Repository navigation
Releases
Releases · centralstatz/readmit
readmit 1.1.0
- Canonical production now rejects discharge episodes with
discharge_time > as_of_timebefore eligibility, risk-provider calls, or
history writes. A discharge exactly atas_of_timeremains valid. - Canonical terminal events now require only
terminal_event_id,episode_id,
event_type, andoccurred_at. Additional columns remain permitted.
Eligibility usesoccurred_at <= as_of_time; an event-availability timestamp
is no longer required or interpreted. - Newly initialized assessment history and
rrp_inspect_history()retain
selected terminal-event identity, type, and occurrence time without
terminal_available_at.
readmit 1.0.0
readmitretains its hospital-readmissions focus and evolves from the
exploratory utilities in 0.0.1 into the Readmission Risk Pool (RRP), a
project-based framework for repeated readmission risk assessment. This
revision intentionally changes the public API and moves from MIT to Apache
License 2.0.rrp_project_init()creates a user-owned project with editable configuration,
source-mapping and risk-provider scripts, branding, and a Shiny application.
Project authors supply their own clinical data mapping and risk logic;
rrp_project_validate()checks the authoring structure, and
rrp_state_init()establishes persistent history.rrp_run()assesses the remaining risk of a first readmission through 30
elapsed days after discharge at a chosen analytical time. It applies shared
eligibility rules to project-supplied canonical data and retains accepted
estimates, provider failures, and ineligible outcomes. Use
rrp_project_status()andrrp_inspect_history()to examine the results.rrp_refresh_products()derives current remaining risk, recorded risk
trajectories, and an operational scope summary as CSV files from retained
history. The project-owned Shiny app opens with empty views before products
exist and displays them after refresh.- A short installed getting-started vignette and a reproducible website
implementation guide demonstrate the project workflow with illustrative
data and risk logic. No clinically validated prediction model is included.