Creative Practice Solutions: AI-Driven Medical Transcription & Billing System
“Neurons Lab has actually been really terrific and helpful, so we are excited to move forward with them”.
Matt Dallmann, President @Creative Practice Solutions
Platforms: iPhone, iPad, Maс
About the Client
The client is a healthcare business consulting group working across the USA. Creative Practice Solutions (CPS) has over 30 years of industry experience.
The CPS team is eager to help improve the functionality and efficiency of both private and public medical establishments through newly adopted digital solutions. In addition, the company helps with billing protocol development, billing services, personalized consulting, and audits.
The medical sector generally includes many issues that can be resolved via innovative technologies like artificial intelligence and machine learning.
The client mentioned that in the medical consulting area, many things require specific reimbursement from commercial insurance organizations. For example, a patient can be denied compensation from the insurance provider if even one or two words in a medical note are missing or misused.
As a result, the Creative Practice Solutions team concluded that the following issues are constantly faced in the healthcare administration domain.
- Medical documentation and coding are performed across a wide variety of office settings making a universal solution difficult
- Current EMR requirements contribute to physician burnout
- Limited budget and time resources
- Lack of historical data
- Non-technical target audience
- Concerns about maintaining HIPAA compliance
One of the areas that continues to be problematic is the creation of medical notes and proper billing terms. Here, medical providers spend a lot of time manually preparing notes and assigning billing codes for invoices, so they need to find a solution to reduce their administrative burden.
Here is how the service flow operated initially:
Step 1: A patient comes to the doctor’s office
Step 2: Appointment is performed and findings are manually dictated or entered into a note/document
Step 3: Medical billing coder extracts diagnosis, procedures, and medications
Step 4: This expert transforms them into codes and applies billing rules
Step 5: Results are reviewed, refined, and sent by the doctor to the insurance company
The idea of the SmartNoteMD project is to effectively organize medical note taking and make the reimbursement processes smoother and faster.
“Despite the many features of current EMR systems, we consistently hear how medical providers are 30-60 notes behind schedule and have to spend their weekends catching up. As a solution, we decided to create an application using AI, machine learning and listening technology to automatically create medical documentation with suggested billing codes in real time. This allows providers to cut down on labor, while increasing revenue through fully supported medical notes”.
Matt Dallmann, President @Creative Practice Solutions
The project aims to simplify the process and save time on documentation for healthcare providers and practitioners.
During the patient examination, the visit is recorded with subsequent automatic processing of the text and the extraction of information about the patient from the text, such as:
- Meta information like height, weight, name, date of birth, etc.
- Diagnoses, allergies, and other individual symptoms
- CPT codes (Current Procedural Terminology) and billing codes
In the long run, the trained AI-empowered system allows users to receive information from the doctor’s visit in real-time, within just 10-20 minutes.
This helps hospitals and practitioners see more patients, resulting in additional revenue. In return, this can be positioned as a great recruitment initiative for specific organizations to attract top physicians and nurse practitioners.
Furthermore, the proper medical note submission to the insurance providers will also increase medical billing efficiency. Accuracy in this matter is essential, especially in the USA. Incorrectly filled insurance claims may result in revenue loss, huge penalties, and lawsuits.
Proposed AI-Based Solution
The solution offered a new way of documenting medical notes and billing codes.
With the help of AI/ML in line with AWS assistance, the program automatically creates a medical note with billing codes in real-time. In addition, the software uses iOS technology to convert handwriting to text and transfer other medical records directly into application fields via the device camera.
The solution was designed to meet the following requirements:
- Create a HIPPA- compliant solution with an enabled audio recording that allows healthcare providers and practitioners to take notes with structured entities
- Attain a high level of accuracy in recognizing specific medical terms in speech and text
- Complete the solution’s launch on 4 operating systems (iOS, iPadOS, macOS) without assembling 4 distinct development teams
- Allow secure medical note exportation in different formats to match with existing billing systems
- Ability to collect medical data securely and store it for further data science operations
- Guarantee that there will not be extra system administration expenditures due to automated serverless infrastructure
- Help patients and doctors meet one-on-one and remotely as if they were in-person
- Provide the modularity of the solution where any new features will not affect the existing ones
In addition, an OpenAPI is being created for integration with clinics to make it possible to automatically enter the information received into the clinic’s EHR system.
As of 2022, the project has been under development for the last two years with Neurons Lab. The project was handed over at the stage when the application and the infrastructure as a whole were ready and working but needed some substantial improvements.
Services Provided by Neurons Lab
A reasonably large system had been developed before Neurons Lab started working on the project. Our strong point was a change to a different approach to the work, which allowed us to simplify the system’s architecture throughout further development.
As well, since the project is large, we had to prioritize development in such a way as to move from small tasks to large ones to spend minimal time immersing ourselves in the operation of the entire system as a whole.
The initial client’s requirements were to continue the backlog of the previous team as well as:
- Performance optimization and improvement
- Development of an OpenAPI for integration with clinics
- Improvement of information processing and interaction with ML components
- Creation of documentation; since the client is not “technical,” it was necessary to describe the documentation for the API, etc.
Utilization of AWS Technologies
As an AWS partner of Advanced Tier, Neurons Lab created a complete serverless solution with a microservices architecture based on AWS managed services.
The current system uses different component, namely:
- The API layer is powered by Amazon API Gateway
- The backend system is based on AWS Lambda functions
- Data layer is based on Amazon Aurora Serverless and Amazon S3
- Amazon API Gateway is integrated with Amazon Cognito, which handles user management and authentication
- ML components for medical note creation are implemented with Amazon Transcribe, Amazon Transcribe Medical, Amazon Translate
As well, Amazon API Gateway is secured with AWS WAF, which is a web application firewall that helps protect your web applications or APIs against common web exploits and bots that may affect availability.
Below, you can check out the scheme presenting the high level project architecture.
How SmartNoteMD Works Now
Here we briefly describe how the SmartNoteMD system operates in 3 simple steps:
1) The software listens in the background during the patient visit. This can be compared to Alexa by Amazon; the system automatically uses particular keywords and phrases to analyze the conversation and create a medical note.
2) The technologies of AI and ML are used to create relevant content for the medical note. A doctor or practitioner can edit the note, add supplements later, and can sign off on the medical note for increased accuracy.
3) Doctors can pick up a note or patient card, export it, and send it via email, text, or airdrop in one of many possible formats (pdf, xml, json, plist, html, reduced json or reduced xml).
Within the app settings, it is possible to select a specialty to improve app accuracy, and the software will use only the applicable codes per treatment type chosen. For even more accuracy, there is an ability to create a custom set of diagnosis and procedure codes from the library.
The app also has a hands-free function, where the software starts recording once the conversation begins and finishes after a few seconds of silence.
You may access the video to see the software in action that explains all the details.
Development Methodology & Process
Here we explain the development process in greater detail to manage and organize the flow. The project implementation process has been set up according to an agile software development methodology, namely Scrum, with 2-week sprints planned.
The following team has been involved in implementing the SmartNoteMD project:
- Principal Applied Scientist –(similar to Tech Lead) a professional with a related Ph.D. in healthcare, physics, biology, or a similar field and 5-7 years of experience. This is a key expert who guides the team and possesses a deep knowledge of the industry domain and technical expertise.
- Backend developers – engineers who build and maintain server-side web and mobile-based apps and process and perform actions with data storage, security, and server-side service functionality.
- Frontend developers – engineers who create the client-side web and mobile-based solutions that users interact with.
- Cloud Architect – IT professional who is in charge of setting up and implementing the company’s cloud computing strategy, followed by cloud management and monitoring.
- ML Delivery Manager – a professional who knows R&D project development methodology (Agile), is responsible for managing the team, and makes sure that all artifacts are completed as stated.
- Business Developer – a professional who aims to grow the client’s business and increase revenue by developing strategies to attract new customers and engage existing ones.
Initially, we conducted a scoping phase where all the technical and business requirements were collected, reorganized, and prioritized into the SoW (scope of work). This document states all of the minor details to release the project in accordance with the set objectives.]
The team used an asynchronous communication model that allowed them to save substantial time on discussions and instead concentrate on the tasks themselves. In this model, the team used Slack chat to share all the materials, ideas, and issues. Other team members had 24 hours to check those and add their feedback in any format they liked (voice message, video, text format). The client was also aware of this approach and was involved in the process!
After each sprint, a demo meeting was held where all the deliverables and results were shared with the client. Then, the pull of new tasks was discussed and chosen for the new sprint and every task was checked for fit-for-purpose to ensure that the product still aligns with business intentions and goals.
Proper project planning, management, and engineering capabilities allowed Neurons Lab to successfully complete the work on the project in a timely manner and according to the client’s needs.
Ongoing Development & SmartNodeMD Opportunities
At this moment, the SmartNoteMD system is being tested among its users in real-life cases, namely independent practitioners, nursing services, and other medical establishments. The plan is to demonstrate how the system works, test its flow, get real user feedback, and make the necessary adjustments.
The numeric indicator the client is focused on is increased accuracy for medical notes in line with reducing the time spent to produce the results.
The future development and project growth involve the following aspects:
- Improving note recognition process
- Improving UX through integrations with different potential mechanisms (Siri etc.)
- Improving usability by adding features such as voiceless recording control, etc.
- Ongoing business development and product management
It is necessary to ensure that all the data collected during these trials is accumulated, stored, and protected correctly in order to advance the system in the following phases.
In addition to our team being substantially involved in Product Management, we provide subject matter expertise in business as well as Account Support between the technical team and our client. We help optimize business and market strategies to align the product with the business goals in order to maximize revenue growth.
Have an idea for using AI/ML in the healthcare domain?
Contact Neurons Lab, an AWS Advanced Tier Services Partner, for a detailed consultation so we can elaborate on how Big Data, Artificial Intelligence, and Machine Learning can add value to your project. Meet our experts to brainstorm about your product.