Data-driven approaches: Creating applications in clinical
settings
Another approach in the search for predictive biomarkers, is the use of
data-driven approaches such as machine learning. These methods have been
used successfully on neuroimaging and clinical data to answer questions
about prognosis and treatment outcomes. In one study, using existing
clinical trial data, clinical variables were used to predict remission
in depression following treatment with selective serotonin re-uptake
inhibitors (SSRIs)[4]. Supervised dimensionality reduction enabled
large numbers of clinical variables to be reduced to a select few
predictive variables, which were then fed into a machine learning
algorithm to predict remission in patients with depression. Applying
this algorithm had a number needed to treat (NNT) of 14 to achieve
remission in depression[1]. Ultimately, machine learning methods
could provide the basis for the creation of tools to support clinical
decision making.