Challenge

The medical service provider performed medical coding of diagnosis for pathology data manually. The accuracy of the ICD–10 coding process affected the payor reimbursement rates. The provider also found it challenging to properly utilize medical coding staff for high-value activities. 

Using Microsoft Azure services, Neal Analytics developed an ensemble of candidate machine learning models capable of classifying diagnosis from the text. A model integrated with diagnosis text, clinical history, code set, and historical labeled data allowed coding addressing automatically. Neal Analytics used text vectorization and augmented feature engineering to translate and process natural language, creating a friendly user interface that requires no training to use. 

Model integration has resulted in increased ICD–10 coding accuracy from manual. The automated claims submission with ICD–10 resulted in better reimbursement and payor reporting.