Healthcare data is messy. Multiple formats, unstandardized fields, missing values, errors, and inconsistencies lead data science teams to ignore patients and predictors or risk spending all their time on data prep. Cornerstone Enterprise learns the structure and connections across data sources, harmonizing the structure. Proprietary ML models then identify and correct errors, which are applied automatically or guided through our UI for the optimal balance of scalability and nuance.

Algorithms learn the structure and relationships for any dataset and create a harmonized schema optimized for analysis. Automatic detection of date and visit columns enables cross-table merging and stacking. No manual configuration is required. Many issues are hidden in the data complexity - but become clear when viewed in just the right context. We find that context and display it automatically for you to adjudicate. Some of the richest data in an EHR is in text form, but it’s also the most time-consuming to prepare. Our built-in algorithms give you a head start, and let you refine the predictions quickly through an intuitive interface. Lab values in different units, varying date formats, and more are also harmonized automatically. Intelligently fill in missing data with imputation methods specialized by data type for efficient model building. Extract the data exactly as you need it for the next step: clean, harmonized tables, with a regulatory-grade audit trail. 

As data scientists, we want the power of intelligent automation without giving up fine-grained control of the process when needed. Cornerstone’s platform is built to give you both. Self-learning algorithms do much of the work, but the interface allows investigation and cleaning down to the level of a specific patient, record, or value. The models can then be applied automatically to new data as it comes in, via API calls that fit into your data pipeline.