Building a flu predictor model for improved patient care

Big data is driving new discoveries in the healthcare industry, allowing medical centers to better understand patients, address illnesses at the onset, and diminish outbreaks. At Kaiser Permanente, one such initiative is focused on the flu, a common virus that can have damaging effects. Each year, direct medical costs from the flu average $10.4 billion, with as many as 49,000 flu-related deaths in the US alone. Jennie Shin explains how Kaiser Permanente developed a sophisticated flu predictor model to better determine where resources were needed and how to reduce outbreaks.

To better treat its patients, Kaiser needed informed data about its resources and an estimation of the required care at each medical center. Kaiser sought to understand its Tamiflu inventory in relation to demand, the effectiveness of its marketing campaigns, and where to appropriately mobilize resources for medical centers with particularly high flu outbreaks. However, working with healthcare data comes with its fair share of challenges, and Kaiser was no exception. Typically, healthcare data is scattered throughout the organization, residing in different departments and different source systems, such as EMRs or HR software, and also comes in structured and unstructured formats. For Kaiser, the biggest challenge to answering its questions was the diversity of data involved and the volume when Kaiser wanted to expand its existing algorithm to include patient data from the entire Southern California organization. Jennie covers the process of building a sophisticated pipeline to ingest and transform a variety of data on a weekly basis for up-to-date forecasting, shares the end results, and explains how this initiative has impacted Kaiser’s preparedness for flu season.