clifR

clifR is an R port of the Python library clifpy for standardizing and analyzing critical care (ICU) data using the Common Longitudinal ICU data Format (CLIF) version 2.0.

Overview

Transform heterogeneous ICU data into standardized, analysis-ready datasets with built-in validation, clinical calculations, and high-performance data processing.

Features

• CLIF 2.0 Tables: 18 core table classes implemented (Patient, Hospitalization, ADT, Vitals, Labs, Diagnoses, Medications, Respiratory Support, Code Status, CRRT, ECMO/MCS, Microbiology, Assessments, Procedures, Position)
• Schema Validation: Automatic validation against CLIF specification
• Clinical Calculations:
  • SOFA (Sequential Organ Failure Assessment) scores
  • Charlson Comorbidity Index (CCI) from ICD-9/ICD-10 codes
  • P/F ratio calculation
  • Ventilator settings analysis
• Medication Analysis:
  • Vasopressor and sedation tracking
  • Antibiotic administration timing
  • Dose unit conversion
• Respiratory Support: Ventilator modes, settings, and compliance metrics
• Unit Conversion: Smart conversion between medical units (doses, temperature, pressure, labs)
• Encounter Stitching: Link related hospital stays
• Wide Dataset Creation: Transform narrow clinical data to time-series format
• High Performance: Leverages DuckDB and arrow for efficient processing
• Timezone-Aware: Proper handling of timestamps across timezones

Installation

# Install development version from GitHub
# install.packages("devtools")
devtools::install_github("AartikSarma/clifR")

Quick Start

library(clifR)

orchestrator <- ClifOrchestrator$new(
  data_directory = "path/to/clif/data",
  filetype = "parquet",
  timezone = "US/Eastern"
)

# Name the tables you want. initialize_tables() with no arguments loads
# ONLY patient, matching clifpy's default of ['patient'].
orchestrator$initialize_tables(
  c("patient", "hospitalization", "adt", "vitals", "labs")
)

# Validate every loaded table
orchestrator$validate_all()

# Access data
vitals <- orchestrator$vitals$df
labs <- orchestrator$labs$df

# Create wide dataset
wide_data <- orchestrator$create_wide_dataset()

# Calculate SOFA scores
sofa_scores <- orchestrator$compute_sofa_scores()

Data files must use the CLIF clif_ prefix — clif_patient.parquet, clif_vitals.parquet, and so on. A bare patient.parquet is not recognized.

CLIF Specification

This package implements the CLIF 2.0 specification for standardized ICU data.

Supported Tables

• Patient demographics
• Hospitalization records
• ADT (Admission/Discharge/Transfer) events
• Vital signs
• Laboratory results
• Respiratory support
• Medications (continuous and intermittent)
• Microbiology (culture, non-culture, susceptibility)
• Diagnoses, procedures, assessments
• And more...

Documentation

• CLIF Specification: https://clif-icu.com/data-dictionary
• Python Library (clifpy): https://github.com/Common-Longitudinal-ICU-data-Format/clifpy
• Package Documentation: [Coming soon]

Development Status

🚧 This package is under active development. Core functionality is being ported from the Python library with systematic cross-validation.

Contributing

This is an R port of clifpy. Contributions are welcome! Please ensure:

• R implementations match Python outputs (see cross-validation tests)
• Code follows tidyverse style guide
• All functions have roxygen2 documentation
• Tests pass with devtools::check()

Related Projects

• clifpy: Python implementation (GitHub)
• CLIF Consortium: https://clif-icu.com/

License

Apache License 2.0

Citation

If you use this package in your research, please cite the CLIF Consortium and the original clifpy library.