meds-reader 0.0.4

A Pythonic interface for MEDS datasets

meds_reader: A Pythonic interface for MEDS datasets

meds_reader is a fast and easy-to-use library for reading and processing patient data stored in MEDS (Medical Event Data Standard) format using a Python-native API.

Inspired by HuggingFace Datasets, meds_reader transforms MEDS datasets into collections of Python objects, that can then be processed using normal Python functions.

[!WARNING] meds_reader is currently in early stage development and the API is still in flux. Please leave feedback and use at your own risk!

Example

# First we need to convert to a database
# That requires running the following command in a terminal
$ meds_reader_convert mimic_meds mimic_meds_reader

# Now we can use Python to interact with our data
import meds_reader

# We can now construct a database
database = meds_reader.PatientDatabase("mimic_meds_reader")

# Databases are dict-like objects that store patient data

# We can iterate through all the patient ids in the database
for patient_id in database:

    # We can retrieve patient data given a patient_id
    patient = database[patient_id]

    # Patient data can be manipulated with normal Python operations
    print(patient.patient_id)
    for event in patient.events:
        print(event.time, event.code)

To learn more, see our full documentation.

Installation

meds_reader can be installed using pip.

pip install meds_reader

Interactive Demo

An interactive demo can be found at Google Colab.