Over the past five years, the potential for web search volume to provide insight into the present has become increasingly apparently. To date, much of these methods have focused on syndromic keywords and are not directly suitable for surveillance of less common or highly variable diseases. The more unique mapping between drug therapy had disease provides a potential workout this problem. We demonstrate the high degree of correlation between search volume and drug utilization and apply this method of keyword generation to modeling drug utilization, patent expirations and TB incidence in the US.