Development of An Immunization Forecasting System: Software & Design

Background:  Immunization forecasting systems are automated software algorithms used to assess patient immunization status and recommend the dates and vaccines that should be administered next.  A new forecasting system was needed at Texas Children’s Hospital (TCH) because the existing off-the-shelf product was no longer maintained and produced inaccurate vaccine forecasts.

Setting:  A new forecasting system was developed collaboratively by a physician, immunization experts, nurse, and software engineer. 

Population:  N/A

Project Description:  To improve the accuracy of immunization recommendations within TCH’s EMR, a forecasting system was developed. The algorithm was built by project staff using no proprietary applications.

Results/Lessons Learned:  A meta data model was created to translate the multitude of ACIP vaccine recommendations into a series of rules and permitted values for each rule. The model was stored in Excel spreadsheets for easy review and editing. A rules engine was built to read the meta data model and create a vaccine forecast. The forecasting process was broken into three steps: (1) evaluation of already administered vaccinations, (2) recommendations of doses required to complete each vaccine series, and (3) final processing of steps 1 and 2 to create an easy-to-read vaccine forecast.  The first step indicated if doses were valid, invalid or missed, while step 2 provided information about the next dose, including dose number, date valid, date due, and date overdue. Step 3 prepared the forecast for display by adding a description, such as “due now” or “overdue,” and label for each recommendation based on patient age (e.g. “DTaP” vs “Tdap”). A description of the rules (ACIP recommendations) applied was also provided for each vaccine recommendation.  Contrary to our expectation, creating the rules engine was the least time-intensive step in development while the meta data model represented the greatest amount of effort.  This team-based approach utilizing a meta data model produced an accurate immunization forecasting system.