eICUDataset: Dataset class for eICU-CRD

A dataset class for handling eICU Collaborative Research Database (eICU-CRD) data. The eICU-CRD is a large dataset of de-identified health records from ICU patients across multiple hospitals in the United States. This class inherits from BaseDataset and provides specialized handling for eICU data structures.

Details

The eICU dataset is centered around ICU stays (patientunitstayid), where:

• A patient (uniquepid) can have multiple hospital admissions (patienthealthsystemstayid)

• Each hospital admission can have multiple ICU stays (patientunitstayid)

• All timestamps are relative offsets (in minutes) from the ICU admission time

Default Tables

The following table is loaded by default:

• patient: Core patient demographics and ICU stay information

Optional Tables

Additional tables can be specified via the tables parameter:

• diagnosis: ICD-9 diagnoses with diagnosis strings

• treatment: Treatment information with treatment strings

• medication: Medication orders with drug names and dosages

• lab: Laboratory measurements with lab names and results

• physicalexam: Physical examination findings

• admissiondx: Primary admission diagnoses per APACHE scoring

Data Source

The eICU-CRD dataset is available at https://eicu-crd.mit.edu/. Access requires completion of the CITI "Data or Specimens Only Research" course and signing a data use agreement.

Note on Patient IDs

In the Python pyhealth implementation, patient_id is a composite of uniquepid and patienthealthsystemstayid to represent a hospital admission. In this R implementation, we use patientunitstayid as the primary identifier to represent individual ICU stays, which aligns better with the event-based architecture of BaseDataset.

See also

BaseDataset, MIMIC3Dataset, OMOPDataset

Super class

RHealth::BaseDataset -> eICUDataset

Methods

Public methods

• eICUDataset$new()

• eICUDataset$clone()

Inherited methods

• RHealth::BaseDataset$collected_global_event_df()
• RHealth::BaseDataset$default_task()
• RHealth::BaseDataset$get_patient()
• RHealth::BaseDataset$iter_patients()
• RHealth::BaseDataset$load_data()
• RHealth::BaseDataset$load_table()
• RHealth::BaseDataset$set_task()
• RHealth::BaseDataset$stats()
• RHealth::BaseDataset$unique_patient_ids()

---

Method new()

Initialize the eICUDataset.

Usage

eICUDataset$new(
  root,
  tables = character(),
  dataset_name = NULL,
  config_path = NULL,
  dev = FALSE,
  ...
)

Arguments

root

Character. Root directory of the eICU dataset containing CSV files (e.g., patient.csv, diagnosis.csv, etc.).

tables

Character vector of additional tables to include beyond the default patient table. Available options: "diagnosis", "treatment", "medication", "lab", "physicalexam", "admissiondx".

dataset_name

Character. Optional custom name for the dataset. Defaults to "eicu".

config_path

Character. Optional path to a custom YAML configuration file. If NULL (default), uses the built-in eICU configuration.

dev

Logical. If TRUE, limits data loading to 1000 patients for rapid prototyping and testing. Default is FALSE.

...

Additional arguments passed to BaseDataset$initialize.

Returns

A new eICUDataset object.

Examples

\dontrun{
# Basic initialization
ds <- eICUDataset$new(
  root = "/data/eicu-crd/2.0"
)

# With additional tables and dev mode
ds <- eICUDataset$new(
  root = "/data/eicu-crd/2.0",
  tables = c("diagnosis", "medication", "lab"),
  dev = TRUE
)
}

---

Method clone()

The objects of this class are cloneable with this method.

Usage

eICUDataset$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

if (FALSE) { # \dontrun{
# Initialize with default patient table only
dataset <- eICUDataset$new(
  root = "/path/to/eicu-crd/2.0",
  dev = TRUE
)

# Initialize with additional clinical tables
dataset <- eICUDataset$new(
  root = "/path/to/eicu-crd/2.0",
  tables = c("diagnosis", "medication", "lab", "treatment"),
  dev = FALSE
)

# Display dataset statistics
dataset$stats()
} # }


## ------------------------------------------------
## Method `eICUDataset$new`
## ------------------------------------------------

if (FALSE) { # \dontrun{
# Basic initialization
ds <- eICUDataset$new(
  root = "/data/eicu-crd/2.0"
)

# With additional tables and dev mode
ds <- eICUDataset$new(
  root = "/data/eicu-crd/2.0",
  tables = c("diagnosis", "medication", "lab"),
  dev = TRUE
)
} # }