Our client is a global leading petcare service provider headquartered in North America operating over 1000 veterinary hospitals and diagnostic centers in US and Canada. They are committed to providing advanced pet care services for the world at large.
A typical day for a veterinarian starts with early-day surgery procedures if any, to allow the pet for post-op recovery procedures throughout the day. The next activity of the day is to examine outpatient pets and emergency walk-ins.
A critical step to be performed during all these activities is to capture key details of the patient and store them digitally which is commonly referred to as Electronic Medical Records (EMR). An EMR has multiple fields which include but are not limited to pet species, breed, age, weight, immunization and vaccination history, past and current diagnoses, allergies and immunogens, etc.
Filling up the details in EMR is a laborious task but every doctor/clinician has to do it religiously. In our client’s case, each veterinarian took almost 20 minutes to complete filling in the details in the PMS per examination. The aggregate time spent on filling EMR details mounted up to consuming at least 50% of veterinarians’ time in a day.
The challenge for the client was bifold.
When the client approached purpleSlate with their challenges, we understood that automating EMR entry in the PMS is the way forward. How to automate data entry into the PMS? Unlike human healthcare, in veterinary care, the pets need to be in constant hold of the veterinarian throughout the examination. So any tool which requires them to use a hand when they were examining the pet, was out of the equation. This is where a Voice solution came in as a blessing.
We at purpleSlate designed a solution where doctors can just talk to the system. The key intent and medical terminologies from their speech will be automatically extracted and will be parsed into the relevant fields in the PMS.
But there was a catch. The system needed to understand complex medical terminologies and more importantly parse the information to the right fields. We achieved a 90%+ accuracy level on voice recognition, and Natural Language Understanding (NLU) of the medical terminologies and language along with their appropriate hierarchy were achieved with rigorous training of the ML models on entity recognition and intent identification. We also trained the models on an extensive corpus of user utterances, including complex medical vocabulary, combined with retraining workflows for the continuous maturity of the ML models. This helped us manage the challenges around accuracy.
And we proved it right with our
purpleSlate offers a host of technology service offerings, with AI and NLP as the crux of its solutions, to solve complex business problems. As machines become increasingly proficient in natural language, conversations, in voice and text will transform into the primary mode of providing and consuming services.