Forecasting
Applications
Open Source Projects
Clinical Decision Expert Panel
In 2011, the CDC convened a panel of immunization forecast experts from over a dozen immunization registry projects across the nation with the stated goal to 'improve the clarity, accuracy and consistency of on-going childhood immunization forecasting and evaluation through the development of a framework for use in automated electronic systems and elsewhere." This panel is making great progress.
November 2011
The first face-to-face work group meeting was held in Atlanta. This meeting was focused on developing the logic for supporting the evaluation piece of the clinical decision support project. The forecasting piece was put off for later. During the meeting Nathan Bunker developed a provisional application that demonstrated how this could work. This application allows for viewing the internal "states" as the algorithm steps through the evaluation.
March 2012
The group was reconvened with many of the same participants for a continuation meeting. A lot of progress had been made since November and the basic state model had been enlarged and developed more. During the meeting the following items were discussed in several sub work groups:
- Reviewed the logic in many of the states to verify that all possible scenarios are covered.
- Reviewed and discussed overall architecture and how such a system would be used.
- Continued work to collect and organize test cases.
- Discussed process for integrating this technical work with the work of ACIP.
- Discussed continuing this work after this initial project is completed this year.
On the technical side, Nathan Bunker was able to continue making changes to the demonstration software and showed it the group just before the meeting was adjourned. This version of the software was partially working but not good enough to demonstrate fully but now forms the basis of the CDS Forecasting Framework.
Current State Diagram
This diagram shows the discrete program states that the CDS work group defined to perform both evaluation and forecast. The previous state diagram in 2011 did not affectively address forecast and other issues. Each white rectangle indicates a state that of the process. The user can step through the forecast state-by-state and verify that the process is working correctly. This ability is particularly useful as we continue to refine the model and the process.
This diagram was completed on the first day of the meeting and by the third day it was apparent that there would again be more changes to this diagram. Nathan Bunker will be updating this diagram in April with the latest changes made by the group. Once this diagram is updated a second version of the software will be released with the basic infrastructure in place.