Model Specification starts from a specification document and sets out the model it describes. You then confirm that the model described is the model you intended to build.
The researcher uploads the document and starts the analysis. A clock shows the run time.
While it runs, the screen shows the time elapsed and the steps being completed. The system reads the document, then the structure of the model and its clinical events, and stops to wait for review of the model's setup.
The run lists each setting as read. A setting the document does not give is marked not stated, and the random seed is marked as set by the system.
Items that need input before the run can continue are listed, and further items are listed for review.
The researcher answers a question in plain words. The system proposes a change and shows it, and nothing is changed until the researcher applies it.
Every entry the researcher makes is listed under Changes you made.
After approval the run continues. It extracts disease progression, mortality, treatment effects, treatment switching, health-state utilities and costs, prepares the narrative and the schematic, and loads the review materials.
The narrative restates the model in plain language, in eight sections from the model description to utilities.
The schematic shows the arms and lines, and the structure map shows the health states, starting with the state where the cohort enters.
The second review has three groups. Needs Resolution blocks the build, Review items do not, and Confirmed holds what already matched. Downloads stay locked until the items that need resolution are resolved.
Before downloading, the researcher checks what the model will do. The review has seven chapters, written by code from the saved tables.
Each value carries a tag that shows where it came from, such as Entered by you, Placeholder, or Set by the system. Questions are listed under Worth checking.
The researcher confirms that this is what was intended. Downloads open for this version, and the approval is recorded with the reviewer name and the date.
The Model Specifications document lists every value with a status. The legend defines the statuses, among them stated, derived, entered, not stated, set by the system and placeholder.
The researcher's own answers are in the document, marked entered, with the name and the date. Values the document does not give stay marked.
Eight downloads are offered: the narrative, the model specifications, the review report, the assumptions, the missing data, the interaction record, the schematic as an image, and the files for the compiler module, which build the model.
Next, Coding builds the confirmed model in R and Excel.