Visual Analytics Kit for Healthcare - overview
Currently real world data (RWD) is playing an increasing role in health care decision making. RWD refers to data collected through healthcare delivery including electronic healthcare records (EHR), insurance claims, pharmacy records, and registries. RWD also includes data made available by patients themselves such as from medical devices, m-health applications and surveys. When analyzed RWD can become real world evidence (RWE), or evidence supporting healthcare decisions not based on a randomized clinical trial. For example, life science companies leverage RWD to assess market uptake of new drugs as well as to support clinical trial design or monitor post market efficacy and safety.
Analysis of RWD has traditionally been done by highly trained programmers, data scientists and biostatisticians with results presented to the end user in a static report. Recent advances in computing, user interface design, and visual analytic methods now allow end users to directly interact with the data in dynamic ways that circumvent the bottleneck of data analysis. Visual Analytics is the science of discovering knowledge in large or complex datasets by combining automated techniques of information extraction, machine learning and data mining with human perceptual and cognitive capabilities.
Visual Analytics Kit for Healthcare provides innovative visualizations for healthcare (e.g. pathway analytics) and integrates with Jupyter Notebook. It provides data scientists with data discovery and insights through interactive visualization features for healthcare specific data types and analysis needs. The toolkit provides a convenient native APIs for various analytical kernels (e.g. Python and R) that enables the user to tightly integrate data mining and visualization into an iterative exploratory data analysis activity. The visualization is created on-the-fly and rendered as an interactive chart within the Jupyter Notebook user interface. The toolkit include healthcare specific visualization and data mining algorithms that enable exploration of complex healthcare data as part of exploratory data analysis tasks.
The toolkit can also be used to build interactive web-based applications for domain users, such as physicians for exploring clincial pathways from historical patient records. A recent solution - Watson Health Pathway Explorer for Traumatology – has been developed and deployed in a large hostpital in Europe to allow doctors to extract insights from historical Polytrauma patient records. The solution includes the following main components: (1) interactive similarity filtering components allowing doctors to dynamically explore and extract information about precision cohort of interest; (2) an interactive pathway visualization component showing the distribution in patient pathways and ratios to outcome variables of interest; (3) a longitudinal visualization component showing lab information of selected patients; and (4) an insight recording mechanism to capture and share insights from the data.