Medical Coding Automation using NLP on Clinical Data


Healthcare, Medical Coding Automation,Artificial Intelligence


Python ,Angular, TypeScript,REST ,SOAP,Web Services

Client Requirements

The client is a USA-based Healthcare provider, A Healthcare Communication Platform that connects care teams and engages patients in real-time. The client wanted to develop a Medical Encoding System/Feature for their provider network spread across the USA to streamline and automate the billing process, which was previously a tedious and hectic task done by people manually by meticulously encoding each keyword in the SOAP Notes.
The requirements were to develop a billing assistant who will provide the ability to detect and review procedures, treatments, and their respective CPTs from the transcript. The Billing Assistant, wherein a healthcare provider records what procedure is performed on a patient. Based on these recordings, the goal was to identify which procedures were performed and their respective CPT codes used to generate bills for the patient.

Thinkitive Solution

Thinkitive’s business analysts and subject matter experts started with the requirement analysis (Discovery) phase. The thinkitive team took multiple calls with the client and created detailed SRS requirement documents and low-fidelity wireframes. Thinkitive established a team of professionals with Machine Learning and Data Science expertise. The Thinkitive team leveraged AWS event-driven architecture, which will be integrated with their platform. Below is a high-level diagram of the system.

Architecture Diagram for Medical Coding Automation using NLP on Clinical Data

Solution Highlights

  • Speech to Text:

    Providers would need to write a transcript and provide respective CPT for the procedures mentioned in the transcript. Writing Transcripts will always Providers would need to write a transcript and provide respective CPT for the procedures mentioned in the transcript. Writing Transcripts will always have a margin of error when done manually. To make it easier for health providers, the system allows the provider to record the conversation between provider and patient, which is converted into text using speech to text model.

  • CPT Prediction:

    Once the text is generated, the model recognizes Clinical Procedures and Treatments from the Transcribed text. CPTs are predicted with respective Clinical Procedures and Treatments. Providers can also review and edit the predicted CPTs.

  • Model Training on New Data:

    The platform provides the option of periodic training when the user feels there's a need to retrain the model, where the provider can select the data to be sent to AI from a billing assistant dashboard web app. Once the data is sent to the model, the training process in SageMaker is initiated.

Value Delivered

Enable patients to find the best specialist available across the entire U.S. Increase provider efficiency to take secure and HIPAA-Compliant Video consultations without physical presence. Patient access to a provider, treatment notes sent back to the referring provider, shared screens and/or files during a virtual consult, secure and HIPAA compliant and more. The provider doesn’t need to manage the EHR; the provider can take notes and push to the EHR with a single click.

  • Medical data is confusing and hard to understand, which makes it difficult and causes many errors while generating bills manually. Billing Assistant has made it easier for providers and billers.

  • Replaced lengthy and time-consuming methods with time-saving and efficient solutions.

  • Reduced margin of error, which used to occur while generating bills manually.

  • Since the billing assistant is reducing the margin of error, it is also contributing to less number of declined insurance claims.