MDQ Validity Concerns
A commentary by Dr. Mark Zimmerman and colleagues from Brown University (J Clin Psych May 2004) challenges the validity of the MDQ. The commentary format does not include an abstract, so you won't find one online; nor the reply in the same issue by Hirschfeld et al. Some excerpts from each are provided below. Obtaining the originals is definitely worth your time if you're concerned about MDQ validity. However, an analysis of the MDQ validity studies appears to indicate that the concern about low sensitivity raised by Zimmerman et al may have little effect on test performance when primary care doctors are ruling out bipolar disorder before giving antidepressants.
Based on the data in the general population study (Hirschfeld et al 2003), particularly the low sensitivity and resultant low positive predictive value, the authors conclude that "the scale is not sufficiently accurate to be used as either a case-finding measure in community studies or a screening scale in clinical practice."
Further, they find the psychiatric outpatient study (Hirschfeld et al 2000) is limited in value because it was:
... conducted with patients already being treated in specialized mood disorders clinics at academic medical centers that had a particular expertise in treating bipolar disorder. It could be expected that patients in ongoing treatment in these clinics had received psychoeducation and were more aware of their diagnoses than are patients usually treated in the community. Also, the prestige of these clinics may have resulted in more prototypical and refractory cases being treated. [This seems a valid concern, which also applies to Miller et al, cited among other validity studies.]
Dr. Hirschfeld and colleagues (a list of well-known mood experts, though nearly all have substantial pharmaceutical company ties as is disclosed with their letter) responded in a letter published in the same issue . They raise 3 points in defense of the MDQ as a screening tool:
1. Very low sensitivities are common with screening tools that look for lifetime symptoms, as opposed to current symptoms. Their examples:
Study Sensitivity (%) Specificity
MDQ in the general population 28 97 Epidemiologic Catchment Area, depression self-report 29 96 Center for Epidemiologic Studies, Depression Scale (CES-D) 34 90 Dysthymia self-report (Roberts et al) 49 74 Michigan Alcohol Screening Test (MAST) 38 97 Alcohol Dependence Data scale (SADD) 11 99
2. By contrast, screening for current symptoms generally yields much higher sensitivities (they offer examples in the 80-90% range).
3. There was a 6-month delay between the mailed-in MDQ result, and the follow-up Structured Clinical Interview, DSM-IV (SCID), which served as the gold standard. When the MDQ was re-administered immediately following the SCID, here's what happened:
Sensitivity (%) Specificity
MDQ 6 months before SCID 28 97 MDQ immediately after SCID 75 94
In conclusion, they acknowledge that:
...there is much to be improved in the ability to screen for bipolar disorder in the community, and we agree that the MDQ has limitations. However, given the public health significance of bipolar disorder and its neglect in epidemiologic and primary care studies, having new tools for screening represents a step forward. The MDQ represents the current state of the art. We await the development of improved screening tools for bipolar disorder.
Here is one more "big picture" view of these validity data concerns, based on analyzing the predictive values which result when the tests are used in low and high-prevalence contexts.