You are on page 1of 6

Journal of Psychosomatic Research 71 (2011) 142147

Contents lists available at ScienceDirect

Journal of Psychosomatic Research

Classication characteristics of the Patient Health Questionnaire-15 for screening somatoform disorders in a primary care setting
Stephanie Krber , Dirk Frieser 1, Natalie Steinbrecher, Wolfgang Hiller
University of Mainz, Department of Clinical Psychology, Germany

a r t i c l e

i n f o

a b s t r a c t
Background: This study examines how effectively the Patient Health Questionnaire-15 (PHQ-15), a selfadministered screening instrument, recognizes somatoform symptoms and somatoform disorders in a German primary care setting. Methods: A selected sample of 308 patients (mean age 47.2 years, 71.4% women) from two regular primary care practices was screened with the PHQ-15 and additionally examined with structured interviews. Their primary care physicians rated symptoms reported in the interview as either medically explained or medically unexplained. Results: Seventy-six percent of the symptoms were judged as medically unexplained. The PHQ-15 correlated signicantly with the total number of symptoms as well as the number of somatoform symptoms (both r=0.63; P.001). A comparison between the most frequently reported symptoms in the interview and the 15 items of the PHQ-15 revealed that even though the PHQ-15 does not differentiate between medically explained and medically unexplained symptoms, it does catch many somatoform symptoms. When used to predict the diagnosis of a somatoform disorder, a cutoff of 10 points in the PHQ-15 was identied as optimal, resulting in a sensitivity of 80.2% and specicity of 58.5%. However, the cutoff has to be adjusted according to specic research or clinical purposes. Conclusion: Several previous results could be conrmed, and under consideration of some limitations, the PHQ-15 seems to be a valuable tool for identifying somatoform symptoms and disorders in primary care. 2011 Elsevier Inc. All rights reserved.

Article history: Received 31 March 2010 Received in revised form 16 November 2010 Accepted 6 January 2011 Keywords: Screening Medically unexplained symptoms Somatoform disorders Primary care

Introduction Medically unexplained symptoms (MUSs) and somatoform disorders (SFDs) are a prevalent phenomenon in the primary care setting [1 4]. Although effective therapeutic strategies for treating this patient group exist [57], diagnostic instruments and screening methods are needed for a reliable case detection. The Patient Health Questionnaire15 (PHQ-15) was developed for this purpose [8]. It is an economical, self-administered screening instrument that has been used as a tool in a number of studies. Moreover, it has been suggested by the DSM-V Workgroup on Somatic Symptom Disorders as a measure of symptom severity for the current proposal of a revised SFD classication [9]. The general purpose of the present study is to further validate the PHQ-15 and to test its screening and classication characteristics. We attempt to replicate previous results and provide new data. The PHQ-15 is a screening instrument for somatic symptom severity (SSS) and has been developed from its precursors, the

Corresponding author. Present Address: University of Erlangen-Nrnberg, University Hospital Erlangen, Department of Psychosomatic Medicine and Psychotherapy, Schwabachanlage 6, 91054 Erlangen, Germany. Tel.: +49 (0) 9131/85 44855. E-mail address: (S. Krber). 1 Present address: salus klinik" Hospital, Friedrichsdorf, Germany. 0022-3999/$ see front matter 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.jpsychores.2011.01.006

Primary Care Evaluation of Mental Disorders (PRIME-MD) [10] and the PRIME-MD Patient Health Questionnaire (PRIME-MD PHQ) [11,12]. These methods aim at economically detecting mental disorders and have been used and validated in a variety of studies [8,1320]. Validity, classication characteristics and responsiveness to change have been reported by these studies. However, to the best of our knowledge, the analyses of classication characteristics were always focused on the PHQ-15s precursor, the PRIME-MD PHQ, but not on the PHQ-15 itself. One exception is a study by van Ravesteijn et al. [20]. However, after conducting a factor analysis, they eliminated two items and thus only studied 13 instead of 15 items. Therefore, in our study, we want to examine the PHQ-15s validity in its most commonly used complete 15-item version. As with every instrument currently used to screen MUSs and SFDs, the PHQ-15 does not differentiate between medically unexplained and medically explained symptoms. To judge this, a medical examination and physicians' opinions are needed. However, it is possible to assess whether a screening instrument is able to capture those symptoms that are most likely medically unexplained and that most often occur in patients with SFD. This is another aim of the current study. We also want to test the assumption of Kroenke et al. that the symptoms inquired about in the PHQ-15 are the most prevalent symptoms reported in the outpatient setting [8,21].

S. Krber et al. / Journal of Psychosomatic Research 71 (2011) 142147


In summary, based on these existing ndings, we aim at answering the following questions: (a) Is the PHQ-15 able to capture physical symptoms in general as well as somatoform symptoms needed to diagnose an SFD in particular? What symptoms are reported most often by patients, and is the PHQ-15 able to capture these symptoms? (b) Is the PHQ-15 a good predictor of the existence of an SFD according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) [22]? What is the optimal threshold to further consider the diagnosis of an SFD? To answer these questions, the present study focuses on the primary care setting because, for most patients with MUSs, their primary care physician is the rst point of contact. Method Procedure First step Screening and interviews took place between February and September 2008 within two primary care practices in Mainz, Germany. Both practices are part of the regular German health system with two general practitioners (GPs) working at each practice. In a rst step, 614 general medical outpatients were screened consecutively with the PHQ-15 questionnaire. We wanted to recruit at least 600 patients with the nal goal of sampling a sufciently high number of patients with a high risk of having an SFD (see below). All patients entering the two practices were asked by the medical assistants if they would complete the PHQ-15. Patients returning to the ofce for the second time within the 8 months of duration of the study completed the questionnaire only once. Those who agreed answered the questionnaire by themselves in the waiting room and then gave it back to the medical assistant. The GPs did not see the completed PHQ-15 questionnaires. Grouping Based on their answers, the 614 patients were then grouped according to Kroenke and colleagues [8] into four categories of SSS: minimal (PHQ-15 score: 04), low (59), medium (1014) and high (15 and above). Their distribution can be seen in Fig. 1. Second step As one of our research interests lies in patients with a high risk of having an SFD, our primary goal was to examine patients with higher levels of SSS. It was therefore, in a second step, our intention to further examine a selected sample of at least 25% of the patients with minimal SSS, at least 40% of the patients with low SSS, at least 70% of the patients with medium SSS and at least 80% of the patients with high SSS (Fig. 1), leading to an oversampling of patients with increased SSS. Finally, the

selected sample consisted of 308 patients: 39 patients with minimal SSS (27% of 145), 106 patients with low SSS (43% of 244), 118 patients with medium SSS (70% of 169) and 45 patients with high SSS (80% of 56). The patients were almost equally recruited from the two practices (51.9% vs. 48.1%). This subsample of 308 patients (50.2% of 614) was further examined in more detail using a diagnostic interview. Interviews were conducted by two trained clinical psychologists who were informed about the PHQ-15 score in advance. After the interview, the symptoms reported by patients were rated by the treating GP as either medically explained or (partly) medically unexplained (see also Assessment Instruments). The sampling procedure is summarized in Fig. 1. Sample The mean age of the 308 patients was 47.2 years (S.D.=16.3, range 1887 years). Two hundred twenty (71.4%) were women. A total of 32.6% lived alone, but most lived with either one (37.5%), two (14.3%), or more (15.6%) other persons in their households. A total of 58.5% had a school education of 12 years or more, a percentage that is about twice as high as in the German general population [23] and thus reects a more academic patient population at both primary care practices. The selected sample of 308 patients did not differ signicantly from the nonselected sample of 306 patients with respect to age, education and number of persons in the households. However, there were signicantly more women in the selected sample (71.4% vs. 62.7%, =5.24, P.05). Assessment instruments PHQ-15 The PHQ-15 is a questionnaire which inquires about 15 somatic symptoms or symptom clusters by asking the patient to rate their severity during the previous 4 weeks on a 3-point scale as either 0 (not bothered at all/not at all), 1 (bothered a little/several days) or 2 (bothered a lot/more than half the days). The symptoms are stomach pain, back pain, headache, chest pain, dizziness, fainting, palpitations, shortness of breath, bowel complaints (constipation or diarrhea), dyspeptic complaints (nausea, gas or indigestion), fatigue, trouble sleeping, pain in joints or limbs, menstrual pain or problems and pain or problems during sexual intercourse. The total score ranges from 0 to 30. According to Kroenke et al. [8,21], these symptoms are the most prevalent symptoms reported in the outpatient setting. Diagnostic interview Our diagnostic interview was based on an interview previously used in a study by Rief and colleagues [24]. It is a highly structured

Minimal PHQ-15=0-4 N=145 24% of 614 Intended selection quota: Real selection quota: 25% of 145 27% of 145 Minimal PHQ-15=0-4 N=39 13% of 308

Consecutive sample (N=614) Low Medium PHQ-15=5-9 PHQ-15=10-14 N=244 N=169 40% of 614 28% of 614 40% of 244 43% of 244 70% of 169 70% of 169

High PHQ-1515 N=56 9% of 614 80% of 56 80% of 56 High PHQ-1515 N=45 15% of 308

Low Medium PHQ-15=5-9 PHQ-15=10-14 N=106 N=118 34% of 308 38% of 308 Selected sample (N=308)

Fig. 1. Sampling procedure.


S. Krber et al. / Journal of Psychosomatic Research 71 (2011) 142147 (7.81 symptoms per patient; 76.0% somatoform, 20.1% somatic). No patient fullled the criteria of a somatization disorder or a conversion disorder according to DSM-IV during the last 12 months (Table 1). A total of 15.9% of the 308 patients fullled the criteria for a pain disorder (adjusted for N=614: 12.0%), and 13.6% for an undifferentiated SFD (adjusted for N=614: 10.9%). Moreover, we inquired about depressive and anxiety disorders during the last 12 months. The prevalence of major depressive disorder (single episode and recurrent) for the 308 patients was 14.9% (adjusted for N=614: 11.3%), and the prevalence of any anxiety disorder was 13.3% (adjusted for N=614: 11.4%).

interview that includes common somatic symptoms (including all symptoms relevant for somatization disorder, irritable bowel syndrome, bromyalgia, neurasthenia, chronic fatigue syndrome) and psychological symptoms (e.g., irritability) prevalent over the last 12 months. These symptoms were then rated by their treating GP as either medically explained or (partly) medically unexplained according to their expertise. The symptoms reported by the patients and the GP's rating as to whether each symptom is an MUS or not were used to determine the diagnosis of an SFD. The operationalization of an SFD can be seen in Table 1. In addition, current and lifetime mental disorders were assessed using the International Diagnostic Checklists (IDCL) [25], an instrument designed to systematically evaluate the DSM-IV criteria of common mental disorders. The reliability and validity of the IDCL are well established [25]. Statistical analysis According to the sampling procedure (Fig. 1), the prevalence rates were adjusted to the original number of N=614 patients by weighting them by the inverse of their probability of selection. For all statistical analyses, the statistical package SPSS 17.0 was used. To assess the relationship between the PHQ-15 and the number of symptoms, Pearson's productmoment coefcients (one-tailed) were calculated. The existence of a depressive or anxiety disorder was controlled for using partial correlations. To test the accuracy of the PHQ-15 and to identify the optimal cutoff to predict the existence of an SFD, the receiver operating characteristics (ROC) sensitivity, specicity and positive diagnostic likelihood ratios (DLRs) as well as an ROC curve were calculated. Positive and negative predictive values were not calculated because of their dependence on prevalence rates [26], which were altered because of the selected sample. Phi coefcients, which express the association between two binary variables, were used to determine the relation between the existence of an SFD and several PHQ-15 cutoffs. Ethical approval The study protocol was approved by the ethics committee of the German Psychological Society, and all patients gave their written informed consent.
Results Prevalence rates The 308 patients achieved a mean PHQ-15 score of 9.82 (S.D.=5.09, range 028) and reported a total of 2838 symptoms in the subsequent interview (9.21 symptoms per patient, S.D.=6.21, range 037). According to the physicians' ratings, 2218 of these symptoms (78.2%) were labeled as somatoform (7.20 symptoms per patient, S.D.= 5.98, range 035) and 516 (18.2%) as somatic. One hundred four symptoms (3.7%) could not be rated due to lack of information. Adjustment to the original number of 614 patients leads to a mean PHQ-15 score of 8.15 and a total of 4797 reported symptoms Table 1 Operationalization and prevalence of SFD diagnoses

PHQ-15 as predictor of the number of reported symptoms The correlation between the PHQ-15 and the total number of reported symptoms in the interview was 0.63 (P.001). The correlation was somewhat smaller when we controlled for the presence of major depressive disorder (partial correlation, r=0.57, P.001) or any anxiety disorder (partial correlation, r=0.61, P.001). The correlation between the PHQ-15 and the number of MUS was also 0.63 (P.001). Again, the values turned out to be slightly smaller when controlling for major depressive disorder (partial correlation, r=0.56, P.001) or any anxiety disorder (partial correlation, r=0.60, P.001). As a next step, we evaluated if the symptom count and the frequency distribution of the symptoms are different when either all symptoms or only MUS are considered. In addition, as Kroenke and colleagues [8,21] previously stated that the PHQ-15 covers the 15 most prevalent symptoms found in the outpatient setting, we tried to assess whether it also covers the 15 most prevalent symptoms of our study. We therefore counted the reported number of symptoms with at least a moderate degree of impairment and compared them with the PHQ-15 items. Moreover, we compared the ranking according to three groups of symptoms/patients: all symptoms reported by all patients, all symptoms reported by patients with an SFD and all MUSs reported by all patients. Table 2 demonstrates that the PHQ-15 does indeed capture at least 10 of the most frequently reported symptoms. Moreover, the rankings in the three groups are quite similar, indicating a similar frequency distribution when either all symptoms or only MUS are considered. With regard to the PHQ-15-items, fatigue, dyspeptic complaints, trouble sleeping and several pain items were found to belong to the most frequently reported symptoms, independent of the group considered. In addition, there were some symptoms not investigated in the PHQ-15 but often reported in the interview (Table 2). These were difculty to relax, problems concentrating, muscular pain, irritability and sweating. Three of these symptoms are of psychological nature rather than physical symptoms and therefore not intended to be captured by the PHQ-15. It is possible that muscular pain might be covered by other items such as back pain or pain in joints or limbs. Sweating seems to be an item that should be considered further, whereas the items shortness of breath, palpitations and especially the items pain or problems during sexual intercourse and fainting appear negligible. It is also evident that the items pain in joints or limbs and menstrual pain or problems drop by 6 and 4 positions, respectively, if only the MUSs are considered. This suggests that these symptoms are more often judged as somatic instead of somatoform compared to symptoms like headache, back pain or stomach pain.

PHQ-15 as predictor for SFDs (DSM-IV) Another question of interest was whether the PHQ-15 predicts an SFD diagnosis. We used a dichotomous variable SFD yes or no which identied patients with undifferentiated SFD or pain disorder (N=91; see above, no patient had somatization disorder or conversion disorder). The diagnosis variable had a point biserial correlation of 0.41 (P.001) with the PHQ-15. Furthermore, we analyzed the accuracy of the PHQ-15 in detecting patients with SFD and evaluated the optimal threshold of the PHQ-15 to distinguish patients with and without SFD. An ROC curve is displayed in Fig. 2. Sensitivity and specicity, positive DLRs as well as phi correlations (between SFD yesno and PHQ-15 above cutoff yes

Minimum requirement for any SFD: 1 very strongly impairing MUS, duration 12 months Somatization disorder 1 very strongly impairing MUS Pain disorder 1 very strongly impairing medically unexplained pain symptom Conversion disorder Undifferentiated SFD

8 MUSs from different body sites according Impairment caused by pain symptoms to DSM-IV, impairment at least moderate outweighs impairment caused by other symptoms 1 MUS began before the age of 30 Duration 12 months Duration 12 months Prevalence in %a 0%

1 very strongly impairing medically unexplained 1 very pseudoneurological symptom strongly impairing MUS Duration 12 Impairment caused by pseudoneurological months symptoms outweighs impairment caused by other symptoms Duration 12 months




Prevalence is adjusted to N=614.

S. Krber et al. / Journal of Psychosomatic Research 71 (2011) 142147 Table 2 Most frequently reported symptoms Rank All patients (N=308) PHQ-15 item (no. of interview categories summarized under this item)a Fatigue (4) Dyspeptic complaints (nausea, gas or indigestion) (6) Back pain (1) Pain in joints or limbs (2) Trouble sleeping (1) Headache (1) Chest pain (2) Stomach pain (1) Bowel complaints (constipation or diarrhea) (2) Menstrual pain or problems (3) Dizziness (1) Shortness of breath (1) Palpitations (1) Pain or problems during sexual intercourse (2) Fainting (1) Further items reported in the interview but not included in the PHQ-15 Difculty to relax Problems concentrating Muscular pain Irritability Sweating 10 14 11 15 17 16 12 15 14 13 7 12 13 14 15 All symptoms (MUS and MES) 1 2 4 3 5 6 7 8 9 12 13 16 19 29 44 Patients with SFD (N=91) All symptoms (MUS and MES) 1 2 3 4 5 6 10 9 7 8 11 18 20 32 49 All patients (N=308)


Only MUS 1 2 4 9 3 5 6 8 10 16 11 18 16 25 46 Fig. 2. Receiver operating characteristic curve for PHQ-15 scores.

a To make a comparison possible, it was necessary to summarize some of the symptoms reported in the interview.

no) are summarized in Table 3. With respect to phi correlations, the coefcients can be interpreted similar to effect sizes (small effect=0.10, medium=0.30, large=0.50) [27]. The area under the curve (AUC) is 0.76 (P.001; CI=0.710.82), indicating that the PHQ-15 has a signicant accuracy and discriminatory power. The data in Table 3 can be used to choose an optimal threshold. The cutoff of 10 (sensitivity 80.2%, specicity 58.5%) leads to the highest rate of correct classications and thus to the lowest rate of false classications (19.8% false negatives, 41.5% false positives). However, this cutoff has the disadvantage that the false-positive rate is quite high. The cutoff of 10 leads to a positive DLR of 1.93, indicating that a PHQ-15 score of 10 or more is almost twice as likely to occur in patients with SFD than in patients without SFD. The phi coefcient is highest when the cutoff of 12 is chosen (phi=0.36; P.001), but almost identical at the cutoff of 10. Therefore, we recommend a cutoff of 10 as optimal. This is further supported by the fact that raising the cutoff from 10 to 11 leads to a loss of sensitivity of 10 percentage points.

unexplained), the analysis of the 15 most frequently reported symptoms revealed a similar frequency distribution when either all symptoms or only MUSs are considered. This indicates that the PHQ15 does indeed capture many typical somatoform symptoms. It can therefore be assumed that there is high concordance between the prediction of somatoform symptoms and physical symptoms in general. This is in line with other studies [19,31] that demonstrated a high association between patients' self-reported symptom counts and clinician-rated somatoform complaints. The fact that the comparison of all symptoms with only somatoform symptoms (Table 2) leads to a similar frequency distribution, as well as the high proportion of 76.0% of MUSs, questions the differentiation between somatic and somatoform. Moreover, previous studies have shown that the presence of somatic symptoms, irrespective of their aetiology, is associated with increased social and psychiatric morbidity [32]. This might also strengthen the argument of some researchers who in the origination process of DSM-V favour the abolition of the somatoformsomatic differentiation and who see the existence of physical symptoms (and their perception, appraisal,...) as more important than their categorization in somatoform or somatic [33,34]. The suggestion of Kroenke et al. that the PHQ-15 symptoms are the most prevalent symptoms reported in the outpatient setting [8,21] could only partly be conrmed in our analysis. The PHQ-15 does indeed capture 10 of the most frequently reported symptoms. However, every third symptom could not be conrmed: there are symptoms reported by Kroenke but not conrmed in our study (e.g.,

Discussion The main intention of the present study was to assess the psychometric properties of the PHQ-15 as a screening instrument for MUS and SFD in the primary care setting. The nding of a proportion of 76.0% of somatoform symptoms in the total symptom count is similar to the results of previous studies. For example, Aiarzaguena and colleagues [28] reported a ratio between somatic and somatoform symptoms of 1 to 4 and Kroenke and Mangelsdorff [29] a ratio of 1 to 3. These results conrm the high relevance that MUSs have in the primary care setting. However, it has to be considered that not every medically unexplained symptom is bothersome. Our prevalence rate of SFDs (22.9%) lies within the wide range of prevalence rates reported in other primary care studies, too. For example, de Waal et al. reported a rate of 16.1%, Mergl et al. a rate of 25.6% and Fink et al. even a rate of 57.5% [1,3,30]. Although the PHQ-15, as with any self-rating instrument, assesses symptoms irrespective of their nature (medically explained vs.
Table 3 Sensitivity, specicity, positive DLRs and phi coefcients for PHQ-15 scores PHQ-15 4 5 6 7 8 9 10 11 12 13 14 15 Sensitivity 100.0% 100.0% 96.7% 93.4% 87.9% 84.6% 80.2% 70.3% 61.5% 48.4% 39.6% 31.9% Specicity 13.4% 18.0% 28.1% 40.6% 47.5% 53.0% 58.5% 65.9% 76.0% 82.5% 88.5% 92.6% Pos. DLR 1.15 1.22 1.35 1.57 1.67 1.80 1.93 2.06 2.57 2.76 3.43 4.32 Phi coefcienta 0.21 0.25 0.28 0.34 0.33 0.35 0.35 0.33 0.36 0.32 0.32 0.32

Numbers in bold print represent the most optimal value (phi coefcient) or value combination (sensitivity and specicity), respectively. a Interpretation: small effect=0.10, medium effect=0.30, large effect=0.50 [27]. P.001.


S. Krber et al. / Journal of Psychosomatic Research 71 (2011) 142147

fainting), as well as symptoms that showed to be relevant in our analysis but are not included in the PHQ-15 (e.g., sweating). The signicant correlations between the PHQ-15 score and the symptom counts conrm that a high PHQ-15 score accompanies a high symptom burden. With respect to manifest disorders, the concordance between the PHQ-15 score and the presence of an SFD can be judged as moderately satisfying. It again shows that the PHQ15, even though it cannot differentiate between medically explained and medically unexplained symptoms, does have the power to predict, to a certain degree, the diagnosis of an SFD. The score of the AUC is signicantly different from 0.5 (which would correspond to an accuracy no better than chance). However, compared to AUC scores reported in other studies with respect to other questionnaires, it is only mediocre. For example, several depression screeners or the General Health Questionnaire (GHQ) show AUC scores between 0.88 and 0.92 [35,36]. However, in their analysis of the PHQ-15, van Ravesteijn and colleagues reported an AUC score of 0.76 as well [20]. De Waal and colleagues analyzed another physical symptom screening questionnaire, the Physical Symptom Checklist-51 (PSC51), which resulted in an AUC score of 0.80 [37]. Similarly, our sensitivity and specicity scores are somewhat lower than reported in many other studies, and in our study, it is especially noticeable that sensitivity goes quite much to the expense of specicity and vice versa. For example, when evaluating their symptom checklist, Kroenke et al. reported a specicity of 77% at a sensitivity score of 85% [38]; in our study, a sensitivity score of 85% is accompanied by a much lower specicity of 53%. For the PHQ-15, van Ravesteijn et al. reported higher scores as well (sensitivity 78%, specicity 71%) [20]. However, the results of de Waal et al. are similar to ours again, as in their analysis of the PSC-51, they reported a specicity of 64% at a sensitivity score of 78% [37]. One possible alternative explanation for the low ROC scores might be the operationalization of SFDs (Table 1). Maybe these diagnoses do not represent a sufcient gold standard. However, we consider this as unlikely, as except from the duration criterion of at least 12 months we dened the SFDs as stringently as possible according to the DSM-IV criteria. In this context, another explanation might be the fact that other studies used different criterion standards. Van Ravesteijn and colleagues [20], as an example, used the Structured Clinical Interview for DSM-IV Axis I disorders as reference. Although both their and our proceedings are based on the same classication system, i.e., DSM-IV, different ways of translating its criteria into concise questions might lead to slightly different results. Moreover, the task of evaluating symptoms as medically explained or medically unexplained is a special challenge in itself. Apart from physicians' abilities to reliably assess this distinction (a point that is going to be discussed below), it has already been hinted at the general reliability of this discrimination. Because of the difculties in deciding which physical phenomenon is medically caused and which one not, and because of concerns towards the fundamental sense of this dichotomy, a wide debate has emerged on this topic, and the current trend more and more seems to favour the abolition of this discrimination [33,34,39]. This trend is also mirrored in the current proposal of the DSM-V Workgroup on Somatic Symptom Disorders [9] that explicitly abdicates the distinction between somatic vs. somatoform complaints. Our analysis led to an optimal cutoff of PHQ-15 10. This cutoff is higher than the one chosen by van Ravesteijn et al. [20], who found a cutoff of 6 to be optimal. However, in their analysis, they had excluded 2 of the 15 symptoms. Our recommendation of a PHQ-15 score 10 is, of course, debatable. We want to stress that researchers and clinicians need to decide which cutoff suits best for their purposes. This decision should take into account possible benets (e.g., early detection and treatment of vulnerable patients) and costs (e.g., time, costs of overlooking diseased patients) [40]. The main benet of a cutoff of 10 is that a high proportion of patients with an SFD will be detected and

eventually treated. However, there is the cost of a high proportion of false positives (41.5%), i.e., patients who are falsely screened as having an SFD, indicating an overinclusive cutoff. This may lead to a large amount of extra time for diagnostic evaluation which is usually scarce in the primary care setting [41]. What do these results mean with respect to the current revision process of the SFD category in DSM-V? Firstly, we interpret our prevalence rates of MUS and SFD as an indicator of their value as important and independent diagnostic entity that should have its place in the new classication systems. This is not self-evident, as some authors question the independent value of the SFD category and ask for its abolition (see for example van der Feltz-Cornelis and van Balkom [42]). Secondly, the suggestion of the DSM-V Workgroup on Somatic Symptom Disorders to include the PHQ-15 as a measure of symptom severity for the current proposal of complex somatic symptom disorder (CSSD) [9] can be reinforced. In this context, a third point of argument is that the use of dimensional measures as the PHQ-15 could reect different degrees of severity and thus build a bridge between the sometimes too strict (as in somatization disorder) and sometimes too loose (as in undifferentiated SFD) classication criteria. Finally, the frequently criticized distinction between medically explained and medically unexplained symptoms that was already mentioned above seems to be gradually overcome, as both the current proposal of CSSD and the PHQ-15 do no more differ between somatoform and somatic symptoms. Several limitations concerning our procedure and analyses have to be mentioned. Concerning the sampling procedure, a self-selection bias cannot be excluded. The participation in our study was voluntary, and features of the sample (e.g., with regard to age, sex or education) might have inuenced the individuals' decisions to participate or not. Moreover, our sampling procedure had consequences for the assessment of ROCs, as the calculation of predictive values was not possible, which would have been a valuable additional indicator. A third limitation is that it can be argued that the use of physicians' ratings as the only source for determining the aetiology of a symptom is not optimal. Further data, such as those obtained through chart reviews [43], could be added. In addition, there were no exactly dened criteria about which medical examinations, laboratory tests etc. had to be conducted to conrm or rule out a somatic cause, but the GPs decided according to their own expertise. Finally, physicians' ratings were not assessed for interrater reliability, as each physician only rated his or her own patients. However, as Kroenke and colleagues [38] stated, the primary care physician's gestalt about a symptom being medically unexplained is quite good and few patients with symptoms initially judged to be somatoform were later found to have occult serious physical disorders at follow-up (p. 271) [38,44]. Moreover, the single rating of the treating primary care physician is the normal case in reality and thus represents ecological validity. A further limitation of our study is that the psychologists who conducted the interviews were not blind to the patients' PHQ-15 scores; thus, an unwanted inuence on their behaviour towards the patients cannot be excluded. Finally, when interpreting the receiver operant characteristics of the PHQ-15, it has to be borne in mind that sensitivity and specicity scores were not controlled for comorbid depressive or anxiety disorders, which might inuence the accuracy. However, this again reects ecological validity, as in practice comorbid diagnoses occur quite frequently in patients with SFD. In conclusion, our analyses result in a mixed picture. As described above, there are some discrepancies between our results and recent studies, and some limitations remain unsolved. However, the present study further strengthens the power of the PHQ-15 as an economical screening instrument in the primary care setting. It seems to be a valid measure of physical symptoms and has a signicant accuracy, and the application of a cutoff score of 10 can help detect patients with an SFD.

S. Krber et al. / Journal of Psychosomatic Research 71 (2011) 142147


[1] De Waal MWM, Arnold IA, Eekhof JAH, van Hemert AM. Somatoform disorders in general practice: prevalence, functional impairment and comorbidity with anxiety and depressive disorders. Br J Psychiatry 2004;184:4706. [2] Arnold IA, de Waal MWM, Eekhof JAH, van Hemert AM. Somatoform disorder in primary care: course and the need for cognitivebehavioral treatment. Psychosomatics 2006;47:498503. [3] Fink P, Srensen L, Engberg M, Holm M, Munk-Jrgensen P. Somatization in primary care: prevalence, health care utilization, and general practitioner recognition. Psychosomatics 1999;40:3308. [4] Hiller W, Fichter MM. High utilizers of medical care a crucial subgroup among somatizing patients. J Psychosom Res 2004;56:43743. [5] Rief W, Martin A, Rauh E, Zech T, Bender A. Evaluation of general practitioners' training: how to manage patients with unexplained physical symptoms. Psychosomatics 2006;47:30411. [6] van der Feltz-Cornelis CM, van Oppen P, Adr HJ, van Dyck R. Randomised controlled trial of a collaborative care model with psychiatric consultation for persistent medically unexplained symptoms in general practice. Psychother Psychosom 2006;75:2829. [7] Bleichhardt G, Timmer B, Rief W. Cognitivebehavioural therapy for patients with multiple somatoform symptoms a randomised controlled trial in tertiary care. J Psychosom Res 2004;56:44954. [8] Kroenke K, Spitzer RL, Williams JBW. The PHQ-15: validity of a new measure for evaluating the severity of somatic symptoms. Psychosom Med 2002;64: 25866. [9] American Psychiatric Association. DSM-5 development complex somatic symptom disorder [Internet]. Arlington: American Psychiatric Association. [updated 2010 Oct 25; cited 2010 Nov 4]. Available from ProposedRevisions/Pages/proposedrevision.aspx?rid=368. [10] Spitzer RL, Williams JBW, Kroenke K, Linzer M, deGruy III FV, Hahn SR, et al. Utility of a new procedure for diagnosing mental disorders in primary care: the PRIMEMD 1000 Study. JAMA 1994;272:174956. [11] Spitzer RL, Kroenke K, Williams JBW. Validation and utility of a self-report version of PRIME-MD: the PHQ Primary Care Study. JAMA 1999;282:173744. [12] Spitzer RL, Williams JBW, Kroenke K, Hornyak R, McMurray J. Validity and utility of the PRIME-MD Patient Health Questionnaire in assessment of 3000 obstetric gynecologic patients: the PRIME-MD Patient Health Questionnaire Obstetrics Gynecology Study. Am J Obstet Gynecol 2000;183:75969. [13] Diez-Quevedo C, Rangil T, Sanchez-Planell L, Kroenke K, Spitzer RL. Validation and utility of the Patient Health Questionnaire in diagnosing mental disorders in 1003 general hospital Spanish inpatients. Psychosom Med 2001;63:67986. [14] Escobar JI, Gara MA, Daz-Martnez AM, Interian A, Warman M, Allen LA, et al. Effectiveness of a time-limited cognitive behavior therapy-type intervention among primary care patients with medically unexplained symptoms. Ann Fam Med 2007;5:32835. [15] Gerber MR, Wittenberg E, Ganz ML, Williams CM, McCloskey LA. Intimate partner violence exposure and change in women's physical symptoms over time. J Gen Intern Med 2008;23:649. [16] Kroenke K, Messina N, Benattia I, Graepel J, Musgnung J. Venlafaxine extended release in the short-term treatment of depressed and anxious primary care patients with multisomatoform disorder. J Clin Psychiatry 2006;67:7280. [17] Becker S, Al Zaid K, Al Faris E. Screening for somatization and depression in Saudi Arabia: a validation study of the PHQ in primary care. Int J Psychiatry Med 2002;32:27183. [18] Grfe K, Zipfel S, Herzog W, Lwe B. Screening psychischer Strungen mit dem qGesundheitsfragebogen fr Patienten (PHQ-D)q: Ergebnisse der deutschen Validierungsstudie. Diagnostica 2004;50:17181. [19] Interian A, Allen LA, Gara MA, Escobar JI, Daz-Martnez AM. Somatic complaints in primary care: further examining the validity of the Patient Health Questionnaire (PHQ-15). Psychosomatics 2006;47:3928. [20] van Ravesteijn H, Wittkampf K, Lucassen P, van de Lisdonk E, van den Hoogen H, van Weert H, et al. Detecting somatoform disorders in primary care with the PHQ-15. Ann Fam Med 2009;7:2328.

[21] Kroenke K, Arrington MW, Mangelsdorff AD. The prevalence of symptoms in medical outpatients and the adequacy of therapy. Arch Intern Med 1990;150:16859. [22] American Psychiatric Association. Diagnostic and statistical manual of mental disorders DSM-IV-TR. 4th edition. Text Revision). Washington, DC: American Psychiatric Association; 2000. [23] Statistisches Bundesamt. Bildungsstand der Bevlkerung Ausgabe 2009 [Educational background in the population 2009] [Internet]. Wiesbaden: Statistisches Bundesamt; [cited 2010 Feb 15]. Available from: bpm.html.cms.cBroker.cls?cmspath=struktur,vollanzeige.csp&ID=1024383. [24] Rief W, Mewes R, Martin A, Glaesmer H, Brhler E. Are psychological features useful in classifying patients with somatic symptoms? Psychosom Med 2010;72: 64855. [25] Hiller W, Zaudig M, Mombour W. Development of diagnostic checklists for use in routine clinical care. Arch Gen Psychiatry 1990;47:7824. [26] Florkowski CM. Sensitivity, specicity, receiver-operating characteristic (ROC) curves and likelihood ratios: communicating the performance of diagnostic tests. Clin Biochem Rev 2008;29(Suppl (i)):S837. [27] Bhner M, Ziegler M. Statistik fr Psychologen und Sozialwissenschaftler. Mnchen: Pearson Studium; 2009. [28] Aiarzaguena JM, Grandes G, Salazar A, Gaminde I, Snchez . The diagnostic challenges presented by patients with medically unexplained symptoms in general practice. Scand J Prim Health Care 2008;26:99105. [29] Kroenke K, Mangelsdorff AD. Common symptoms in ambulatory care: incidence, evaluation, therapy, and outcome. Am J Med 1989;86:2626. [30] Mergl R, Seidscheck I, Allgaier A, Mller H, Hegerl U, Henkel V. Depressive, anxiety, and somatoform disorders in primary care: prevalence and recognition. Depress Anxiety 2007;24:18595. [31] Rost KM, Dickinson WP, Dickinson LM, Smith RC. Multisomatoform disorder: agreement between patient and physician report of criterion symptom explanation. CNS Spectr 2006;11:3838. [32] Kisely S, Goldberg D, Simon G. A comparison between somatic symptoms with and without clear organic cause: results of an international study. Psychol Med 1997;27:10119. [33] Mayou R, Kirmayer LJ, Simon G, Kroenke K, Sharpe M. Somatoform disorders: time for a new approach in DSM-V. Am J Psychiatry 2005;162:84755. [34] Sharpe M, Mayou R, Walker J. Bodily symptoms: new approaches to classication. J Psychosom Res 2006;60:3536. [35] Lwe B, Grfe K, Zipfel S, Witte S, Loerch B, Herzog W. Diagnosing ICD-10 depressive episodes: superior criterion validity of the patient health questionnaire. Psychother Psychosom 2004;73:38690. [36] Goldberg DP, Gater R, Sartorius N, Ustun TB, Piccinelli M, Gureje O, et al. The validity of two versions of the GHQ in the WHO study of mental illness in general health care. Psychol Med 1997;27:1917. [37] de Waal MWM, Arnold IA, Spinhoven P, Eekhof JAH, Assendelft WJJ, van Hemert AM. The role of comorbidity in the detection of psychiatric disorders with checklists for mental and physical symptoms in primary care. Soc Psychiatry Psychiatr Epidemiol 2009;44:7885. [38] Kroenke K, Spitzer RL, deGruy FV, Swindle R. A symptom checklist to screen for somatoform disorders in primary care. Psychosomatics 1998;39:26372. [39] Kroenke K. Somatoform disorders and recent diagnostic controversies. Psychiatr Clin North Am 2007;30:593619. [40] Swets JA, Dawes RM, Monahan J. Psychological science can improve diagnostic decisions. Psychological Science in the Public Interest 2000;1:126. [41] Cherry DK, Woodwell DA, Rechtsteiner EA. National Ambulatory Medical Care Survey: 2005 summary. Advance Data From Vital and Health Statistics 2007;387: 140. [42] Van der Feltz-Cornelis CM, van Balkom AJLM. The concept of comorbidity in somatoform disorder a DSM-V alternative for the DSM-IV classication of Somatoform disorder. J Psychosom Res 2010;68:7999. [43] Smith RC, Korban E, Kanj M, Haddad R, Lyles JS, Lein C, et al. A method for rating charts to identify and classify patients with medically unexplained symptoms. Psychother Psychosom 2004;73:3642. [44] Martina B, Bucheli B, Stotz M, Battegay E, Gyr N. First clinical judgment by primary care physicians distinguishes well between nonorganic and organic causes of abdominal or chest pain. J Gen Intern Med 1997;12:45965.