We built a Natural Language Processing (NLP) system for helping the HMA doctors to record the patient information as they speak in real time, saving several million per annum in medical transcriptions.
Our client is a Fortune 500 Hospital chain, which has centers across 80 different locations. They were spending more than $2 Million in preparing medical transcription every year.
The client was losing on both time and money. The Doctor used to record every patient’s conversation, which then went to a medical transcription company. The whole process was costly, time-consuming and error-prone. The client wanted an elegant but smart mobile application, which would not just work as a voice to text converter but also would be able to do Real-Time categorization of doctors’ notes into different categories.
We used Nuance to record the patient info as they speak in real time and categorize them into EMR leveraging IBM Watson. The engine would take the input data, analyze the information and the final EHR document would be generated.
- The Nuance engine was trained to medical-specific terms. As the doctor dictates the text, the nuance engine would convert the text in real time into categories like Vitals, Pre-existing ailments, Current medication, Allergies and other categories
- The engine would use the pre-saved templates from the EMR which would control the data that would be displayed. The business rules were configured in the templates
- The NLP annotator will look for specific tags to define templates and provide templates found
- The nuance engine was trained to identify the incomplete sections and would be marked for the doctor to fill in the missing vital information
- The Nuance engine was also trained to identify conflicts if present in the dictated notes
- The prev history of the patient would be visible to the doctor if the patient was treated by the doctor earlier
- The dictated text could be saved as a draft or converted into an EMR
- IBM watson
HashedIn wowed the client with a feature enriched product. Some of the impacts were:
- Automated understanding of physician narratives reduced the time and effort to document encounter notes with discrete data
- Improved quality of overall encounter note through real-time extraction of, and feedback on, discrete data in free-form text
- Reduced time to create a complete record – estimated 1+ hour per day
- Improved accuracy of final note – time lag and rework queues
- Saved several millions per annum on operational and labor cost through automation