Data-driven healthcare company
Using data science to analyze healthcare provider free-text notes
Overview
THE CHALLENGE
Our innovative data-driven healthcare client wanted the ability to quickly analyze healthcare provider free-text task notes to spot new opportunities for improving patient care.
THE SOLUTION
Ollion used custom built text processing algorithms to group more than 150,000 free-text healthcare tasks into 11 task groups to create this advanced healthcare analytics solution.
THE OUTCOME
The client is now able to identify and analyze these task groups according to their category without any additional efforts from the provider, patient, or data user.
The challenge
Our data-driven healthcare client wanted to easily analyze the more than 150,000 task notes their healthcare workers took when visiting patients. The challenge our client faced was that the descriptions of these tasks were all free-text, so analysis was nearly impossible.
Their hope was to find a way to automatically identify and categorize free-text notes into task groups to be able to spot trends, uncover issues, and ultimately provide better care.
Ollion had previously helped this client build out a modern data platform and enable internal and external reporting. Combined with Ollion’s deep healthcare and data science expertise, partnering with Ollion for this advanced healthcare analytics project was a no-brainer.
The solution
Ollion created a Latent Dirichlet allocation model in Python to analyze all the free-form task notes from our client. This model goes through each task and assigns it a group based on patterns and key words found within the text for the task. Thus, similar tasks are expected to be assigned to the same group, while different tasks will be in different groups. Ollion then went through each of the groups with our client to assign a group name to each of the 11 groups created.
Ollion ran the model in Python through Apache Airflow and was able to leverage the existing data warehouse in Snowflake to assign new tasks as they came in.
The outcome
With this advanced healthcare analytics and data science solution, the client is able to analyze the data and metrics from these free-form text notes through these newly formed task groups. They can now spot trends and find opportunities to improve patient care quickly and easily, without having to manually read through or categorize notes. The model also automatically categorizes new notes as they come in, so the client always has the most up-to-date information.
