VigiLens: Clinical Safety Command Center

Real-time clinical safety monitoring with Tableau Next and Data Cloud. Turn scattered signals into prioritized risk scores and trigger instant CAPA workflows in Salesforce to save lives.

Comment

Inspiration

Clinical trials are an $80+ billion industry, but the currency that matters most isn't money—it's human trust. Currently, safety data is siloed across Electronic Data Capture (EDC) systems, site logs, and spreadsheets. When a safety signal (like a spike in Adverse Events) is missed or delayed, patients get hurt, and life-saving therapies are stalled. We realized that safety teams don't need more reports; they need a Command Center.

We asked: "What if we could detect a safety threat the moment it happens and trigger a compliance action in one click?" That question built VigiLens.

What it does

VigiLens is a closed-loop safety monitoring system that transforms scattered clinical data into immediate action.

Ingests & Harmonizes: It pulls raw data (Adverse Events, Protocol Deviations, Site Training Logs) into Salesforce Data Cloud.

Detects Risk: Using a custom Semantic Model, it calculates a live Risk Signal Score for every trial site.

Visualizes Urgency: The Tableau Next dashboard ranks sites by risk. A dynamic threshold line (CAPA ≥ 15) visually separates normal sites from critical threats (highlighted in red).

Triggers Action: Users click a "Create Risk Action" deep link directly on the dashboard.

Executes Workflow: This launches a Salesforce Flow with context passed automatically (Site Name, Risk Score, AE Counts), creating an audit-ready CAPA (Corrective and Preventive Action) record in the CRM.

How we built it

The architecture relies on the synergy between Salesforce Data Cloud and Tableau Next:

Data Layer: We ingested CSV datasets representing Sites, Patients, and Adverse Events into Data Cloud. We built a Snowflake Schema in the Semantic Model Builder to map relationships (e.g., Adverse_Event →→ Patient →→ Site) while strictly avoiding cyclic conflicts.

Analytics Layer: We used Tableau Next to build the visualizations. We utilized Level of Detail (LOD) calculations or aggregate measures to derive the Risk Signal Score (Sum of Severity Grades).

Integration Layer: The "Magic Button" was built using a URL Action in Tableau. We constructed a dynamic URL that passes dashboard parameters (&riskScore={{value}}) directly into Salesforce Flow variables.

Action Layer: We built a Screen Flow in Salesforce that accepts these input variables, allows for human review ("Human-in-the-Loop"), and commits a Risk_Action__c record to the database for auditability.

Challenges we ran into

Semantic Model Cyclic Conflicts: Early in the build, linking all objects created "Cyclic Dependency" errors. We resolved this by restructuring our graph into a clear hierarchy (Parent-Child) and ensuring all joins flowed upstream to the Site object.

Data Type Mismatches in URL Actions: Salesforce Flow is strict about data types. Tableau was sending "12.0" (decimal) which caused the Flow to crash because it expected an Integer. We solved this by creating a calculated field in Tableau: STR(INT(COUNT([Adverse_Event]))) to sanitize the data before transmission.

Visualizing "Invisible" Risks: Initially, we just listed sites. We realized this wasn't actionable. We added the Reference Line (Threshold = 15) and conditional coloring to create a visual "Call to Action" that judges and users could understand instantly.

Accomplishments that we're proud of

The "One-Click" Integration: We successfully passed data from a Tableau visualization into a Salesforce Flow without any code—just parameter mapping.

The UI Design: Moving away from a cluttered map to a clean, sorted Leaderboard that tells a clear story: "Red means Danger."

Real-World Applicability: This isn't just a toy; the "Insight →→ Action →→ Audit" loop is exactly what FDA/EMA regulators require for clinical oversight.

What we learned

Tableau Next significantly speeds up the modeling phase compared to traditional desktop modeling.

Data Governance: The importance of defining field categories (like Latitude / Longitude) at the Data Lake Object level to ensure downstream tools recognize them correctly.

The Power of Narrative: Tech matters, but connecting the tech to patient safety is what makes the solution valuable.

What's next for VigiLens: Clinical Safety Command Center

Einstein AI: Integrating predictive modeling to flag sites before they breach the threshold based on historical trends.

Slack Integration: Pushing "High Risk" alerts directly to safety monitors' mobile devices via Salesforce Flow.

Geospatial Drill-downs: Re-enabling the global map view with drill-down capabilities for regional monitoring.

Built With

Share this project:

Updates