JupyterHealth is a project that develops open source infrastructure for building and deploying digital health tools that give patients, clinicians, and analysts more control over health data.
The core workflow¶
JupyterHealth is built around four steps:
Collect patient data. An organization sets up a study in a JupyterHealth Exchange. Patients join the study, choose what to share, and connect their relevant devices, apps, or health records to send data to the Exchange. The Exchange only accepts the kinds of data from a patient that they have consented to share.
Manage access to data. An administrator or data engineer configures access to the Exchange. The Exchange controls which users can access the data, and which applications can read it through the Exchange’s APIs.
Analyze data and build data products. A data scientist, researcher, or developer pulls data from the Exchange into an environment like the Hub for analysis. They can turn results into a data product, like a dashboard or website, using tools like Voilà or Jupyter Book.
Share data products with clinicians and patients. Data products can then be deployed as applications or dashboards for clinicians and patients to use. Clinician-facing applications can also be launched directly from an EHR, so a clinician can open the application from a patient’s chart without leaving their usual workflow.
See Components for the documentation of each tool in this workflow.
Design principles¶
JupyterHealth follows the same design principles as Jupyter:
Modular. Each piece should be useful on its own, outside the full JupyterHealth platform. For example, an institution that has its own computing infrastructure could run an Exchange and connect it to that infrastructure instead of a Hub.
Built on open standards. Data is stored and served with standards like FHIR and Open mHealth, so it works with other tools in health care.
Open source and replicable. The code and deployment configuration are public, so anybody can run their own deployment.