Role:
HCI Research Scientist & Data Scientist.
Collaborations:
Development team, researchers, and subject matter experts.
Development team, researchers, and subject matter experts.
Skills:
User-centered design, visual analytics, and infovis.
End-to-end process:
Interviews, Exploratory Data Analysis, Social Network Analysis, Sketching, Validation, Information Visualization, and creation of metrics.
Industry:
Healthcare insurance.
Research problem:
How to identify physicians that excel in their activities?
How to identify underlying connections among physicians and patients?
Final solution:
A set of metrics to identify related to multiple aspects related to physicians activity.
Related publications:
Initial exploration of claims database allowed the identification of visual patterns revealing mutual relationships in the physician-physician network.
This network was designed to show connections between pairs of physicians (nodes). Hence, physicians A and B are connected (edges) if a patient visited first A and after that visited physician B.
After exploring connections' attributes and validating our analyses with subject matter experts, we proposed a metric to reveal strong mutual referral relationships in the physician-physician network.
The metric allowed the analysis of how these networks differ considering location and health care service offerings.
Comparing the mutual referral metric (left) with PageRank (right) for the same physician-physician network, it is possible to see that the mutual referral highlights pairs of strongly connected physicians, even in the borders of the network. On the other hand, PageRank highlights high degree nodes and its connections.
The data analysis project resulted in set of metrics and mutual referral is just one of them. At the end, our team showed how client's analytics team could replicate and apply the proposed set of metrics in their daily activity, identifying physicians that excel and their underlying network connections.
