RT Journal
A1 Bombardier CH, Temkin NR, Fann JR
T1 TRaumatic brain injury and major depressive disorder—reply
JF JAMA
JO JAMA
YR 2010
FD August 25
VO 304
IS 8
SP 857
OP 858
DO 10.1001/jama.2010.1171
UL http://dx.doi.org/10.1001/jama.2010.1171
AB 
Second, Thombs uses sensitivity and specificity data from a study by Fann et al2 to argue that if the PHQ-9 has a specificity of 0.89, repeated screenings would result in a large increase in false positives. However, the specificity data are based on 135 independent cases,2 whereas in our study, observations are likely positively correlated because each participant was assessed multiple times. In situations with positively correlated data, the number of false positives would not escalate to the degree suggested by Thombs. Using his assumptions, with 6 assessments per person, about 50% of those without MDD would screen positive at least once. If the true depression rate were 10%, 10 of 100 people would be true positives and 45 would be false positives, giving an expected positive predictive value of 10/(10 + 45) = 18%. However, 181 of the PHQ-9 positive cases in our study were administered the Structured Diagnostic Interview for DSM-IV (SCID) at least once, and 115 of them met SCID criteria for MDD. This gives an observed positive predictive value of 64% and suggests a false-positive rate for the repeated screening much lower than Thomb's assumptions yield.