EHRShotDataset: Dataset class for EHRShot Benchmark

A dataset class for handling EHRShot data. EHRShot is a benchmark for few-shot evaluation on Electronic Health Records (EHR) data, covering multiple predictive tasks including operational outcomes, lab values, new diagnoses, and medical imaging findings.

Details

The EHRShot benchmark provides standardized evaluation across multiple clinical prediction tasks:

Operational Outcomes:

• guo_los: Length of stay prediction

• guo_readmission: Hospital readmission prediction

• guo_icu: ICU admission prediction

Lab Values:

• lab_thrombocytopenia: Low platelet count prediction

• lab_hyperkalemia: High potassium level prediction

• lab_hypoglycemia: Low blood sugar prediction

• lab_hyponatremia: Low sodium level prediction

• lab_anemia: Anemia prediction

New Diagnoses:

• new_hypertension: New hypertension diagnosis

• new_hyperlipidemia: New hyperlipidemia diagnosis

• new_pancan: New pancreatic cancer diagnosis

• new_celiac: New celiac disease diagnosis

• new_lupus: New lupus diagnosis

• new_acutemi: New acute myocardial infarction diagnosis

Medical Imaging:

• chexpert: CheXpert multi-label chest X-ray finding classification

Data Structure

The dataset expects the following CSV files in the root directory:

• ehrshot.csv: Main events table with clinical codes

• splits.csv: Train/validation/test split assignments

• <task_name>.csv: Label files for each prediction task

Website

For more information, visit: https://som-shahlab.github.io/ehrshot-website/

See also

BaseDataset, BenchmarkEHRShot

Super class

RHealth::BaseDataset -> EHRShotDataset

Methods

Public methods

• EHRShotDataset$new()

• EHRShotDataset$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 EHRShotDataset.

Usage

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

Arguments

root

Character. Root directory of the EHRShot dataset containing CSV files (e.g., ehrshot.csv, splits.csv, task label files).

tables

Character vector of tables to include. Should include at minimum "ehrshot" for events and "splits" for data splitting. Additional task-specific label tables can be added (e.g., "guo_los", "lab_thrombocytopenia"). Available task tables include:

• Operational outcomes: "guo_los", "guo_readmission", "guo_icu"

• Lab values: "lab_thrombocytopenia", "lab_hyperkalemia", "lab_hypoglycemia", "lab_hyponatremia", "lab_anemia"

• New diagnoses: "new_hypertension", "new_hyperlipidemia", "new_pancan", "new_celiac", "new_lupus", "new_acutemi"

• Imaging: "chexpert"

dataset_name

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

config_path

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

Examples

\dontrun{
# Basic initialization with single task
ds <- EHRShotDataset$new(
  root = "/data/ehrshot",
  tables = c("ehrshot", "splits", "guo_los")
)

# With multiple tasks and dev mode
ds <- EHRShotDataset$new(
  root = "/data/ehrshot",
  tables = c("ehrshot", "splits", "lab_thrombocytopenia", "new_hypertension"),
  dev = TRUE
)
}

---

Method clone()

The objects of this class are cloneable with this method.

Usage

EHRShotDataset$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

if (FALSE) { # \dontrun{
# Initialize with default ehrshot table only
dataset <- EHRShotDataset$new(
  root = "/path/to/ehrshot/data",
  tables = c("ehrshot", "splits", "guo_los"),
  dev = TRUE
)

# Initialize with multiple task tables
dataset <- EHRShotDataset$new(
  root = "/path/to/ehrshot/data",
  tables = c("ehrshot", "splits", "lab_thrombocytopenia", "new_hypertension"),
  dev = FALSE
)

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


## ------------------------------------------------
## Method `EHRShotDataset$new`
## ------------------------------------------------

if (FALSE) { # \dontrun{
# Basic initialization with single task
ds <- EHRShotDataset$new(
  root = "/data/ehrshot",
  tables = c("ehrshot", "splits", "guo_los")
)

# With multiple tasks and dev mode
ds <- EHRShotDataset$new(
  root = "/data/ehrshot",
  tables = c("ehrshot", "splits", "lab_thrombocytopenia", "new_hypertension"),
  dev = TRUE
)
} # }