Medacist was selected for 5 Best Big Data Companies to Watch 2020 by the Silicon Review. Check out the full article for a detailed interview on what David Brzozowski, Jr., CTO had to say about Medacist’s technologies current and future.
How Medacist improves Drug Diversion Prevention using big data and AI?
The core competency of a SaaS business is to provide value to its users. At Medacist our goal is to deliver insight to our clients that can be acted on, ultimately serving as the catalyst that streamlines the decision-making process. The value Medacist is able to deliver to clients is measured by the speed and accuracy in which we are able to impact the decision-making process of our clients. Outlined below are some of the techniques Medacist leverages to indefinitely tune our solutions and optimize our clients’ ultimate value.
Reinforcement Learning Feedback Through Reinforcement Learning, analysis can be tuned based on feedback provided by end-users. For example, in RxAuditor Investigate (Medacist’s Diversion Prevention Solution), end-users of the application can provide feedback on the analysis presented. This feedback is then fed back into our neural network for consideration on subsequent analysis. As a result, end-users that are more active in providing feedback will see subsequent analysis become more accurate.
Unsupervised Learning “Clustering” Through clustering, we can build a profile of like entities (health systems, hospitals, nurses, doctors, patients, etc…) based on the observed characteristics. By clustering entities with similar characteristics, Medacist can build sample sets that are more significant, resulting in a more accurate analysis. In addition, by leveraging these entity profiles, we are able to synthetically create sample sets that include the entirety of the Medacist’s client base (“Synthetic Demographic Profiling”). This, in turn, extends the comparative analysis scope beyond the client’s organization, which has a limited statistical significance by nature. Another approach to building a large sample set would be to rely on historical data within the client’s organization’s same scope. Although there is value in trending analysis, not all analysis will benefit from historical sample sets as they take a long time to gather and are volatile as the format of the source data can frequently change. Through our “Synthetic Demographic Profiling” methodology, we can build a sample set in less time, resulting in an overall increase in our clients’ speed in identifying at-risk-individuals.