Problem to solve: Coding statistic model characteristics is unnecessarily complicated by having a codeable concept for the model characteristic and a separate codeable concept for the method of addressing that characteristic.

Use case and best practices: Coding a meta-analysis using the Paule-Mandel method of Tau estimation. Code the Paule-Mandel method was used for Tau estimation. It is obvious in this coding that tau was being estimated (so no need to separately code that Tau estimation was done)

Importance and tradeoffs: Reporting statistical models is often incomplete. At times the misapplication or misinterpretation of the statistical model can lead to significantly different results. We are not aware of a scaled system for clear communication of statistical models and creating a standard for this electronic data exchange is also establishing such a system. This proposed FHIR change simplifies the resource with no obvious negative consequences. There has only been 1 instance of Evidence Resources using the more complicated model so far and it can be changed easily, so backwards compatibility is not a concern.

Changes to be made:

1. Statistic.modelCharacteristic.value[x], remove valueCodeableConcept option
2. Statistic.modelCharacteristic.code becomes 1..1
3. Add 56 codes to the StatisticModelCode code system
   • The display names will be derived from this spreadsheet: https://docs.google.com/document/d/1YWmkUFPUYOfpOaRhrqkcDJ7wg2HFSkMO/edit
   • We will come up with codes and definitions (or placeholder text) for these entries in the Statistic Model Code System Development Steering Group meeting on November 2nd.