MDQ versus BSDS: accuracy


  • MDQ: Mood Disorders Questionnaire
  • BSDS: Bipolar Spectrum Diagnostic Scale
  • HCL-32: Hypomania Checklist, 32-item

What’s the best screening tool for bipolar disorder in primary care? An international team recently reviewed the evidence. Carvalho

Bottom line: the BSDS has the fewest studies. Overall it appeared to work as well as the others.  HCL-32 is better than the MDQ for finding more subtle versions of bipolar disorder (Bipolar II and Bipolar NOS).

Conclusion: in my opinion, the BSDS is more compatible with primary care: it’s much faster than the HCL-32.  In one head-to-head study these two tools performed equally well.Smith

That’s why I’ve used the BSDS in the screening tool I built for primary care, MoodCheck (no money in it for me; it’s public sector, unlike the MDQ).

More gory details; stop now unless obsessed, like me

Once again, this review shows that predictive values of these tests are simply inadequate.  In their section 3.3.3, the authors note that if one looks at all the studies and places specificity at 69% (by adjusting cut-offs), then for a simulated primary care cohort of 100 patients, at a bipolar prevalence of 15% (15 out of the 100 patients have bipolar disorder, by some presumed gold standard), the sensitivity is 68% for the MDQ and 78% for the BSDS.  Not good but not awful.

But when you take those numbers through a 2×2 table to calculate predictive values, you get (for the BSDS):

  • predictive value of a negative test: 3 / 59 + 3 =  95% , very good
  • predictive value of a positive test: 12/ 12 + 26 = 32%, very poor

Most positives are false positives at a 15% prevalence rate, even with a reasonably good test.   If the test it weak, as it is here, then the predictive value of a positive test is unacceptably low.