To extract the text from the billing records & patients’ charts and identify anomalies.
A US based company handling medical billing records wanted to optimize the process, specifically related to billing errors (billing done for wrong medical issue), discover out-of-network instances (patient who met with an accident was sent to a network hospital but treated & billed by an out of network doctor) and detect frauds (Too many doctors referring to same pharmacies). Client wanted us to build an application that can amplify the detection capabilities and detect inconsistencies.
During the initial phase, lot of challenges were foreseen such as text extraction tables as well as from plain documents, dynamic data elements which can be added to the patient history etc. Our scope was to create an application that can extract the text from the billing records, patients’ charts and create a network of events using graphs networks.
The core of our solution lies in the Network Graphs that can create relationships between multiple entities. Anomalies were identified which are typically out-of-network elements and are shared with the concerned team for further investigation.
This application had helped in the improvement of detection capabilities by 85%. It also helped in addressing multiple questions & use cases and could be easily scaled up, based on the requirements. Graph databases as a core architecture gave the flexibility to the client, in adding new data columns with database changes and queries.