r-kfre

Kidney Failure Risk Equation (KFRE) Tools.

Table of Contents

• KFRE Risk Prediction Tools (R)
• Installation
• GitHub (Development)
• Dependencies
• Quick Start
  1. Toy Data
  2. Vectorized Predictions with RiskPredictor
  3. Single-Person Predictions
  4. Add KFRE Risk Columns to a DataFrame
  5. CKD Staging & ESRD Outcome Labels
  6. uPCR → uACR Conversion
  7. Evaluation Metrics (AUC-ROC, AP, Brier…)
  8. Plot ROC / PR Curves
• Running Tests
• API Surface (Exports)
• Notes on Parity with Python
• References
• License

KFRE Risk Prediction Tools (R)

kfre is an R implementation of helpers around the Kidney Failure Risk Equation (KFRE), including:

• Vectorized and per-person KFRE predictions (4-, 6-, and 8-variable models; 2- and 5-year horizons)
• Convenience wrappers to add KFRE risk columns to a data.frame
• uPCR → uACR conversion utility
• CKD stage labeling utilities
• Evaluation helpers (AUC-ROC, Average Precision, Brier, etc.)
• Publication-style ROC and Precision–Recall plots

Installation

install.packages("kfre")

GitHub (Development)

# install.packages("remotes")
remotes::install_github("lshpaner/kfre_r")

Dependencies

Core imports: R6, stats, ggplot2, pROC, precrec Suggested for tests/vignettes: testthat (>= 3.0.0), knitr, rmarkdown

Quick Start

1. Toy data

toy <- data.frame(
  age = c(55, 72),
  sex_txt = c("male", "female"),
  eGFR = c(45, 28),
  uACR = c(120, 800),
  dm = c(1, 0),
  htn = c(1, 1),
  albumin = c(4.2, 3.4),
  phosphorous = c(3.3, 4.6),
  bicarbonate = c(24, 22),
  calcium = c(9.1, 9.8),
  stringsAsFactors = FALSE
)

cols <- list(
  age = "age",
  sex = "sex_txt",
  eGFR = "eGFR",
  uACR = "uACR",
  dm = "dm",
  htn = "htn",
  albumin = "albumin",
  phosphorous = "phosphorous",
  bicarbonate = "bicarbonate",
  calcium = "calcium"
)

2. Vectorized predictions with RiskPredictor

rp <- RiskPredictor$new(df = toy, columns = cols)

# 4-variable KFRE (2-year), North America constants
p4_2y <- rp$predict_kfre(
  years = 2, is_north_american = TRUE,
  use_extra_vars = FALSE, num_vars = 4
)

# 6-variable KFRE (5-year)
p6_5y <- rp$predict_kfre(
  years = 5, is_north_american = TRUE,
  use_extra_vars = TRUE, num_vars = 6
)

# 8-variable KFRE (2-year)
p8_2y <- rp$predict_kfre(
  years = 2, is_north_american = TRUE,
  use_extra_vars = TRUE, num_vars = 8
)

p4_2y
p6_5y
p8_2y

3. Single-person predictions

# Male, 55yo, 2-year risk (4-var)
rp$kfre_person(
  age = 55, is_male = TRUE,
  eGFR = 45, uACR = 120,
  is_north_american = TRUE, years = 2
)

# Female, 72yo, 5-year risk (6-var)
rp$kfre_person(
  age = 72, is_male = FALSE,
  eGFR = 28, uACR = 800,
  is_north_american = TRUE, years = 5,
  dm = 0, htn = 1
)

# Female, 72yo, 2-year risk (8-var)
rp$kfre_person(
  age = 72, is_male = FALSE,
  eGFR = 28, uACR = 800,
  is_north_american = TRUE, years = 2,
  albumin = 3.4, phosphorous = 4.6, bicarbonate = 22, calcium = 9.8
)

4. Add KFRE risk columns to a data.frame

toy_kfre <- add_kfre_risk_col(
  df = toy,
  age_col = "age",
  sex_col = "sex_txt",
  eGFR_col = "eGFR",
  uACR_col = "uACR",
  dm_col = "dm",
  htn_col = "htn",
  albumin_col = "albumin",
  phosphorous_col = "phosphorous",
  bicarbonate_col = "bicarbonate",
  calcium_col = "calcium",
  num_vars = c(4, 6, 8),
  years = c(2, 5),
  is_north_american = TRUE,
  copy = TRUE
)

names(toy_kfre)
head(toy_kfre)
# Adds:
# kfre_4var_2year, kfre_4var_5year,
# kfre_6var_2year, kfre_6var_5year,
# kfre_8var_2year, kfre_8var_5year

5. CKD staging & ESRD outcome labels

# ESRD outcome within 2 years (duration is in days → converted to years)
out <- data.frame(
  eGFR = c(95, 25),
  ESRD_flag = c(1, 1),
  followup_days = c(200, 1000)
)

out <- class_esrd_outcome(
  df = out,
  col = "ESRD_flag",
  years = 2,
  duration_col = "followup_days",
  prefix = "esrd",
  create_years_col = TRUE
)
# Adds: ESRD_duration_years and esrd_2_year_outcome

# CKD stage labels
out <- class_ckd_stages(
  df = out,
  egfr_col = "eGFR",
  stage_col = "stage",
  combined_stage_col = "stage_combined"
)

table(out$stage)
table(out$stage_combined)

6. uPCR → uACR conversion

df_pcr <- data.frame(
  sex = c("female","male","female"),
  dm  = c(1,0,1),
  htn = c(1,1,0),
  pcr = c(150, 600, 50)
)

acr <- upcr_uacr(
  df_pcr,
  sex_col = "sex",
  diabetes_col = "dm",
  hypertension_col = "htn",
  upcr_col = "pcr",
  female_str = "female"
)

acr

7. Evaluation metrics (AUC-ROC, AP, Brier…)

Your data.frame must include:

• Outcome columns named like *_2_year_outcome / *_5_year_outcome
• Prediction columns named kfre_{n}var_{year}year, e.g. kfre_4var_2year

met <- eval_kfre_metrics(
  df = toy_kfre,                 # must contain truth + prediction columns
  n_var_list = c(4, 6, 8),
  outcome_years = c(2, 5),
  decimal_places = 4
)

met
# Rows: Metrics; Cols: "{2_year|5_year}_{4|6|8}_var_kfre"

8. Plot ROC / PR curves

# Basic: compute & plot both ROC and PR (no files written)
plot_kfre_metrics(
  df = toy_kfre,
  num_vars = c(4, 6, 8),
  plot_type = "all_plots",
  mode = "both",              # compute + plot
  show_years = c(2, 5)
)

# Save to disk (PNG/SVG)
plot_kfre_metrics(
  df = toy_kfre,
  num_vars = c(4, 6),
  plot_type = "auc_roc",
  mode = "both",
  show_years = c(2, 5),
  save_plots = TRUE,
  image_path_png = "plots",
  image_prefix = "kfre"
)

Running Tests

If you’ve cloned the repo:

library(devtools)
devtools::load_all(".")
devtools::test()

You should see unit tests for both the end-to-end flow and the evaluation utilities.

API surface (exports)

• RiskPredictor (R6)
  • $predict_kfre(years, is_north_american, use_extra_vars, num_vars)
  • $kfre_person(...)
• Wrappers:
  • predict_kfre(df, columns, years, is_north_american, use_extra_vars, num_vars)
  • add_kfre_risk_col(...)
• Utilities:
• upcr_uacr(...)
• perform_conversions(...)
• class_esrd_outcome(...)
• class_ckd_stages(...)
• eval_kfre_metrics(...)
• plot_kfre_metrics(...)

Notes on parity with Python

The R implementations are designed to mirror the Python versions (naming, shapes, and expected columns). Where packages differ (e.g., ROC/PR computation), we use pROC and precrec to maintain metric parity.

References

• Tangri N, Grams ME, Levey AS, et al. (2016). Multinational assessment of accuracy of equations for predicting risk of kidney failure: A meta-analysis. JAMA, 315(2), 164–174. doi:10.1001/jama.2015.18202

• Tangri N, Stevens LA, Griffith J, et al. (2011). A predictive model for progression of chronic kidney disease to kidney failure. JAMA, 305(15), 1553–1559. doi:10.1001/jama.2011.451

• Sumida K, Nadkarni GN, Grams ME, et al. (2020). Conversion of urine protein-creatinine ratio or urine dipstick protein to urine albumin-creatinine ratio for use in CKD screening and prognosis. Ann Intern Med, 173(6), 426–435. doi:10.7326/M20-0529

License

kfre is distributed under the MIT License. See LICENSE for more information.