Journal of Psychosomatic Research 72 (2012) 242–247

Contents lists available at SciVerse ScienceDirect

Journal of Psychosomatic Research

Healthcare costs incurred by patients repeatedly referred to secondary medical carewith medically unexplained symptoms: A cost of illness study

Christopher Burton a, Kelly McGorm a, Gerry Richardson b, David Weller a, Michael Sharpe c,⁎a Centre for Population Health Sciences, University of Edinburgh, United Kingdomb Centre for Health Economics, University of York, United Kingdomc Psychological Medicine Research, Department of Psychiatry, University of Oxford, United Kingdom

⁎ Corresponding author at: Psychological Medicine RDepartment of Psychiatry, Warneford Hospital, OxforTel.: +44 1865 226397; fax: +44 1865 793101.

E-mail address: michael.sharpe@psych.ox.ac.uk (M.

0022-3999/$ – see front matter © 2011 Elsevier Inc. Alldoi:10.1016/j.jpsychores.2011.12.009

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 9 June 2011Received in revised form 13 December 2011Accepted 20 December 2011

Keywords:Cost of illnessMUSPrimary careReferralsSomatoform disorders

Background: Some patients are repeatedly referred from primary to secondary care with medically unex-plained symptoms (MUS). We aimed to estimate the healthcare costs incurred by such referrals and to com-pare them with those incurred by other referred patients from the same defined primary care sample.Methods: Using a referral database and case note review, all adult patients aged less than 65 years, who hadbeen referred to specialist medical services from one of five UK National Health Service primary care practicesin a five-year period, were identified. They were placed in one of three groups: (i) repeatedly referred withMUS (N=276); (ii) infrequently referred (IRS, N=221), (iii) repeatedly referred with medically explainedsymptoms (N=230). Secondary care activities for each group (inpatient days, outpatient appointments,emergency department attendances and investigations) were identified from primary care records. The asso-ciated costs were allocated using summary data and the costs for each group compared.

Results: Patients who had been repeatedly referred with MUS had higher mean inpatient, outpatient andemergency department costs than those infrequently referred (£3,539, 95% CI 1458 to 5621, £778 CI 705 to852 and £99, CI 74 to 123 respectively. The mean overall costs were similar to those of patients who hadbeen repeatedly referred with medically explained symptoms.Conclusions: The repeated referral of patients with MUS to secondary medical care incurs substantialhealthcare costs. An alternative form of management that reduces such referrals offers potential costsavings.

© 2011 Elsevier Inc. All rights reserved.

Introduction

Many patients attending medical services have symptoms whichcannot be adequately explained by organic disease, so-called medi-cally unexplained symptoms (MUS) [1,2]. This clinical descriptionincludes a range of symptom syndromes such as irritable bowel syn-drome, fibromyalgia, and non-cardiac chest pain and tension typeheadache [3–6]. Patients with MUS may be high users of medical ser-vices [7–11]: whilst some are high users of primary care [12], thegreatest costs are incurred by those who are high users of secondarycare. One cause of a high use of secondary care is the repeated referralto specialists by their primary care doctor [13]. Patients with MUS arehowever unlikely to benefit from such repeated referral as secondarycare medical services are designed primarily to identify and treat dis-ease whereas the patients with MUS: (a) are likely to have alreadyhad many previous negative investigations and are unlikely to be

esearch, University of Oxfordd OX3 7JX, United Kingdom.

Sharpe).

rights reserved.

reassured by more tests [14] and (b) have symptoms that are morelikely to be relieved by addressing causes other than disease, suchas depression and anxiety [15].

The decision of the primary care doctor to refer a patient to sec-ondary care is a critical and potentially modifiable step in increasingthe costs of care. That is why this study focuses on referrals. Alterna-tives to repeated referral such as enhanced assessment and bettertreatment of depression and anxiety in primary care could potentiallybe more effective and cost-effective [16]. Therefore knowledge of thesecondary care costs incurred by repeated referral indicates the po-tential savings that could be achieved by such an alternative manage-ment strategy.

We have reported elsewhere our findings on a novel sample ofpatients selected from a defined primary care population as havingbeen repeatedly referred to secondary care (defined as at least threetimes in five years) where they had received repeated diagnoses ofMUS. We found the prevalence of such patients to be approximately1% (1.1%, 95%CI 1.0–1.2) of 26,252 patients aged between 18 and65 years who were registered at one of five primary care practices[13]. Approximately half of these patients had an anxiety or depres-sive disorder [15].

243C. Burton et al. / Journal of Psychosomatic Research 72 (2012) 242–247

In this paper we use data obtained from this sample together withstandard cost estimates to estimate the costs incurred in the second-ary care of such patients. In order to aid interpretation of these costswe compared them to those incurred by: (a) patients who had beeninfrequently referred; (b) patients who had been repeatedly referredfor symptoms that the specialist had concluded were symptoms ofdisease (medically explained symptoms). The infrequently referredcomparison group provides an indication of the savings that couldbe made by reducing the number of referrals. The repeated referralfor medically explained symptoms (MES) comparison group providesan estimate of the costs of care of patients similarly frequently re-ferred but with needs better served by disease focused secondarycare.

Methods

The study was based in five National Health Service general pri-mary care practices in Edinburgh, UK. The practices comprised 30 pri-mary care doctors and 39,562 registered patients. Data collection tookplace between March 2003 and October 2005.

Identification of patients

We used national secondary care activity data (SMR00; ScottishMorbidity Records, Information Services Division of NHS Scotland)to identify all the patients from the participating practices who hadbeen newly referred to a range of medical specialties at least threetimes over the previous five years. We only included specialtieswhich commonly accept referrals for outpatient diagnosis of symp-toms, rather than those concerned with treatment of already diag-nosed problems (such as cardiothoracic surgery, palliative medicine,obstetrics or oncology); a list of eligible specialties is given in thefootnote to Fig. 1. We chose a threshold of at least three new referralsbecause this identified the most frequently referred 10% of patients in

Patients registered with pr39,562

Age 18-65 and one or more14,034

Excluded at stage of searpractice records 2 1,37

3 or more referrals todesignated clinics1 1,312

Not referred in 5th year of3,721

RRMUS267

RRMUS221

Fig. 1. Flowchart showing the stage

our population. It also defined approximately 1% of the practice pop-ulation as frequently referred with MUS; a number of patients thatpresent a reasonable target for alternative management strategies inprimary care.

We then linked this referral data to the individual practice data-bases using a unique patient identifier to identify the frequently re-ferred patients. Within each practice, a researcher (KM) searchedboth handwritten and computerised records to extract data on eachreferral; this included the reason for the referral, the specialty re-ferred to and the specialist's final diagnosis. Only referrals for theassessment of symptoms to an eligible specialist were included. Thespecialists' final diagnosis for that referral episode was categorisedas being one of MUS or MES according to pre-specified criteria.Where there was uncertainty, cases were adjudicated jointly by apsychiatrist and a primary care practitioner (MS and DW). Thesemethods had been tested in an earlier pilot study [17] and havebeen reported in detail elsewhere [13]. We tested the inter-raterreliability of identifying a referral outcome as MUS or MES on a ran-dom sample of 20 cases and found good agreement (pooled kappa0.76). We also identified a sample of patients from the participatingpractices who had only been referred once in five years (the finaldiagnoses of these patients may have been MES or MUS).

Patients

Using the data obtained from case note review patients were allo-cated to one of three groups: (i) patients who had been repeatedlyreferred with MUS; these patients had been referred at least threetimes in the five-year study period and at least two of these referralsled to a final specialist diagnosis of MUS; (ii) patients who had beeninfrequently referred to secondary care; these patients had beenreferred only once in the five years; (iii) patients who had been re-ferred at least three times in the five-year study period and all ofthese referrals led to a final specialist diagnosis of MES.

actice

referral

Outside age range or not referred25,528

ching8

Only one referral todesignated clinics1

4,505

Ineligible number of referrals todesignated clinics1 8,217

data

IRS230

s of recruitment to the study.

Table 1Healthcare costs used in the model.

Health care activity Cost (£)

Clinical encountersInpatient day 409New outpatient appointment 93Follow-up outpatient appointment 62Missed outpatient appointment 46Unscheduled emergency department attendance 52

Specific investigationscolonoscopy/sigmoidoscopy 171CT scan 209Endoscopy 171MRI 309Ultrasound 100

CT: computerised tomography; MRI: magnetic resonance imaging;

244 C. Burton et al. / Journal of Psychosomatic Research 72 (2012) 242–247

Data collection

We measured healthcare usage for each patient group by record-ing the number and content of contacts documented in the hospitalcorrespondence section of the primary care medical records overthe preceding five years. Specifically, we recorded inpatient days,new outpatient attendances and follow-up appointments. The num-ber of missed appointments at outpatient clinics was counted fromletters from clinics to the practice. Unscheduled contacts at hospitalemergency departments were also noted from the correspondencein the primary care notes. We also recorded the specific number of in-vestigations (including laboratory tests, imaging and procedures suchas endoscopy) resulting from each outpatient referral. However wedid not record investigations that occurred during inpatient stays orwhich were directly ordered by the primary care doctor.

Allocation of costs

We estimated average costs for an inpatient day and for an outpa-tient appointment using unit costs available from the InformationServices Division of NHS Scotland [18]. These average costs used aweighted average, obtained by weighting the unit cost of appoint-ments or inpatient days for each specialty by the number of appoint-ments or inpatient days for that specialty in Scotland. Follow-upappointments were costed at two-thirds the cost of a new outpatientappointment [19].

We also estimated the cost of specific investigations from the re-ferral by applying unit costs for all investigations costing more than£100 at the time of the study. Because the average cost of investiga-tions per referral is already included in the outpatient specialtycosts we report the costs of specific investigations separately. Allcosts used in the model are summarised in Table 1.

Table 2Resource use per patient by study group over the five-year period.

Group RRMUS

(N=267)

Mean (95% ci)

Inpatient day 9.38 (4.3 to 14.46)New outpatient appointment 4.931 (4.71 to 5.15)Follow-up outpatient appointment 7.18 (6.22 to 8.14)Missed outpatient appointment 1.14 (0.89 to 1.39)Emergency department attendance 2.35 (1.89 to 2.81)

RRMUS: repeatedly referred with MUS; RRMES: repeatedly referred with medically explain

Sensitivity analysis

We repeated the analysis using comparable outpatient costs forthe NHS in England [19], (inpatient costs are reported differently be-tween the two systems so we were unable to compare these).

Statistical analysis

All data were manually entered into a database and analysed in R2.9.1. We compared healthcare contacts, total and of each type, be-tween groups using the difference in means with 95% confidence in-tervals derived from t-tests [20] and by non-parametric bootstrapmethods [21].

Ethical approval

The study received ethical approval from the local National HealthService Research Ethics Committee.

Results

Participants

14,034 patients, representing 53% of the practices' population aged between 18and 64, were identified in the hospital activity database as having had at least one re-ferral to specialist medical care over the five-year study period. Fig. 1 shows the num-bers at each stage of recruitment of the samples. 718 patients met our inclusion criteriafor referrals for symptoms; of these 267 had been repeatedly referred for MUS; 221had been infrequently referred and 230 had been repeatedly referred for MES.

More details about the characteristics of these patients are given in a previous pub-lication [13]. The patients who had been repeatedly referred for MUS had received amean of 4.9 referrals (range 3 to 13) and consequent episodes of secondary care inthe five-year study period. The number and percentage of patients in each groupwho had been referred to each secondary care speciality is shown in the Appendix.

Number and estimated cost of health care contacts

The inpatient and outpatient resources used during these referral episodes areshown in Table 2. Perhaps unsurprisingly the resource use of the patients who hadbeen repeatedly referred with MUS was much greater than that of the infrequentlyreferred patients. However it was also similar to that of the patients who had been re-peatedly referred with MES. The distribution of inpatient days was skewed: 15 pa-tients (across all three groups) had between 30 and 60 inpatient days and five(three patients repeatedly referred with MUS and two patients repeatedly referredwith MES) had between 92 and 612 days.

The costs associated with this resource use are summarised in Table 3. The tableshows the mean cost incurred over five years for outpatient, inpatient and emergencydepartment treatment. Specific outpatient investigations ordered by the specialists arealso shown separately and described in greater detail in Table 4.

Within the sample repeatedly referred with MUS (n=273) the majority (69%) hadbeen referred five or fewer times in the five-year period (that is an average of onlyonce per year). The remaining 31% had been referred more frequently (up to 13times). It is notable that the total costs incurred by these two groups were similar;the total cost of the relatively infrequent referral of a greater number of patients wassimilar to that of the smaller number of very frequent referrals.

Similar values to those reported above were obtained by non-parametric boot-strapping. A sensitivity analysis replacing Scottish outpatient costs with comparablevalues from England also made no meaningful difference to the findings.

IRS RRMES

(N=221) (N=230)

Mean (95% ci) Mean (95% ci)

0.73 (0.39 to 1.06) 8.18 (6.19 to 10.17)1.00 (0.99 to 1.00) 4.17 (4.00 to 4.35)1.18 (0.88 to 1.47) 8.13 (7.04 to 9.21)0.28 (0.17 to 0.38) 0.86 (0.64 to 1.08)0.45 (0.35 to 0.56) 1.92 (1.60 to 2.25)

ed symptoms; IRS: infrequently referred with symptoms.

Table 3Average healthcare costs (£) incurred per patient by study group over the five-year period.

Item Average costs per patient (£) Comparison between groups

RRMUS IRS RRMES RRMUS vs IRS RRMUS vs. RRMES

Difference 95% CI Difference1 95% CI

Estimated overall costs2

Inpatient costs 3837 296 3,347 3539 (1458 to 5621) 491 (−1737 to 2718)Outpatient costs 957 178 932 778 (705 to 852) 25 (−78 to 127)Emergency Department costs 112 24 100 99 (74 to 123) 22 (−7 to 52)Total 4916 500 4379 4416 (2315 to 6517) 537 (−1723 to 2798)

Specific costs3

Investigation costs 319 59 216 260 (224 to 296) 102 (56 to 149)

RRMUS: repeatedly referred with MUS; RRMES: repeatedly referred with medically explained symptoms; IRS: infrequently referred with symptoms.1 Bootstrap mean and confidence intervals as follows: inpatient costs 524 (−1375 to 3083); outpatient costs 23 (−82 to 119); emergency department costs 21 (−7 to 51); total

costs 551 (−1359 to 3128).2 Estimated costs were obtained by multiplying outpatient appointments and inpatient days by weighted average costs.3 Actual costs were obtained by multiplying each investigation by its unit cost. Estimated outpatient costs include an element for investigation, therefore actual investigations

costs were not added to these.

245C. Burton et al. / Journal of Psychosomatic Research 72 (2012) 242–247

Discussion

Main findings

This study examined referrals from primary to secondary medicalcare from a defined primary care population and found that the re-peated referral of patients with symptoms deemed by the assessingspecialist to be MUS incurs substantial costs. These costs are, notsurprisingly, much higher than those incurred by patients who wereinfrequently referred; the difference allows an estimate of the maxi-mum cost saving that could be made if the patients with MUS wereonly referred once. The overall cost of repeatedly referring patientswith MUS was similar to that of repeatedly referring patients withsymptoms of disease; the specific cost incurred by investigationswas even greater for patients with MUS.

Strengths and limitations

A strength of this study was the defined sample of primary carepatients that included a broad socio-economic range. Referrals andcosts were also systematically identified using record linkage andhand-searching of primary care records which ensured completenessof the data. We used the final opinion of the hospital specialist todecide whether patients' symptoms were medically explained ornot, rather than that of the referring primary care doctor, which islikely to be less accurate [22].

The study also had limitations: Our sample was recruited fromonly one UK city. We chose a threshold for frequent referrals of atleast three referral episodes in five years for the reasons given; how-ever we acknowledge that other thresholds could have been chosen.Our main cost estimates used average rather than speciality specificcosts. We included all secondary care costs, although we acknowl-edge that inpatient costs may not have been as sensitive to doctors'referral behaviour as outpatient costs or investigations. We did not

Table 4Number of investigations and number of patients receiving them over the five-year period

Test RRMUS (N=273) IRS (N

Tests Patients Cost (£) Tests

CT scan 65 60 22% 13,585 9MRI 92 67 25% 28,428 10Ultrasound 162 117 44% 16,200 25colonoscopy/sigmoidoscopy 85 65 24% 14,535 13Upper GI endoscopy 72 55 21% 12,312 19Total 476 364 85,060 76

RRMUS: repeatedly referred with MUS; RRMES: repeatedly referred with medically etomography; MRI: magnetic resonance imaging; GI: gastrointestinal.

attempt to measure primary care costs; however these were likelyto be relatively small (and in a capitation-based remuneration systemsuch as the UK NHS, little influenced by activity). We did not collectdata on attendance at outpatient clinics other than those listed; pre-scriptions or on other healthcare costs incurred privately by patients,such as alternative therapies. Nor did we measure resources usedoutside the health care setting (even though it is likely that the fre-quently referred patients consumed resources of social care resourcesand other sectors). Whilst this means that we have adopted a rela-tively narrow cost perspective, we argue that the inclusion of theseother costs would be likely to increase the relative cost of caring forpatients who had been repeatedly referred with MUS.

Studies of healthcare costs are inevitably specific to the costs andnature of the health care system studied. Are these finding generalisa-ble? First we carried out a sensitivity analysis using English ratherthan Scottish NHS costs; the findings of this reanalysis were not sub-stantially different. Second,we argue that thefindings are potentially gen-eralisable to other countries with similar primary care systems; that isthose with registered patient lists and a strong gatekeeper role as regardsaccess to secondary care. Third, whilst referral practices may have chan-ged since the study was done we think it unlikely that the findingswould be markedly different if the study was repeated today.

In order to put the costs of referral in context we compared themwith those of two other groups from the same sample whomet differ-ent criteria but who were selected using the same procedure. Howev-er neither is a perfect control. In particular the infrequently referredpatient group included those with referrals for both medicallyexplained and unexplained symptoms.

We have found that high costs are incurred by the repeated refer-ral of patients with MUS. We have assumed that the associated clini-cal activities were, at least in part, unnecessary to improving thepatient's outcome. However we could only be sure that these costswere unjustified by undertaking a prospective comparison of alter-native management strategies. Nonetheless whilst a small number

.

=221) RRMES (N=230)

Patients Cost (£) Tests Patients Cost (£)

8 4% 1881 36 29 13% 752410 5% 3090 47 35 15% 14,52322 10% 2500 108 77 33% 10,80011 5% 2223 47 34 15% 803719 9% 3249 52 41 18% 889270 12,943 290 216 49,776

xplained symptoms; IRS: infrequently referred with symptoms; CT: computerised

246 C. Burton et al. / Journal of Psychosomatic Research 72 (2012) 242–247

of referrals may help the primary care doctor to exclude disease, thevery phenomenon of repeated referral suggests that the previous re-ferrals were unhelpful. This lack of benefit is in keeping with what isknown about the limited value of reassurance [23]. Patients whohave been repeatedly referred with MUS have high rates of depres-sive and anxiety disorders, which frequently go untreated [14].There is also some evidence that alternative psychiatric and psycho-logical management strategies for patients with multiple chronicMUS (also referred to as somatization disorder) can be more effec-tive than usual care, at least within the North American health caresystem [16]. In summary, it seems most unlikely that the costs in-curred in repeatedly assessing patients referred with MUS were jus-tified by improved patient outcomes.

Comparison with other studies

Other studies of costs related to the management of patients withMUS have taken different approaches: they have investigated the healthcare costs in frequent attendees at secondary care clinics [9], in medicalinpatients [10], and in the general population [11]. However, no previousstudy has examined what we regard as the critical issue in determiningcosts; the decision by the primary care doctor to refer the patient fromprimary to secondary medical care. By using number of referrals to selectour sample from a defined primary care population we were faced withmaking a choice of how many referrals constituted ‘repeated referral’.The cut-off we chose of three or more new referrals over five years iden-tified 10% of all the adults aged less than 65 years of agewho had been re-ferred to a specialist over the five years and identified a group with amean of approximately five referrals.

Implications for policy, practice and research

There are three main implications of this study: Firstly, the criterionwe used in this study of at least three referrals for the diagnosis of symp-toms which in at least two cases were judged by the specialist to be MUS(which includes both syndromes, such as irritable bowel syndrome,fibro-myalgia or chronic pelvic pain, and symptoms, such as palpitations orpain with normal investigations) has utility in identifying patients whoare costly tomanage. The findings also indicate that the potentially exces-sive costs of care of patients withMUS is not limited to the extremely fre-quently referred or polysymptomatic group. Secondly, the repeatedreferral of these patients to amedical specialist is at least as costly as refer-ral of patients with symptoms of disease, and is associated with greaterexpenditure on investigations seeking disease. This tells us thatwe shouldnot regard referral for assessment ofMUS as cheap. Thirdly, given the costand likely lack of benefit of the patters of repeated referral for specialistassessment of patientswithMUS, [23,24] newandmore cost-effective ap-proaches to management should be explored. These have the potentialboth to save costs and to improve patient outcomes; a new approachthat costs less than the difference between those repeatedly referredwith MUS and infrequently referred patients (average of £4000 overfive years) would be potentially cost saving. New approaches might in-clude prompts embedded in clinical information systems to remind prac-titioners that the patient has already had repeated referrals resulting inspecialist diagnoses of MUS and better assessment and treatment of de-pression, anxiety and panic disorders [14]. The design and evaluation ofsuch new approachesmust be a priority if wewish to both reduce unnec-essary healthcare costs and improve the quality of care we currently pro-vide for the large number of patients who are currently repeatedlyreferred to secondary care with MUS.

Conclusions

The repeated referral of patients with MUS from primary care tomedical specialists generates high healthcare costs and is of dubiousbenefit to patients. There is consequently an urgent need to develop

and evaluate potentially more cost-effective methods of identifyingassessing and treating this important group of patients includingthe better assessment and treatment of depression and anxiety. Thisstudy quantifies the potential maximum savings that could be made.

Competing interests

All authors declare that they have no competing interests.

Acknowledgements

We wish to thank the staff of ISD for their collaboration, andthe primary care staff and patients who took part in this study. Weare grateful to Dr Andrew Walker for his advice on the design ofthe study. This work was supported by the Chief Scientist Office ofthe Scottish Government Health Directorate [CZH/4/37]; the funderhad no involvement in the conduct or reporting of the study.

Appendix A

Referrals to clinical specialties: proportion of referrals from eachpatient group.

Number of referrals

IRS RMES RMUS

221

1035 1351

% (N)

% (N) % (N)

Medicine

27.2 26.4 33.4 Cardiology 5.2 (11) 5.3 (55) 5.7 (77) Endocrinology 0.4 (1) 2.3 (24) 1.4 (19) Gastroenterology 10 (22) 5.7 (59) 7.1 (96) General Medicine 2.4 (5) 2.2 (23) 4.2 (57) Neurology 4.8 (11) 4 (41) 7.6 (103) Rehabilitation 0 0.1 (1) 0.3 (4) Respiratory 1.6 (4) 2.4 (25) 3.4 (46) Rheumatology 2.8 (6) 4.4 (46) 3.7 (50) Surgery 33.6 37.4 30 Surgery (General) 14.4 (32) 11 (116) 10.4 (141) Orthopaedics 14.4 (32) 21 (213) 14.9 (201) Urology 4.8 (11) 5.6 (58) 4.7 (63) Specialties 36.4 31.4 30.6 Ear, Nose & Throat 18.8 (42) 13 (129) 11 (149) Dermatology 2 (4) 4.2 (43) 1.9 (26) Ophthalmology 4 (9) 5.1 (53) 2.8 (38) Genito-UrinaryMedicine

0.4

(1) 0.4 (4) 0.1 (1)

Gynaecology

11.2 (25) 9.2 (95) 14.8 (200) Other 2.8 (6) 4.8 (50) 6 (81)

RRMUS: repeatedly referred with MUS; RRMES: repeatedly referred with medicallyexplained symptoms; IRS: infrequently referred with symptoms.

References

[1] Sharpe M. Medically unexplained symptoms and syndromes. Clin Med 2002;2(6):501–4.

[2] Burton C. Beyond somatisation: a review of the understanding and managementof medically unexplained physical symptoms (MUPS). Br J Gen Pract 2003;53:233–41.

[3] Hamilton J, Campos R, Creed F. Anxiety, depression and management of medicallyunexplained symptoms in medical clinics. J R Coll Physicians Lond 1996;30(1):18–20.

[4] Carson AJ, Ringbauer B, Stone J, McKenzie L, Warlow C, Sharpe M. Do medicallyunexplained symptoms matter? A prospective cohort study of 300 new referralsto neurology outpatient clinics. J Neurol Neurosurg Psychiatry 2000;68(2):207–10.

[5] Reid S, Wessely S, Crayford T, Hotopf M. MUS in frequent attenders of secondaryhealth care: retrospective cohort study. BMJ 2001;322(7289):767.

[6] Maiden NL, Hurst NP, Lochhead A, Carson AJ, Sharpe M. Medically unexplainedsymptoms in patients referred to a specialist rheumatology service: prevalenceand associations. Rheumatology (Oxford) 2003;42(1):108–12.

[7] Barsky AJ, Borus JF. Somatization and medicalization in the era of managed care.JAMA 1995;274(24):1931–4.

247C. Burton et al. / Journal of Psychosomatic Research 72 (2012) 242–247

[8] Williams ER, Guthrie E, Mackway-Jones K, James M, Tomenson B, Eastham J, et al.Psychiatric status, somatisation, and health care utilization of frequent attendersat the emergency department: a comparison with routine attenders. J PsychosomRes 2001;50(3):161–7.

[9] Reid S, Wessely S, Crayford T, Hotopf M. Frequent attenders with medically unex-plained symptoms: service use and costs in secondary care. Br J Psychiatry2002;180:248–53.

[10] Hansen MS, Fink P, Frydenberg M, Oxhoj ML. Use of health services, mental ill-ness, and self-rated disability and health in medical inpatients. Psychosom Med2002;64(4):668–75.

[11] Grabe H, Baumeister S, John U, Freyberger H, V+¦lzke H. Association of mentaldistress with health care utilization and costs: a 5-year observation in a generalpopulation. Soc Psychiatry Psychiatr Epidemiol 2009;44(10):835–44.

[12] Verhaak PF, Meijer SA, Visser AP, Wolters G. Persistent presentation of MUS ingeneral practice. Fam Pract 2006;23(4):414–20.

[13] McGorm K, Burton C, Weller D, Murray G, Sharpe M. Patients repeatedly referredto secondary care with symptoms unexplained by organic disease: prevalence,characteristics and referral pattern. Fam Pract 2010;27(5):479–86.

[14] Penzien DB, Rains JC. Reassuring patients about normal test results. BMJ2007;334(7589):325.

[15] Burton C, McGorm K, Weller D, Sharpe M. Depression and anxiety in patients re-peatedly referred to secondary care with medically unexplained symptoms: acase control study. Psychol Med 2011;41(3):555–63.

[16] Smith RC, Lyles JS, Gardiner JC, Sirbu C, Hodges A, Collins C, et al. Primary care cli-nicians treat patients with medically unexplained symptoms: a randomized con-trolled trial. J Gen Intern Med 2006;21(7):671–7.

[17] Smith BJ, McGorm KJ, Weller D, Burton C, Sharpe M. The identification in primarycare of patients who have been repeatedly referred to hospital for medicallyunexplained symptoms: a pilot study. J Psychosom Res 2009;67(3):207–11.

[18] Information Services Division. Scottish Health Service Costs. http://www.isdscotland.org/isd/3780.html2005 Accessed 21/2/2011.

[19] Department of Health. NHS Reference Costs 2004–2005. London: NSRC1; 2006.[20] Thompson SG, Barber JA. How should cost data in pragmatic randomised trials be

analysed? BMJ 2000;320(7243):1197–200.[21] Barber JA, Thompson SG. Analysis of cost data in randomized trials: an application

of the non-parametric bootstrap. Stat Med 2000;19(23):3219–36.[22] Morriss R, Dowrick C, Salmon P, Peters S, Dunn G, Rogers A, et al. Cluster rando-

mised controlled trial of training practices in reattribution for medically unex-plained symptoms. Br J Psychiatry 2007;191:536–42.

[23] Spiegel BM, Gralnek IM, Bolus R, Chang L, Dulai GS, Naliboff B, et al. Is a negativecolonoscopy associated with reassurance or improved health-related quality oflife in irritable bowel syndrome? Gastrointest Endosc 2005;62(6):892–9.

[24] McDonald IG, Daly J, Jelinek VM, Panetta F, Gutman JM. Opening Pandora's box: theunpredictability of reassurance by a normal test result. BMJ 1996;313(7053):329–32.