Inspiration
Studies report that the most effect interventions do not lead to substantial improvements in adherence and outcomes because most interventions are too complex and are not tailored to the individual's needs. Thus, we introduce MARVL (Medication Adherence Risk eVaLuation), a pharmacist's best friend to pair the most appropriate adherence intervention with the patient.
What it does
We trained a model using a large epidemiological survey to understand medication adherence from more than 1,000 patients. The data set includes more than 100 important individual-level characteristics such as age, sex, race, and health insurance. We developed machine learning algorithms to predict the patient's risk and possible reasons of non-adherence using variables that were clinically relevant and most predictive of medication adherence. Then we used the risk score and the reasons for non-adherence to recommend personalized solutions.
MARVL will help pharmacists identify at-risk patients upon picking up their medications, and provide personalized strategies to improve non-adherence, which will ultimately improve health outcomes and reduce the economic burden of medication non-adherence.
How we built it
The publicly available data was obtained from Roper Center from Cornell University. The model was trained in R and Python using a machine learning ensemble method including neural networks. The user interface was built with R Shiny.
Accomplishments that we're proud of
We have a prediction accuracy of 93.52% for medication non-adherence.
Log in or sign up for Devpost to join the conversation.