LoginSign Up
← Back to all sparks
C

cfr

ANALYTICS
Velocity0.0

R package to estimate disease severity and under-reporting in real-time, accounting for reporting delays in epidemic time-series

epiverse-trace.github.io/cfr ↗

cfr packaged delay-corrected severity estimation, then went quiet on maintenance.

epidemiologyseverity-estimationoutbreak-analyticsepiverser-packages
◆Current state
The package estimates disease severity and case ascertainment while correcting for the delay between a case being reported and its outcome being known. After the 0.1.1 rework of the estimation internals and a maintainer handover to Adam Kucharski, activity dropped to a vignette and an R-devel compatibility patch in February 2025.
◆Where it's heading
The methodological work is done and consolidated: likelihood approximation is now selected automatically from outbreak size and an initial severity estimate, and the internals were renamed with a dot prefix to close the public surface down to cfr_static(), cfr_rolling() and the data-preparation generic. Releases since have been documentation and compatibility only.
◆Prediction
The 0.1.0 notes flagged time-varying ascertainment as future work and it has not appeared in the two releases since; nothing in these entries indicates when or whether it lands.

◆Recent moves

  1. 1y ago

    R-devel difftime patch and an individual-data vignette

    A compatibility release: unique() on difftime wrapped in as.numeric() for R 4.5.0, an epiparameter class-name documentation update, and a new vignette on estimating from individual-level data. The estimators themselves are untouched.

    View source ↗
  2. 2y ago

    Severity estimator picks its own likelihood approximation

    The estimation internals were reworked: .estimate_severity() now selects between binomial, Poisson and normal likelihood approximations based on outbreak size and an initial severity estimate, and output columns were renamed from severity_mean to severity_estimate. Existing workflows break on the rename, and CFR values can shift slightly against 0.1.0.

    View source ↗
  3. 2y ago

    Delay-corrected severity and ascertainment estimation on CRAN

    ⚡ SPARK

    The initial release, packaging static, expanding-window and time-varying severity estimation with correction for reporting-to-outcome delays, plus static ascertainment against a baseline severity. It ships with 1976 Ebola and Covid-19 example data and an incidence2 method.

    View source ↗