Project Type: Train Natural Language Processing (NLP) Model
Industry: Digital Healthcare
The client provides software to Skilled Nursing Homes to reduce costs and find new revenue opportunities. As reimbursement rules become more convoluted, our client helps optimize revenue, control costs, improve efficiency and deliver insightful business intelligence using its state of the art AI and Machine Learning Platform.
The co-founders had years of experience in post-acute care and built a successful software platform for nursing homes to manage rising costs and analyze information from nursing facilities. As the company grew, the founders realized that as reimbursement models become increasingly more complex, small and mid-sized nursing facilities would be overburdened with administrative work and wouldn’t be focusing on what’s important: helping their patients. The client needed to figure out a way to pull unstructured data from EHR (Electronic Health Record) Systems and classify and identify them so that nursing homes were qualifying for the correct insurance contracts. They knew they needed NLP, but didn’t know how to implement it.
Skiplist understands how to create simple software solutions to complex problems by building Thoughtful Software. The team at Skiplist suggested a combination of managed software and custom solutions to make this happen. Skiplist deployed AWS Comprehend and built a custom tool to use training data to train the NLP model to classify what data would be matched with which rules to find missed revenue opportunities. Skiplist taught the client's team to discern the differences between raw data and processed data, and demonstrated the importance of having good training models to build successful NLP models.
The Skiplist team achieved a 90%+ accuracy model for its data quality using the NLP model they built. This was a significant improvement over other solutions they looked at such as MonkeyLearn. The client now has a powerful back-end engine that delivers actionable insights for its customers, but with a simple front-end so that it maintains a straightforward user experience