Point of View

Misuse of the Mood Disorders Questionnaireas a case-finding measure and a critique of theconcept of using a screening scale for bipolardisorder in psychiatric practice

Introduction

Bipolar disorder (BD) is a serious illness resultingin significant psychosocial morbidity and excessmortality. Recent research has suggested that BD,when defined to include milder variants, such asbipolar II disorder and BD not otherwise specified,is more prevalent than has been previouslyreported (1–11). Studies of depressed psychiatricand primary care patients show that the disorder isunder-diagnosed (5, 6, 10). Even for patientseventually diagnosed with BD, the lag betweeninitial treatment seeking and a correct diagnosis is

often more than ten years (12, 13). Failure torecognize BD in depressed patients incurs serioustreatment and clinical costs, including the under-prescription of mood-stabilizing medications, anincreased risk of rapid cycling due to possible over-prescription of antidepressant medications, andincreased costs of care (5, 14–16).BD is prone to being overlooked, in part,

because its diagnosis is more often based onretrospective report than cross-sectional assess-ment. When symptomatic, patients with BD aremuch more likely to experience symptoms ofdepression and anxiety than symptoms of mania

Zimmerman M. Misuse of the Mood Disorders Questionnaire as acase-finding measure and a critique of the concept of using a screeningscale for bipolar disorder in psychiatric practice.Bipolar Disord 2012: 14: 127–134. � 2012 The Author.Journal compilation � 2012 John Wiley & Sons A ⁄S.

Objectives: Under-recognition of bipolar disorder (BD) is common andincurs significant costs for individuals and society. Clinicians are oftenencouraged to use screening instruments to help them identify patientswith the disorder. The Mood Disorder Questionnaire (MDQ) is the mostwidely studied measure for this purpose. Some studies, however, haveused the MDQ as a case-finding instrument rather than a screening scale.Such inappropriate use of screening scales risks distorting perceptionsabout many facets of BD, from its prevalence to its consequences.

Methods: Studies using the MDQ were reviewed to identify thosereports that have used the scale as a case-finding measure rather than ascreening scale.

Results: Multiple studies were identified in the BD literature that usedthe MDQ as a diagnostic proxy. The findings of these studies weremisinterpreted because of the failure to make the distinction betweenscreening and case-finding.

Conclusions: Inappropriate conclusions have been drawn regarding theprevalence, morbidity, and diagnostic under-recognition of BD in studiesthat rely on the MDQ as a diagnostic proxy. A conceptual critique isoffered against the use of self-administered screening questionnaires forthe detection of BD in psychiatric settings.

Mark Zimmerman

Department of Psychiatry and Human Behavior,

Brown Medical School, Providence, RI, USA

doi: 10.1111/j.1399-5618.2012.00994.x

Key words: bipolar disorder – case-finding –

Mood Disorders Questionnaire – screening

Received 3 May 2011, revised and accepted for

publication 1 December 2011

Corresponding author:

Mark Zimmerman, M.D.

Bayside Medical Building

235 Plain Street

Providence, RI 02905

USA

Fax: 401-444-7057

E-mail: mzimmerman@lifespan.org

Bipolar Disorders 2012: 14: 127–134 � 2012 John Wiley and Sons A/S

BIPOLAR DISORDERS

127

or hypomania (17, 18). Because patients with BDoften do not present in the manic or hypomanicphase of the illness, the presence of these episodesrequires a history of these symptoms to be elicited,which may be difficult when patients presentduring a period of depression. Recommendationsfor improving the detection of BD include carefulclinical evaluations inquiring about a history ofmania and hypomania and the use of screeningquestionnaires (2, 8, 9, 19, 20).Several self-report screening questionnaires have

been developed to improve the detection of BD(21–24), the most widely studied being the MoodDisorders Questionnaire (MDQ) (22, 25). The firstpaper on the validity of the MDQ was published adecade ago, and since then, it has been translatedinto multiple languages, and two dozen studieshave examined its performance as a screening tool(26). Although the MDQ was intended as ascreening instrument, several studies describedbelow have used the scale as a diagnostic proxyand subsequently drew conclusions of questionablevalidity about the prevalence, identification, anddiagnosis of BD. These studies have been pub-lished in high-profile peer-reviewed journals andhave been subsequently cited in other articlesabout the widespread frequency and clinical sig-nificance of unrecognized BD. Moreover, thefrequency of these studies seems to be increasing,with several having been published in the past year(27–32).The purpose of the present article is twofold.

The first goal is to draw attention to the accumu-lating number of studies that have drawn inappro-priate conclusions based on the use of the MDQ asa case-finding measure. The conclusions of thesestudies are of doubtful validity because the oper-ating characteristics of the MDQ are not goodenough for it to function as a case-finding tool. Thesecond goal is to challenge the conceptual basis ofrecommendations to use self-administered ques-tionnaires to screen for BD in clinical psychiatricpractice.

A brief review of the statistics of diagnostic testing

In studies of the performance of self-report screen-ing scales, four statistics are most commonlycomputed: sensitivity, specificity, positive predic-tive value, and negative predictive value. Thesestatistics are easy to compute when a table ofassociation between the test and the gold standardis presented, as in Figure 1.The sensitivity (true positive rate) of a screening

test refers to how well the test identifies individualswith the illness. When computing sensitivity, the

numerator is the number of ill persons who arecorrectly identified as ill by the test, and thedenominator is the total number of ill persons[a ⁄ (a + c)]. Specificity (true negative rate) refers tohow well the screening test identifies individualswithout the illness. When computing specificity,the numerator is the number of persons withoutthe illness who are correctly identified by the test asnot having the illness, and the denominator is thetotal number of persons without the illness[d ⁄ (b + d)].The predictive values of a test are generally more

clinically meaningful than a test�s sensitivity andspecificity. The predictive value statistics indicatethe probability that an individual is ill or not ill,given that the screening test identifies them as illor not ill. Accordingly, positive predictive valuerefers to the probability that a person who isidentified as ill by the test actually has the illness.When computing positive predictive value, thenumerator is the same as it is in computingsensitivity (i.e., the number of ill persons who arecorrectly identified as ill by the test); however, thedenominator is now the total number of personsthat the test identifies as ill [a ⁄ (a + b)]. Negativepredictive value refers to the probability that aperson who is identified as not ill by the test actuallyis not ill. When computing negative predictivevalue, the numerator is the same as it is incomputing specificity (i.e., the number of not illpersons who are correctly identified as not ill by thetest), although the denominator is the total numberof persons that the test identifies as not ill[d ⁄ (c + d)].Evaluations of an instrument often focus on the

sensitivity of the measure if it is to be used forscreening. High sensitivity means that most peoplewho have the disorder screen positive on the test.

a b

c d

Gold standard diagnosis

Present Absent Total

Positive

Screening test

Negative

a + b

c + d

Total a + c b + d a + b + c + d

Sensitivity = a/(a + c). Specificity = d/(b + d). Positive predictive value = a/(a + b). Negative predictive value = d/(c + d).

Fig. 1. Table of association between a diagnostic gold stand-ard and screening test.

Zimmerman

128

However, the tradeoff for selecting a threshold onthe test resulting in high sensitivity is lowerspecificity. As a result of lowered specificity, moreindividuals who screen positive do not in fact havethe disorder (i.e., they are false positives).If an instrument is used for case-finding as

opposed to initial screening, then it is misleadingto focus on the measure�s sensitivity. Rather, thepositive predictive value of the instrument is themost important statistic. If the positive predictivevalue of a measure such as the MDQ is high andthe vast majority of individuals who screen positiveon the MDQ have BD, then authors can moreconfidently extrapolate their findings to the diag-nosis of BD. If, however, the positive predictivevalue is modest and the majority of individualswho screen positive do not have BD, but insteadhave another psychiatric disorder, then the findingsof these studies refer to the implications ofunspecified types of psychopathology rather thanBD specifically.The positive predictive value of a test is strongly

affected by the prevalence of the disorder beingscreened for (33, 34). At constant sensitivity andspecificity, the positive predictive value will belower in samples with a lower prevalence. In fact,in samples with low prevalence it is difficult toestablish an adequate positive predictive value (33).The impact of disorder prevalence on positivepredictive value is illustrated by comparing theresults of the three studies of the MDQ inpsychiatric outpatients with mixed diagnoses (35–37). In each study, the Structured Clinical Inter-view for DSM-IV (38) was used as the diagnosticstandard, and the MDQ was scored according tothe recommendations of Hirschfeld et al. (22). Theresults of the three studies are summarized inTable 1. The sensitivity and specificity of the MDQwere similar in the studies of Hardoy et al. (35) andZimmerman et al. (37), yet because the prevalenceof BD was approximately 20% higher in Hardoyet al.�s (35) sample, the positive predictive valuewas much higher. By contrast, in the Zimmermanet al. (37) and Konuk et al. (36) studies, theprevalence of BD was the same, and the smalldifference in positive predictive value directly

reflected the difference in specificity between thestudies.A low prevalence of BD, and corresponding low

positive predictive value, is the basis for thecriticism of studies that have used the MDQ as acase-finding measure in general population andprimary care samples. Because of low prevalenceand the consequent low positive predictive value ofthe MDQ in these studies, the majority of subjectswho screen positive will have other psychiatricdisorders rather than BD. To highlight this issue,Zimmerman et al. (39) reported that patients whoscreened positive on the MDQ were more likely tobe diagnosed with borderline personality disorderthan to be diagnosed with BD. Thus, in reading thesummaries below of studies of the correlates ofMDQ caseness, the reader should ask whether thefindings could be accounted for by another psy-chiatric disorder, such as borderline personalitydisorder, rather than BD.

Inappropriate conclusions and inferences from

studies using the MDQ as a case-finding

instrument

In a study of the general population, Calabrese andcolleagues (40) compared MDQ-positive andMDQ-negative subjects on indices of psychosocialimpairment. The authors interpreted the results inreference to diagnosable BD. The stated reason forthe study was because, �it is essential that healthcare providers fully understand the nature andintensity of the psychosocial impact of BD in theUnited States, so that appropriate diagnosticprocedures and treatments are offered earlier inthe course of the disease to prevent humansuffering and reduce the burden of illness� (40,p. 426). In the discussion, the authors cautionedthe reader to interpret the data carefully becausethe MDQ is a screening tool, not a diagnosticmeasure. However, they did not raise the criticallyimportant issue of low positive predictive valueand that most individuals in a general populationsample who screen positive on the MDQ will nothave BD. While the authors described and dis-cussed the findings in terms of the differences

Table 1. Studies of the operating characteristics of the Mood Disorders Questionnaire (MDQ) in three studies of psychiatric outpatients

Study N

Prevalenceof bipolardisorder, %

Prevalence ofMDQ cases, % Sensitivity, % Specificity, %

Positivepredictivevalue, %

Negativepredictivevalue, %

Hardoy et al. 2005 (35) 154 33.1 31.2 66.7 86.4 70.8 84.0Konuk et al. 2007 (36) 309 11.7 27.8 63.9 76.9 26.7 94.2Zimmerman et al. 2009 (37) 480 10.8 20.4 63.5 84.8 33.7 95.0

Misuse of the MDQ

129

between MDQ-positive and MDQ-negative sub-jects, the clearest evidence that they were extrap-olating their data to diagnosable BD was reflectedby the title of the paper �Impact of bipolar disorderon a US community sample.�Another report from this study examined health

service utilization in respondents who were MDQpositive (41). In contrast to the title of the paper byCalabrese et al., the title of this paper identifiedit as a study of individuals who screened positivefor BD (�Use of health care services among personswho screen positive for bipolar disorder�). Whilethe title refers to screening, the language of thearticle suggests that the authors interpreted MDQ-positives as true cases. They found that more thanhalf of the MDQ-positives were not diagnosed withBD by treating clinicians, and indicated that theseclinicians �failed to detect� or �misdiagnosed� BD inthese individuals. They further examined the med-ications received by the MDQ-positive cases andfound that a low percentage were prescribed moodstabilizers and a higher percentage were prescribedantidepressants. The authors indicated that thesepatients were �inappropriately treated.� As evidencethat the authors viewed the MDQ-positives as truecases, the first sentence of the discussion sectionstated, �The results of this US-population-basedstudy suggest that bipolar disorder is frequentlyundetected or misdiagnosed, even among patientswho consult psychiatrists.� (41, p. 1532). Later in thesame paragraph, the authors wrote, �That fewerthan one in four respondents who screened positiveand who consulted a primary care physician receiveda diagnosis of bipolar disorder is alarming …�Elsewhere in the discussion, the authors wrote,�The underuse of mood stabilizers among thesepatients with bipolar disorder is particularly wor-risome …� To be sure, towards the end of thediscussion, the authors offered the caveat that,�Unequivocally, a limitation of the study is that theMDQ is a screening instrument for bipolar disor-der and not a diagnostic instrument� (41, p. 1533).However, they then go on to undermine this caveatby briefly summarizing studies suggesting that theMDQ has adequate sensitivity and specificity, butthey do not discuss the issue of limited positivepredictive value. And, despite their caveat aboutthe MDQ being a screening instrument, theynonetheless concluded, �The results of this US-population-based study indicate that bipolardisorder is an under-diagnosed and often inappro-priately treated illness. Underdiagnosis and inap-propriate treatment of bipolar disorder wereparticularly common in the primary care settingbut were also alarmingly high among psychiatrists,who did not detect bipolar disorder among

approximately half of respondents who screenedpositive for the disorder on the MDQ� (41, p. 1533).Clearly, the language of the authors was consistentwith the MDQ interpreted as a diagnostic proxy.Goldney and colleagues (42) administered the

MDQ to more than 3000 residents of Australia. Inthe introduction, the authors noted the challenge indiagnosing BD and the significant psychosocialimpact of this condition. They described the MDQas, �an advance in providing a reliable and validinstrument to assess the prevalence of bipolar I andII disorders.� The prevalence of MDQ-positivecases was 2.5%, and compared to MDQ-negativesubjects, MDQ-positive respondents reportedpoorer health-related quality of life, poorer healthstatus, and more frequently were treated by mentalhealth care specialists. In the discussion, theauthors compared the prevalence rate to thatfound in other epidemiological studies of BD andindicated that, �the 2.5% lifetime prevalence forbipolar I and II disorders represents a credibleassessment in this Australian population.� Therewas no discussion that the MDQ was a screening,not a diagnostic, instrument. Consistent with theiruse of the MDQ as a diagnostic proxy, the title ofthe paper was �Bipolar I and II disorders in arandom and representative Australian population.�Another paper based on the Australian epide-

miological data set examined the financial costassociated with BD (42, 43). In the methods sectionof this paper, the authors indicated that the MDQ�is considered a valid screening instrument forcommunity settings since it effectively screens outalmost all true non-bipolar cases and does notover-estimate or detect too many false positivecases of bipolar disorder.� It is not clear how manyfalse positives are �too many.� In a critique of acommunity study using the MDQ in the USA,Zimmerman et al. (34) suggested that the false-positive rate was higher than the true-positive rate,and the positive predictive value of the scale wasonly approximately 25%. While the authors of theAustralian study referred to the MDQ as ascreening scale, they interpreted it as a diagnosticmeasure. In the discussion they stated, �the MDQmeasures the lifetime prevalence of bipolar disor-ders,� and the title of the paper was �Bipolardisorders in Australia: a population-based study ofexcess costs.�Das et al. (44) found that nearly 10% of 1157

patients seeking primary care at an urban generalmedical practice servicing low-income individualsscreened positive on the MDQ. MDQ-positivecases had more depressive, anxiety, and substanceuse disorders, suicidal ideation, mental health care,and functional impairment. Throughout the results

Zimmerman

130

section, the authors referred to the MDQ-positivepatients as �participants who screened positive forbipolar disorder,� although they also indicated thatMDQ-positive cases reflected �probable� BD andnoted how �remarkably� few patients had beenpreviously diagnosed with BD. In discussing thereasons for the low diagnostic rate, the authorsoffered four possible explanations for the low rateof previous BD diagnoses without mentioninglow positive predictive value. Most of the intro-duction and discussion focused on the clinical andpublic health significance of BD. The usual caveatabout the MDQ being a screening instrument wasincluded in the discussion, but this did not precludethe authors from concluding that BD is under-recognized in primary care and that primary carephysicians should receive greater education aboutthe recognition and treatment of BD.A second report from the low-income primary

care study by Olfson et al. (45) compared MDQ-positive and MDQ-negative depressed patients.Depression was �diagnosed� according to the self-report Patient Health Questionnaire (46). In con-trast to the Das et al. (44) report, the title of thispaper did not mention screening (�Bipolar depres-sion in a low-income primary care clinic�). Also incontrast to Das et al.�s study, throughout the textand tables, the patients who screened positive werereferred to as having BD and not as screenpositives. The depressed patients who screenedpositive reported more suicidal ideation, low self-esteem, comorbid substance use disorders, lifetimehistory of hallucinations, and psychiatric hospitali-zations. Olfson et al. suggested that this providedconstruct validity for the MDQ. However, thesame differences would be expected betweendepressed patients with and without borderlinepersonality disorder (47). As with the other studiesdescribed above, there was no acknowledgement offalse positives and the limited positive predictivevalue of the MDQ.Kamat et al. (48) used the MDQ to screen for

BD in 1360 patients with major depressive disorderand found a positive screen rate of 7%. Through-out most of the article, they were careful to usecorrect terminology, noting a positive MDQ screenas opposed to a BD diagnosis, although in oneinstance in the conclusion, they wrote inappropri-ately that the �bipolar disorder misdiagnosis ratewas 7%.� More concerning than that (given thatthe error was only a single instance), they reviewedthe burden associated with an MDQ-positivescreen and never entertained the possibility thatthis burden might have been due to other disorders(such as borderline personality disorder) as opposedto BD. Such neglect of alternative explanations is

common, given the presumption in these articlesthat a positive screen on the MDQ �is likely to�reflect BD, a presumption that is not supported byavailable research.Other studies that have used the MDQ have

similarly drawn conclusions about the prevalence,diagnosis, and recognition of BD. For example,Carta et al. (49) administered the MDQ to patientsbeing treated for fibromyalgia, found a higherprevalence of MDQ-positive cases compared to acontrol group and raised a caution in prescribingantidepressants to patients with fibromyalgia.Similarly, Wilke et al. (27), in a paper entitled�Fibromyalgia and bipolar disorder: a potentialproblem?,� included the MDQ as part of thestandard assessment battery of patients withfibromyalgia and reported that one-quarter of thepatients screened positive on the scale. Theyvariously referred to these patients as having�probable,� �putative,� and �potential� BD and,similarly to Carta et al. (49), discussed concernsabout placing patients with fibromyalgia at riskwhen prescribing antidepressants. Lim and col-leagues (30) administered the MDQ to depressedpatients initiating interferon treatment for chronichepatitis C infection. As in the report by Olfsonand colleagues (45), depression was �diagnosed�according to the self-report Patient Health Ques-tionnaire (46). Lim et al. reported that patientswho screened positive for BD were more likely todiscontinue interferon due to psychiatric com-plications. They referred to the MDQ-positivegroup as having �manic traits.� While they notedthat the MDQ is not a diagnostic measure, theycountered this caution by also noting that �a highpercentage of these patients are likely to have atrue diagnosis of BD disorder rather than MDD;�admittedly, the authors did not define what theymeant by high percentage. Further, they did notdiscuss the issue of positive predictive value; thus,they did not indicate that it is likely that only aminority of individuals who screened positivewould have been diagnosed with BD.Hirschfeld et al. (50) sent the MDQ to a demo-

graphically representative random sample of morethan 125000 adults in the USA and receivedcompleted forms from more than 85000. Theadjusted prevalence rate for MDQ-positive screenswas 3.7%. The authors suggested that the preva-lence of bipolar I and II disorders was higher thanhad been previously thought, with implications forthe public health significance of BD in the USA.They highlighted the finding that only 20% of thesubjects who screened positive had been previouslydiagnosed with BD, whereas 31% had been diag-nosed with major depression. They therefore

Misuse of the MDQ

131

concluded that many patients with bipolar I and IIdisorders may be misdiagnosed with unipolardepression. Positive screens were associated withhigher rates of substance use disorders, youngerage, and lower income, and interpreted this interms of the correlates of BD. Similarly to theother studies described above, the issue of positivepredictive value was not raised, and the authors didnot consider the possibility that the findings couldbe explained by the MDQ identifying non-bipolarforms of psychopathology. While, undoubtedly,some individuals who screen positive on the MDQhave BD, in general population and primary caresamples where the prevalence is expected to be low,the majority of individuals who screen positive willnot be found to have BD upon a more compre-hensive diagnostic evaluation.Recently, misuse of the MDQ as a diagnostic

proxy has begun to appear in studies of psychiatricpatients. van den Berg and colleagues (31), in apaper entitled, �Manic symptoms in patients withdepressive and ⁄or anxiety disorders,� administeredthe MDQ to more than 2000 subjects with a moodor anxiety disorder but who had not been diag-nosed with BD. The subjects were administered asemi-structured interview assessing depressive,anxiety, and alcohol use disorders, but not BD.Screening positive on the MDQ was associatedwith increased rates of anxiety disorders, alcoholdependence, and a lifetime history of suicideattempt. Contrary to the authors� prediction,results on the MDQ were not associated withreverse vegetative symptoms of depression. Indiscussing their findings, the authors noted that alimitation of the study was the lack of diagnosticinterview for BD but indicated that the MDQ-positive group consisted of two subgroups: thosewith unrecognized BD and those with subsyndro-mal manic symptoms. They did not consider thepossibility that some, probably the majority, of theMDQ positives were false positives.Kiejna et al. (28) and Dudek et al. (29), in two

reports from the same study, found that an MDQ-positive screen was significantly more common inpatients with treatment-resistant depression thannontreatment-resistant depression and concludedthat bipolarity was associated with worse responseto antidepressant medication.

A conceptual critique of using a screening scale for

BD in clinical practice

The conceptual basis for developing a screeningmeasure for BD has never been well articulated.Indeed, BD is sometimes unrecognized in patientswith psychiatric illness. Some may even argue that

BD is often unrecognized. The question is why?Are clinicians failing to inquire about a history ofmania or hypomania? Are they failing to inquireproperly? Are they asking questions that areinsufficiently sensitive for detecting BD? Are theyfailing to correctly interpret patients� responses toappropriate inquiry?It could be argued that understanding the reason

for under-detection, while illuminating, should notpreclude the use of a test that could improve therecognition of BD. While the sentiment behind thisargument is understandable, and efforts to im-prove diagnostic recognition and accuracy arelaudable, advocates of using screening scales forBD have not adequately discussed the twopotential negative consequences of routine use ofthe scale.First, the sensitivity of the scale is less than 65%

(26). It is important for a screening scale to havehigh sensitivity because the more time-intensivefollow-up diagnostic inquiry will presumably onlyoccur in patients who screen positive. Clinicianswho rely on the screening scale as the first stage ina two-stage process for diagnosing BD will miss thediagnosis in approximately one-third of patientswith BD.Second, the modest positive predictive value of

the scale raises concern that routine use of theMDQ in clinical practice may result in BD over-diagnosis if clinicians over-rely on the scale for caseidentification without adequate follow-up evalua-tion. This problem is not unique to BD. In anysample in which the prevalence of a disorder isrelatively low, the majority of individuals whoscreen positive will not have the disorder uponadministration of the more valid and comprehen-sive diagnostic assessment. If, as illustrated above,researchers are prone to misinterpret the results onthe MDQ to indicate actual caseness rather thanscreen positives, then I would expect clinicians tofall into the same trap.No one would argue that use of the MDQ

eliminates the need for a competent psychiatricevaluation. Because of the MDQ�s modest positivepredictive value, most patients who screen positivewill not have BD; thus, patients who screenpositive on the MDQ need to be evaluated todetermine if, in fact, they have BD. The questionthat should be asked, then, is why should a self-administered screening questionnaire for BD beused by mental health clinicians? There are nospecial questions on the MDQ that are unfamiliarto a competent clinician. If a psychiatrist is awareof the importance of screening for BD, under whatcircumstances would reliance on a paper-and-pencil screening questionnaire be preferable to

Zimmerman

132

direct inquiry in screening for a history of mania orhypomania? There is no substitute for a competentclinical interview.It could be argued that a screening scale will

improve the efficiency of the diagnostic evaluation.This is true. If the goal of screening is to improveefficiency by identifying for the clinician whichpatients do not require inquiry for a history of BD,then a self-report scale with high sensitivity couldbe of value if the measure also has high negativepredictive value. High negative predictive valueindicates that the clinician need not inquire for thedisorder because it is unlikely to be present.Clinicians who are too time-pressured to inquirefor a history of BD could therefore benefit fromroutine use of the MDQ. However, given theclinical importance of accurately diagnosing BD, Iwould have concerns about clinicians who indi-cated that they did not have the time to routinelyinquire for a history of mania or hypomania.

Conclusions

While there is clear evidence that BD is sometimesunder-diagnosed in clinical practice, the reasons forunder-diagnosis havenotbeenwell investigated, andit is unknown whether the potentially improveddetection resulting from screening will be offset byover-diagnosis of BD.Anumber of researchers haveover-interpreted MDQ results to suggest caseness,and this raises concerns that this could occur inclinical practice. Experts should therefore heedcaution before recommending routine clinical useof any self-administered screening questionnaire forBD in the absence of studies simultaneously exam-ining both the benefits (e.g., improved detection)and costs (e.g., over-diagnosis and subsequentoverprescribing of medication) of screening. In fact,in the absence of a single study comparing the risksand benefits of using a self-report questionnaire toscreen for BD in clinical practice, questions shouldbe raised as towhether recommendations to use suchscales should be allowed in industry-sponsoredpresentations because of an apparent conflict ofinterest. There should be a concern that recommen-dations to use such scales are primarily intended toincrease the number of individuals who are diag-nosed with BD in order to sell more medications,and that these recommendations come withoutevidence that the use of such scales result in overallimproved outcome.

Disclosure

The author of this paper has no financial conflicts of interest toreport.

References

1. Merikangas KR, Akiskal HS, Angst J et al. Lifetime and12-month prevalence of bipolar spectrum disorder in theNational Comorbidity Survey replication. Arch Gen Psy-chiatry 2007; 64: 543–552.

2. Bowden CL. Strategies to reduce misdiagnosis of bipolardepression. Psychiatr Serv 2001; 52: 51–55.

3. Katzow JJ, Hsu DJ, Ghaemi SN. The bipolar spectrum: aclinical perspective. Bipolar Disord 2003; 5: 436–442.

4. Manning JS, Haykal RF, Connor PD, Akiskal HS. On thenature of depressive and anxious states in a family practicesetting: the high prevalence of bipolar II and relateddisorders in a cohort followed longitudinally. ComprPsychiatry 1997; 38: 102–108.

5. Ghaemi SN, Boiman EE, Goodwin FK. Diagnosingbipolar disorder and the effect of antidepressants: anaturalistic study. J Clin Psychiatry 2000; 61: 804–808.

6. Ghaemi SN, Ko JY, Goodwin FK. ‘‘Cade�s disease’’ andbeyond: misdiagnosis, antidepressant use, and a proposeddefinition for bipolar spectrum disorder. Can J Psychiatry2002; 47: 125–134.

7. Hirschfeld RM. Bipolar spectrum disorder: improving itsrecognition and diagnosis. J Clin Psychiatry 2001; 62(Suppl. 14): 5–9.

8. Hirschfeld RM, Vornik LA. Recognition and diagnosisof bipolar disorder. J Clin Psychiatry 2004; 65 (Suppl. 15):5–9.

9. Yatham LN. Diagnosis and management of patients withbipolar II disorder. J Clin Psychiatry 2005; 66 (Suppl. 1): 13–17.

10. Hantouche EG, Akiskal HS, Lancrenon S et al. Systematicclinical methodology for validating bipolar-II disorder:data in mid-stream from a French national multi-site study(EPIDEP). J Affect Disord 1998; 50: 163–173.

11. Perugi G, Akiskal HS, Lattanzi L et al. The high preva-lence of ‘‘soft’’ bipolar (II) features in atypical depression.Compr Psychiatry 1998; 39: 63–71.

12. Hirschfeld RM, Lewis L, Vornik LA. Perceptions andimpact of bipolar disorder: how far have we really come?Results of the national depressive and manic-depressiveassociation 2000 survey of individuals with bipolar disor-der. J Clin Psychiatry 2003; 64: 161–174.

13. Lish J, Dime-Meenan S, Whybrow P, Price R, HirschfeldR. The National Depressive and Manic-depression Asso-ciation (DMDA) survey of bipolar members. J AffectDisord 1994; 31: 281–294.

14. Birnbaum HG, Shi L, Dial E, Oster EF, Greenberg PE,Mallett DA. Economic consequences of not recognizingbipolar disorder patients: a cross-sectional descriptiveanalysis. J Clin Psychiatry 2003; 64: 1201–1209.

15. Shi L, Thiebaud P, McCombs JS. The impact of unrecog-nized bipolar disorders for patients treated for depressionwith antidepressants in the fee-for-services CaliforniaMedicaid (Medi-Cal) program. J Affect Disord 2004; 82:373–383.

16. Matza LS, Rajagopalan KS, Thompson CL, de LissovoyG. Misdiagnosed patients with bipolar disorder: comor-bidities, treatment patterns, and direct treatment costs.J Clin Psychiatry 2005; 66: 1432–1440.

17. Judd L, Akiskal H, Schettler P et al. The long-term naturalhistory of the weekly symptomatic status of bipolar Idisorder. Arch Gen Psychiatry 2002; 59: 530–537.

18. Judd L, Akiskal H, Schettler P et al. A prospectiveinvestigation of the natural history of the long-term weeklysymptomatic status of bipolar II disorder. Arch GenPsychiatry 2003; 60: 261–269.

Misuse of the MDQ

133

19. Isometsa E, Suominen K, Mantere O et al. The MoodDisorder Questionniare improves recognition of bipolardisorder in psychiatric care. BMC Psychiatry 2003; 3: 4.

20. Ketter TA. Strategies for monitoring outcomes in patientswith bipolar disorder. Primary Care Companion J ClinPsychiatry 2010; 12: 10–16.

21. Depue RA, Krauss S, Spoont MR, Arbisi P. GeneralBehavior Inventory identification of unipolar and bipolaraffective conditions in a nonclinical university population.J Abnorm Psychol 1989; 98: 117–126.

22. Hirschfeld R, Williams J, Spitzer R et al. Development andvalidation of a screening instrument for bipolar spectrumdisorder: The Mood Disorder Questionnaire. Am JPsychiatry 2000; 157: 1873–1875.

23. Angst J, Adolfsson R, Benazzi F et al. The HCL-32:towards a self-assessment tool for hypomanic symptoms inoutpatients. J Affect Disord 2005; 88: 217–233.

24. Bowden CL, Singh V, Thompson P et al. Development ofthe bipolar inventory of symptoms scale. Acta PsychiatrScand 2007; 116: 189–194.

25. Hirschfeld R, Holzer C, Calabrese J et al. Validity of themood disorder questionnaire: a general population study.Am J Psychiatry 2003; 160: 178–180.

26. Zimmerman M, Galione JN. Screening for bipolar disor-der with the Mood Disorders Questionnaire: a review.Harv Rev Psychiatry 2011; 19: 219–228.

27. Wilke WS, Gota CE, Muzina DJ. Fibromyalgia andbipolar disorder: a potential problem? Bipolar Disord2010; 12: 514–520.

28. Kiejna A, Pawlowski T, Dudek D et al. The utility ofMood Disorder Questionnaire for the detection of bipolardiathesis in treatment-resistant depression. J Affect Disord2010; 124: 270–274.

29. Dudek D, Rybakowski JK, Siwek M et al. Risk factors oftreatment resistance in major depression: association withbipolarity. J Affect Disord 2010; 126: 268–271.

30. Lim C, Olson J, Zaman A, Phelps J, Ingram KD.Prevalence and impact of manic traits in depressed patientsinitiating interferon therapy for chronic hepatitis C infec-tion. J Clin Gastroenterol 2010; 44: 141–146.

31. van den Berg B, Penninx BW, Zitman FG, Nolen WA.Manic symptoms in patients with depressive and ⁄ oranxiety disorders. J Affect Disord 2010; 126: 252–256.

32. Rouillon F, Gasquet I, Garay RP, Lancrenon S. Screeningfor bipolar disorder in patients consulting generalpractitioners in France. J Affect Disord 2011; 130: 492–495.

33. Phelps JR, Ghaemi SN. Improving the diagnosis of bipolardisorder: predictive value of screening tests. J Affect Disord2006; 92: 141–148.

34. Zimmerman M, Posternak M, Chelminski I, Solomon D.Using questionnaires to screen for psychiatric disorders: acomment on a study of screening for bipolar disorder in thecommunity. J Clin Psychiatry 2004; 65: 605–610.

35. Hardoy MC, Cadeddu M, Murru A et al. Validation of theItalian version of the ‘‘Mood Disorder Questionnaire’’ for

the screening of bipolar disorders. Clin Pract EpidemolMent Health 2005; 1: 8.

36. Konuk N, Kiran S, Tamam L, Karaahmet E, Aydin H,Atik L. Validation of the Turkish version of the mooddisorder questionnaire for screening bipolar disorders.Turkish J Psychiatry 2007; 18: 147–154.

37. Zimmerman M, Galione JN, Ruggero CJ et al. Perfor-mance of the Mood Disorders Questionnaire in a psychi-atric outpatient setting. Bipolar Disord 2009; 11: 759–765.

38. First MB, Spitzer RL, Gibbon M, Williams JBW. Struc-tured Clinical Interview for DSM-IV Axis I Disorders –Patient edition (SCID-I ⁄P, version 2.0). New York:Biometrics Research Department, New York State Psy-chiatric Institute, 1995.

39. Zimmerman M, Galione JN, Ruggero CJ et al. Screeningfor bipolar disorder and finding borderline personalitydisorder. J Clin Psychiatry 2010; 71: 1212–1217.

40. Calabrese JR, Hirschfeld RM, Reed M et al. Impact ofbipolar disorder on a U.S. community sample. J ClinPsychiatry 2003; 64: 425–432.

41. Frye MA, Calabrese JR, Reed ML et al. Use of health careservices among persons who screen positive for bipolardisorder. Psychiatr Serv 2005; 56: 1529–1533.

42. Goldney RD, Fisher LJ, Grande ED, Taylor AW, Haw-thorne G. Bipolar I and II disorders in a random andrepresentative Australian population. Aust N Z J Psychi-atry 2005; 39: 726–729.

43. Fisher LJ, Goldney RD, Dal Grande E, Taylor AW,Hawthorne G. Bipolar disorders in Australia. A popula-tion-based study of excess costs. Soc Psychiatry PsychiatrEpidemiol 2007; 42: 105–109.

44. Das A, Olfson M, Gameroff M et al. Screening for bipolardisorder in a primary care practice. JAMA 2005; 293: 956–963.

45. Olfson M, Das AK, Gameroff MJ et al. Bipolar depressionin a low-income primary care clinic. Am J Psychiatry 2005;162: 2146–2151.

46. Kroenke K, Spitzer R, Williams J. The PHQ-9. Validity ofa brief depression severity measure. J Gen Intern Med2001; 16: 606–613.

47. Galione J, Zimmerman M. A comparison of depressedpatients with and without borderline personality disorder:implications for interpreting studies of the validity of thebipolar spectrum. J Personal Disord 2010; 24: 763–772.

48. Kamat SA, Rajagopalan K, Pethick N, Willey V, BullanoM, Hassan M. Prevalence and humanistic impact ofpotential misdiagnosis of bipolar disorder among patientswith major depressive disorder in a commercially insuredpopulation. J Manag Care Pharm 2008; 14: 631–642.

49. Carta MG, Cardia C, Mannu F et al. The high frequencyof manic symptoms in fibromyalgia does influence thechoice of treatment? Clin Pract Epidemol Ment Health2006; 2: 36.

50. Hirschfeld R, Calabrese J, Weissman M et al. Screeningfor bipolar disorder in the community. J Clin Psychiatry2003; 64: 53–59.

Zimmerman

134