sociodemographic characteristics of our sample were typical of FSS patients in primary care. Our patients had multiple, heterogeneous bodily complaints with a broad overlap between functional disturbance in organ systems, pain and fatigue/exhaustion. The distribution of the SSS was well-balanced in both analysed groups and represented patients from the whole severity spectrum of FSS. 4.3. Health effects and costs Our regression analysis showed a significant positive intervention effect on QALYs. The minimal important difference of the SF-6D drawn from data of 11 studies ranges from 0.011 to 0.097 with a mean of 0.04 indicating that our group difference is small, but relevant [33]. Compared to baseline, SF-6D utility indices improved modestly, but more pronounced in the intervention group. A comparison of retrospective baseline costs with the costs during treatment and follow-up shows that even if intervention costs were ignored costs slightly increased over time in both groups, but

Fig. 2. Cost-effectiveness acceptability curves for base-case and somatic symptom severity levels (PHQ-15).

48

A. Konnopka et al. / Journal of Psychosomatic Research 90 (2016) 43–50

Fig. 3. Cost-effectiveness acceptability curves for base-case and scenario analyses.

considerably stronger in the control group. This finding was unexpected, as we expected that the achieved reduction in SSS would be associated with a decrease in costs. A differentiated look at single cost categories revealed that costs did indeed slightly decrease in outpatient services but considerably increased for rehabilitations. This increase might be interpreted as an effect of the physician training, which increased the physician's awareness for FSS and induced rehabilitation prescriptions. In this view, the difference in rehabilitation prescriptions between intervention and control group might be interpreted as follows: whereas physicians in the intervention group could offer the patients a psychotherapy, physicians in the control group had no such opportunity, which might have resulted in a higher percentage of rehabilitation prescriptions. However, since rehabilitations can (regularly) only be prescribed once every four years in Germany, this effect can at least for the German health care system - be regarded singular. Both groups showed a slight increase of absenteeism costs at T1, which is in part connected to the increase in inpatient rehabilitations, but comparing T2 with baseline costs, costs in both groups remained almost unchanged. Also, after controlling for baseline-differences there were no relevant differences between CGI and EMC. Presenteeism costs however showed a different picture. First, presenteeism costs showed a clear decrease in the intervention group, but almost no decrease in the control group. Second, connected to this finding, presenteeism costs were also lower for CGI than EMC after controlling for baseline differences. One explanation for these findings may be that GPs in both groups equally restrict certifications of sick leave as a result of the GP training they had received, resulting in a relatively small difference in absenteeism costs between intervention and control group. On the other hand, presenteeism costs are more strongly associated with patient behaviour, enabling a more pronounced effect of CGI on these costs [34]. 4.4. Cost-effectiveness As point estimate CGI dominated EMC. To show the uncertainty surrounding this point estimate, we used a NMB-approach. Restricting the analysis to direct costs (payer's perspective) or adding presenteeism costs yielded considerably better 95% cut-offs compared to the base case. This finding resulted from the cost differences discussed above. Whereas direct costs and presenteeism costs showed a clear costdifference between CGI and EMC in favour of CGI, there was almost no cost difference between CGI and EMC regarding absenteeism costs. Therefore both scenarios (neglecting absenteeism costs or adding

presenteeism costs) resulted in lower p-values for the effect of CGI on net-benefit and consequently a higher probability of CGI to be costeffective. Subgroup analyses of different SSS levels at baseline showed that the probability of cost-effectiveness was highest for patients with severe SSS. This finding was primarily driven by effectiveness which showed increasing QALY gains with increasing baseline SSS. Thereby, it supports guideline recommendations that patients with mild, uncomplicated courses of FSS should be treated particularly by the GP, and patients with more severe FSS should additionally be offered specific psychotherapeutic interventions [4]. However, our follow-up may have been too short to reflect differences in disease progress between the treatment groups. Thus, although we found that CGI is not cost-effective in mild and moderate cases, one might speculate that cost-utility of CGI for mild and moderately ill patients might be much better in a longterm perspective, given that CGI would reduce disease progression in these patients. Until now, some full economic evaluations of CBT-based interventions have been conducted in the field of FSS, mostly focussing on specific syndromes; e.g. fibromyalgia [8], chronic fatigue [10–12,35], irritable bowel syndrome [9,36] or health anxiety [6]. The study most comparable to ours stems from Ravesteijn et al. who evaluated mindfulness-based CBT for FSS patients using SF-6D based QALYs as measure of effectiveness [5]. As we, Ravesteijn et al. found only a small inter-group difference in QALYs of 0.012, which was connected to an ICER of 56,637EUR/QALY from a societal perspective, compared to cost savings in our study. The results of the different studies lack comparability for different reasons (e.g. different health care systems, different syndromes, different measures of effectiveness). Nevertheless there is a tendency for CBT-based interventions to either dominate its comparators or to show ICERs clearly below commonly accepted thresholds for cost-effectiveness. The highest ICER aside the one found by Ravesteijn et al. was measured by Severens et al. [12] who compared CBT with guided support groups and natural disease course for patients with chronic fatigue. Whilst CBT dominated guided support groups, the ICER of CBT compared to natural course was 21,375 per QALY. 4.5. Strengths and limitations Strengths of this study are (1) a naturalistic primary care recruitment setting; (2) a large sample size, and (3) a broad assessment of cost categories. The study has the following limitations: (1) Health care utilisation and loss of productivity were measured via self-report

A. Konnopka et al. / Journal of Psychosomatic Research 90 (2016) 43–50

questionnaires, which may be subject to memory bias. (2) Presenteeism was measured on a 10-point Likert-scale adapted from Reilly et al. which has not been sufficiently validated until now [24]. (3) The time horizon of the study was limited to 12 months. 4.6. Implications for clinical practice In relation to stepped and collaborative care models according to severity profiles as recommended by current guidelines [4], our collaborative approach could help to bridge the gaps between primary care and specialized psychotherapy; regarding severity profiles we found that our intervention was particularly cost-effective for patients with severe somatic symptom severity. The finding that presenteeism crucially influences incremental cost-utility calls for preventive strategies in occupational medicine – also in allegedly mild cases. Applying therapy for FSS patients in group formats enhances its economic appeal, as intervention costs per patient are lower. 5. Conclusion

49

Appendix Table 1 (continued)

Sum per patient (n = 301)

Quantity Unit cost

Costs

–

4.21

–

Training and supervision for group therapy (intervention group only) Training of physicians Personal (1 physician, 1 psychologist) 2 ∗ 12 h 2 ∗ 28 per hour Room rent (1 group à 12 h) 1 200 Printed materials 18 10 Sum – – Sum per patient (n = 171) – – Supervision of physicians: 3 groups á 3 sessions á 1, 5 13.5 h 2 * 28 h (1 physician, 1 psychologist) Supervision of supervisors: 3 groups á 3 sessions á 0, 6.75 h 75 h Sum – Sum per patient (n = 171) – Total cost per patient Intervention group Control group

672 200 180 1052 6.15 756

1 ∗ 28

189

– –

945 5.52 15.88 4.21

Earnings of 28EUR per hour are according to tariff BAT IIa; costs for room rent and printed materials are assumptions.

CGI added to a GP training for FSS patients seems to be cost-effective, in particular for patients with severe symptoms. This finding is primarily due to the intervention's effectiveness. At least for the time horizon considered, the intervention was not associated with cost savings, but rather a slight increase of costs. Authors' contributions WH, HHK, JS, BW, RS, and CK conceived and designed the trial. WH, HHK, AK, NS, RS, CK and DS conceived the idea for this paper. WH, BW, JS, RS and HHK obtained funding and supervised the study. CK, DS and RS collected the data. AK and NS analysed the data. All authors interpreted the data. AK and RS drafted the manuscript. All authors critically revised the manuscript and approved the final version. AK and RS had full access to all data and take responsibility for the integrity of the data and the accuracy of the data analysis. Conflict of interest All authors report no financial or other relationships relevant to the subject of this article. Ethic committee approval The study was approved be the ethics committee of the University of Heidelberg. Acknowledgements The authors thank all involved GPs, their staff, the participating patients, and the contributing research assistants. The study was funded by the German Federal Ministry of Education and Research (BMBF), Grant Number 01GK0601 (speciAL). Appendix A. Appendix

Appendix Table 1 Calculation of intervention fix costs in Euro. Quantity Unit cost

Costs

GP-training (control and intervention group) Personal (1 physician, 1 psychologist)

2 ∗ 15.5

28 per hour

868

Room rent (1 group à 15, 5 h) Printed materials Sum

h 1 40 –

200 5 –

200 200 1268

References [1] P. Henningsen, S. Zipfel, W. Herzog, Management of functional somatic syndromes, Lancet 369 (2007) 946–955. [2] M.W. de Waal, I.A. Arnold, J.A. Eekhof, A.M. van Hemert, Somatoform disorders in general practice: prevalence, functional impairment and comorbidity with anxiety and depressive disorders, Br. J. Psychiatry J. Ment. Sci. 184 (2004) 470–476. [3] A. Konnopka, R. Schaefert, S. Heinrich, C. Kaufmann, M. Luppa, W. Herzog, H.-H. König, Economics of medically unexplained symptoms: a systematic review of the literature, Psychother. Psychosom. 81 (2012) 265–275. [4] C.M. van der Feltz-Cornelis, R. Hoedeman, E.J. Keuter, J.A. Swinkels, Presentation of the multidisciplinary guideline medically unexplained physical symptoms (mups) and somatoform disorder in The Netherlands: disease management according to risk profiles, J. Psychosom. Res. 72 (2012) 168–169. [5] H. van Ravesteijn, J. Grutters, T. olde Hartman, P. Lucassen, H. Bor, C. van Weel, v. der Wilt GJ, A. Speckens, Mindfulness-based cognitive therapy for patients with medically unexplained symptoms: a cost-effectiveness study, J. Psychosom. Res. 74 (2013) 197–205. [6] E. Hedman, E. Andersson, N. Lindefors, G. Andersson, C. Ruck, B. Ljotsson, Costeffectiveness and long-term effectiveness of Internet-based cognitive behaviour therapy for severe health anxiety, Psychol. Med. 43 (2013) 363–374. [7] H. Seivewright, J. Green, P. Salkovskis, B. Barrett, U. Nur, P. Tyrer, Cognitivebehavioural therapy for health anxiety in a genitourinary medicine clinic: randomised controlled trial, Br. J. Psychiatry J. Ment. Sci. 193 (2008) 332–337. [8] J.V. Luciano, F. D'Amico, M. Cerda-Lafont, M.T. Penarrubia-Maria, M. Knapp, A.I. Cuesta-Vargas, A. Serrano-Blanco, J. Garcia-Campayo, Cost-utility of cognitive behavioral therapy versus U.S. Food and drug administration recommended drugs and usual care in the treatment of patients with fibromyalgia: an economic evaluation alongside a 6-month randomized controlled trial, Arthritis Res. Ther. 16 (2014) 451. [9] E. Andersson, B. Ljotsson, F. Smit, B. Paxling, E. Hedman, N. Lindefors, G. Andersson, C. Ruck, Cost-effectiveness of internet-based cognitive behavior therapy for irritable bowel syndrome: results from a randomized controlled trial, BMC Public Health 11 (2011) 215. [10] P. McCrone, L. Ridsdale, L. Darbishire, P. Seed, Cost-effectiveness of cognitive behavioural therapy, graded exercise and usual care for patients with chronic fatigue in primary care, Psychol. Med. 34 (2004) 991–999. [11] P. McCrone, M. Sharpe, T. Chalder, M. Knapp, A.L. Johnson, K.A. Goldsmith, P.D. White, Adaptive pacing, cognitive behaviour therapy, graded exercise, and specialist medical care for chronic fatigue syndrome: a cost-effectiveness analysis, PLoS One 7 (2012), e40808. [12] J.L. Severens, J.B. Prins, G.J. van der Wilt, J.W. van der Meer, G. Bleijenberg, Costeffectiveness of cognitive behaviour therapy for patients with chronic fatigue syndrome, QJM Mon. J. Assoc. Physicians 97 (2004) 153–161. [13] R. Schaefert, C. Kaufmann, B. Wild, D. Schellberg, R. Boelter, R. Faber, J. Szecsenyi, N. Sauer, E. Guthrie, W. Herzog, Specific collaborative group intervention for patients with medically unexplained symptoms in general practice: a cluster randomized controlled trial, Psychother. Psychosom. 82 (2013) 106–119. [14] R. Schaefert, C. Kaufmann, R. Bölter, R. Faber, B. Wild, J. Szecsenyi, N. Sauer, W. Herzog, Somatoform syndromes in primary care. Manual for interdisciplinary group therapy, Psychotherapeut 56 (2011) 315–324. [15] K. Christensen, P. Fink, T. Toft, L. Frostholm, E. Ornbol, F. Olesen, A brief case-finding questionnaire for common mental disorders: the cmdq, Fam. Pract. 22 (2005) 448–457. [16] P. Henningsen, N. Hartkamp, T. Loew, M. Sack, C.E. Scheidt, G. Rudolf, Guidelines and Source book [somatoforme störungen. Leitlinien und quellentexte], Schattauer, Stuttgart, 2002.

50

A. Konnopka et al. / Journal of Psychosomatic Research 90 (2016) 43–50

[17] R. Schaefert, G. Benedikt, N. Sauer, S. Wilke, W. Herzog, J. Szecsenyi, S. Zipfel, P. Henningsen, Funktional project early recognition and treatment of functional/ somatoform complaints in primary care, Notfall Hausarztmedizin 31 (2005) 580–589. [18] S. Wessely, P.D. White, There is only one functional somatic syndrome, Br. J. Psychiatry J. Ment. Sci. 185 (2004) 95–96. [19] M.E. Charlson, P. Pompei, K.L. Ales, C.R. MacKenzie, A new method of classifying prognostic comorbidity in longitudinal studies: development and validation, J. Chronic Dis. 40 (1987) 373–383. [20] J.E.J. Ware, C.D. Sherbourne, The mos 36-item short-form health survey (sf-36). I. Conceptual framework and item selection, Med. Care 30 (1992) 473–483. [21] J.E. Brazier, J. Roberts, The estimation of a preference-based measure of health from the sf-12, Med. Care 42 (2004) 851–859. [22] K. Kroenke, R.L. Spitzer, J.B. Williams, The phq-15: validity of a new measure for evaluating the severity of somatic symptoms, Psychosom. Med. 64 (2002) 258–266. [23] D. Chisholm, M.R. Knapp, H.C. Knudsen, F. Amaddeo, L. Gaite, B. van Wijngaarden, Client socio-demographic and service receipt inventory—European version: development of an instrument for international research. Epsilon study 5. European psychiatric services: inputs linked to outcome domains and needs, Br. J. Psychiatry Suppl. (2000) s28–s33. [24] M.C. Reilly, A.S. Zbrozek, E.M. Dukes, The validity and reproducibility of a work productivity and activity impairment instrument, PharmacoEconomics 4 (1993) 353–365. [25] C. Krauth, F. Hessel, T. Hansmeier, J. Wasem, R. Seitz, B. Schweikert, [Empirical standard costs for health economic evaluation in Germany — a proposal by the working group methods in health economic evaluation], Gesundheitswesen 67 (2005) 736–746. [26] K. Bundesvereinigung, Gebührenordung für ärzteDownload: https://www. Aerztekammer-bw.De/20/goae/volltext.Pdf (access: May/2009).

[27] Deutschland. RLAf: Rote liste service gmbh. Frankfurt am main: Ecv editio cantor verlag, 2009. [28] S. Bundesamt, Verdienste und arbeitskosten 2006, Statistisches Bundesamt, Wiesbaden, 2009. [29] K. Bundesvereinigung, Einheitlicher bewertungsmaßstab (ebm)Download: http:// www.Kbv.De (access: June/2013). [30] M.J. Azur, E.A. Stuart, C. Frangakis, P.J. Leaf, Multiple imputation by chained equations: what is it and how does it work? Int. J. Methods Psychiatr. Res. 20 (2011) 40–49. [31] W.G. Manning, A. Basu, J. Mullahy, Generalized modeling approaches to risk adjustment of skewed outcomes data, J. Health Econ. 9 (2004) 197–204. [32] N. Zethraeus, M. Johannesson, B. Jonsson, M. Lothgren, M. Tambour, Advantages of using the net-benefit approach for analysing uncertainty in economic evaluation studies, PharmacoEconomics 21 (2003) 39–48. [33] S.J. Walters, J.E. Brazier, Comparison of the minimally important difference for two health state utility measures: eq-5d and sf-6d, Qual. Life Res. 14 (2005) 1523–1532. [34] A. Konnopka, C. Kaufmann, H.H. Konig, D. Heider, B. Wild, J. Szecsenyi, W. Herzog, S. Heinrich, R. Schaefert, Association of costs with somatic symptom severity in patients with medically unexplained symptoms, J. Psychosom. Res. 75 (2013) 370–375. [35] M.J. de Boer, G.J. Versteegen, K.M. Vermeulen, R. Sanderman, M.M. Struys, A randomized controlled trial of an internet-based cognitive-behavioural intervention for non-specific chronic pain: an effectiveness and cost-effectiveness study, Eur. J. Pain 18 (2014) 1440–1451. [36] P. McCrone, M. Knapp, T. Kennedy, P. Seed, R. Jones, S. Darnley, T. Chalder, Costeffectiveness of cognitive behaviour therapy in addition to mebeverine for irritable bowel syndrome, Eur. J. Gastroenterol. Hepatol. 20 (2008) 255–263.