Open-source · communicable disease control demo.sentinelsurveillance.app
An open-source platform for communicable disease surveillance, outbreak management, and contact tracing.
The grievance
Our surveillance systems were not ready for what was asked of them, and they failed us.
When adding one question to a questionnaire takes a help-desk ticket and a staff retraining session, the system cannot keep pace with an outbreak. Sentinel was built to address this: surveillance staff configure the system themselves as requirements change.
Design principles
The epidemiologist or surveillance manager controls the system, rather than whoever maintains the code. Adding a question to a salmonella questionnaire, for example whether a case ate bean sprouts, does not require a developer.
Geographies, demographics, diseases, case definitions, surveys and reports are all configurable through the interface, with no code change and no schema migration.
The design target is the end user in a large operation. A changed field or a relabelled question does not require a training session: the contact tracer clicks next, and the new question is already there.
The system accommodates the user, rather than the reverse.
The platform
Case intake, investigation, isolation and contact tracing in one place.
Workbench Patient search, your outbreaks and your task queue, shown at sign-in
Investigation Labs, classification, exposures and tasks for a single case
Case definitions Machine-readable criteria, built by the epidemiologist without code
Contact tracing Team supervision: assignments, call attempts and progress
Data governance Data can be staged, committed or flagged for review depending on needs
Under the hood
Sentinel's core data structures are patients, cases, contacts and outbreaks. Everything around them is configured by the user. Four components handle most of the intake and investigation work.
Picks up HL7 messages from a file drop, matches them to the right patient using strategies you configure, and parses results with LOINC and SNOMED. You choose how data from a HL7 message is entered. Data enters automatically - only if you say so.
Machine-readable case definitions are evaluated in the background against labs, symptoms and exposures. The epidemiologist decides whether they apply automatically, flag for review, or stay manual. Each evaluation keeps a full history and can be overridden.
Surveys carry conditional logic and versioning. Responses map into case and patient records: trusted fields save automatically, ambiguous ones queue for review. A new question reaches every tracer without a code change.
Hierarchical outbreaks, interview queues, supervisor dashboards, bulk contact import, contact-to-case conversion, and a mind-map view of relationships.
Open source
The code is public and free to use, share and modify. Sentinel is in active development and open to contributors. The repository and a live demo are linked below.
Questions, or want to get involved? [email protected]
Pip · the watcher