DeepPatientLevelPrediction Installation Guide

Introduction

This vignette describes how to install the Observational Health Data Sciences and Informatics (OHDSI) DeepPatientLevelPrediction package on Windows, macOS, and Linux.

Software Prerequisites

Windows Users

Under Windows, install:

R (https://cran.r-project.org/) - R >= 4.1.0; the latest release is recommended
Python - Python 3.14 is recommended; Python >= 3.10 is supported
An R development environment, if desired
Java when using a JDBC-based database connection
Rtools when installing packages from source

Mac/Linux Users

Under macOS and Linux, install:

R (https://cran.r-project.org/) - R >= 4.1.0; the latest release is recommended
Python - Python 3.14 is recommended; Python >= 3.10 is supported
An R development environment, if desired
Java when using a JDBC-based database connection
Xcode command line tools on macOS when installing packages from source (xcode-select --install)

Installing the Package

Install the released package from CRAN:

install.packages("DeepPatientLevelPrediction")

The development version can instead be installed from GitHub. Development versions may contain changes that have not yet been released.

install.packages("remotes")
remotes::install_github("OHDSI/DeepPatientLevelPrediction@develop")

Python environment

Model training and inference use PyTorch through reticulate. Python is not required to install or load the R package. When Python functionality is first used, reticulate can create a managed environment from the package requirements on systems with internet access.

You can verify the active interpreter with:

library(DeepPatientLevelPrediction)
reticulate::py_config()

Advanced users, users with strict reproducibility requirements, or users in air-gapped environments can manage the Python environment themselves and tell reticulate which interpreter to use. One option is to create the environment with uv and Python 3.14:

uv python install 3.14
uv venv --python 3.14
uv pip install polars tqdm pyarrow duckdb nvidia-ml-py numpy
uv pip install "torch==2.12.1" --index https://download.pytorch.org/whl/cpu/

The second uv pip install command installs the CPU build of PyTorch. If you want to train on a GPU, install the PyTorch build that matches your CUDA setup instead.

To force reticulate to use a manually managed interpreter, set RETICULATE_PYTHON in .Renviron.

For Linux/macOS:

RETICULATE_PYTHON="/path/to/project/.venv/bin/python"

For Windows:

RETICULATE_PYTHON="C:/path/to/project/.venv/Scripts/python.exe"

Then restart your R session.

Python 3.9 is end-of-life and should not be used. Python 3.10 is still supported, but Python 3.14 is recommended.

Accessing the torch helper or starting model training triggers reticulate to resolve the Python requirements if RETICULATE_PYTHON is not configured.

library(DeepPatientLevelPrediction)
torch$randn(10L)

This should print out a tensor with ten different values.

On Windows, close other R sessions that are using DeepPatientLevelPrediction or its dependencies before updating the package; open sessions can lock installed files.

Testing Installation

library(DeepPatientLevelPrediction)

torch$randn(10L)

The R-side model settings can be created without initializing Python:

modelSettings <- DeepPatientLevelPrediction::setResNet(
  numLayers = 2L,
  sizeHidden = 64L,
  hiddenFactor = 1L,
  residualDropout = 0,
  hiddenDropout = 0.2,
  sizeEmbedding = 64L,
  estimatorSettings = DeepPatientLevelPrediction::setEstimator(
    learningRate = 3e-4,
    weightDecay = 1e-6,
    device = "cpu",
    batchSize = 128L,
    epochs = 3L,
    seed = 42L
  ),
  hyperParamSearch = "random",
  randomSample = 1L
)

stopifnot(inherits(modelSettings, "modelSettings"))

To run an end-to-end patient-level prediction example, continue with the first-model vignette.

Acknowledgments

Considerable work has been dedicated to providing the DeepPatientLevelPrediction package.

citation("DeepPatientLevelPrediction")

## To cite package 'DeepPatientLevelPrediction' in publications use:
## 
##   Fridgeirsson E, Reps J, Chan You S, Kim C, John H (2026).
##   _DeepPatientLevelPrediction: Deep Learning for Patient-Level
##   Prediction_. R package version 2.4.0,
##   <https://ohdsi.github.io/DeepPatientLevelPrediction/>.
## 
## A BibTeX entry for LaTeX users is
## 
##   @Manual{,
##     title = {DeepPatientLevelPrediction: Deep Learning for Patient-Level Prediction},
##     author = {Egill Fridgeirsson and Jenna Reps and Seng {Chan You} and Chungsoo Kim and Henrik John},
##     year = {2026},
##     note = {R package version 2.4.0},
##     url = {https://ohdsi.github.io/DeepPatientLevelPrediction/},
##   }

Please reference this paper if you use the PLP Package in your work:

Reps JM, Schuemie MJ, Suchard MA, Ryan PB, Rijnbeek PR. Design and implementation of a standardized framework to generate and evaluate patient-level prediction models using observational healthcare data. J Am Med Inform Assoc. 2018;25(8):969-975. doi:10.1093/jamia/ocy032.

Egill Fridgeirsson

2026-10-09