Cityblock Health is a community-centered healthcare company that focuses on delivering personalized medical and social care to low-income neighborhoods. In addition to providing care in a clinical setting, we improve outcomes by addressing social determinants of health. SDoH are non-clinical, environmental factors that influence access to healthcare resources, the likelihood of developing chronic diseases, and overall clinical outcomes for individuals and communities. The SDoH that impact patient care are often under-represented, in part because they are primarily captured within free-text unstructured data sources such as clinical notes.
A major driver of health outcomes is access to stable housing, so we built a housing classification model to identify members in need of housing and enable our Community Health Partners to work with our members to address that need. We used Spark NLP for Healthcare to process raw text data into sentence-level embeddings; used active learning to manually annotate our data; and deep learning with TensorFlow to build the classification model.
In this talk, we will address some of the nuances of implementing our model, present our findings, and discuss how we are using our model results to help our members.
