MDD

Smoller, Perlis, Iosifescu et al.,

Tianxi showed that the incremental value of billing codes to NLP'd characterizations of depressed patients was minimal at a wide range of false positive rates.

Then a longer discussion regarding controls.

• Utilization is a bad match, because increased utlization is confounded with psychiatric overlay.
• Completely healthy individuals is a bad match because we would be selecting for populations with a lower burden of risk alleles for numerous diseases.

(The following courtesy of Patience Gallagher):

Isaac Kohane, Susanne Churchill, Jordan Smoller, Roy Perlis, Sergey Goryachev, Shawn Murphy, Dan Iosifescu, Victor Castro, Tianxi Cai, Vivian Gainer, Wouter Hoogenboom, Margarita Sordo, Stefanie Block, Patience Gallagher