Add Pancreas Donor Risk Index (PDRI)

Computes PDRI and appends it to df (default column name: "PDRI"). If df is grouped (dplyr), optional mean-imputation is performed within groups. Original columns are never modified.

Usage

add_pdri(
  df,
  variant = NULL,
  mean_impute = FALSE,
  pdri_col = "PDRI",
  verbose = FALSE,
  return_terms = FALSE
)

Arguments

df

A data frame (may be a dplyr grouped data frame).

variant

NULL (default; auto-detect), or "PA" (pancreas-alone), or "KP" (kidney-pancreas).

mean_impute

Logical. If TRUE, mean-impute missing numeric inputs within the current group (or overall if ungrouped) for the calculation only. For indicator terms, NA is treated as 0 only when mean_impute = TRUE; otherwise NA propagates to PDRI.

pdri_col

Name of the output column. Default "PDRI".

verbose

If TRUE, print inference and sanity messages.

return_terms

Logical. If TRUE, also attaches a list-column PDRI_terms with per-row linear predictor contributions (for QA/testing).

Value

df with a new pdri_col column; grouping is preserved.

Details

Required base columns: DON_GENDER, DON_AGE_IN_MONTHS, DON_CREAT, DON_RACE, DON_HGT_CM, DON_COD_DON_STROKE, REC_PA_PRESERV_TM, DON_NON_HR_BEAT If DON_BMI is missing, it is computed from DON_WGT_KG and DON_HGT_CM. PA additionally uses REC_PREV_KI for an interaction term; if absent, the interaction is skipped (equivalent to KP behavior).

Examples

if (FALSE) {
# Default: auto-detect PA vs KP
kp |>
  dplyr::group_by(REC_TX_YEAR) |>
  add_pdri(mean_impute = TRUE)

# Force PA
pa |> add_pdri(variant = "PA")

# Inspect term breakdown (QA)
# tmp <- kp |> dplyr::slice(1:3)
# res <- add_pdri(tmp, return_terms = TRUE)
# res$PDRI_terms[[1]]  # additive components for the first row
}