University-Wide Analytics Use Case Scenarios

Overview

This document outlines key use cases for the Washington University in St. Louis data lake, supported by the ICS Platform Engineering Team. Each use case demonstrates how federated governance, platform support, and stewardship enable analytics across research, clinical operations, finance, and student services.

Research Domain

Example: Cohort Analysis Using the Data Lake

Washington University researchers can leverage the data lake to perform cohort discovery, phenotyping, and longitudinal analyses. The platform integrates structured EHR data and clinical notes, as well as clinical data mapped to the OMOP data model for standards based, self-service analytics.

Support Model:

• Domain Data Steward: Defines cohort definitions and phenotypes.
• Compliance Approver: Ensures IRB approval is recorded before access.
• ICS Platform Engineering Team: Provisions workspaces, enforces data classification, manages runtime and access audit logs.
• Enterprise Data Steward: Provides onboarding training and documentation for cohort workflows.

Clinical Operations Domain

Example: Real-Time Outcome Prediction Pipelines

The data lake supports automated ML pipelines for patient outcome predictions, such as ICU transfers and discharge probabilities. These pipelines deliver actionable insights to clinicians, improving patient care and operational efficiency.

Support Model:

• Domain Data Steward & Custodian: Define clinical data elements and agree on model outputs.
• Compliance Approver / CPO: Ensures patient identifiers and PHI handling policies are enforced.
• ICS Platform Engineering Team: Builds ETL pipelines, orchestrates batch runs or streaming ingestion, manages job scheduling, and secures dashboards.
• Reporting/Audit Engineer: Creates usage dashboards, audit logs, and tracks KPIs.

Finance / Administrative Operations Domain

Example: Monthly and On-Demand Expense Reporting by Department

The finance team leverages the data lake to generate detailed expense reports by department. These reports can be produced on a monthly schedule or on demand, providing insights into spending patterns and budget utilization.

Support Model:

• Domain Data Steward: Defines expense categories, reporting templates, and validation rules.
• Compliance Approver: Ensures financial data access complies with university policies and audit requirements.
• ICS Platform Engineering Team: Automates data ingestion from financial systems, schedules report generation, and secures access to reports.
• Enterprise Data Steward: Provides training on report interpretation, data access, and self-service tools.

Student Services / Educational Analytics Domain

Example: Analytics on Student Performance and Retention

The data lake enables analysis of student performance, retention, and outcomes using registrar records, surveys, and performance metrics. Predictive models identify at-risk students and support intervention strategies.

Support Model:

• Domain Data Steward: Defines student outcome metrics, data quality rules, and privacy-safe identifiers.
• Compliance Approver / CPO: Ensures adherence to FERPA and educational privacy standards.
• ICS Platform Engineering Team: Manages tokenized or masked student datasets, provides domain-specific workspace segments, and sets access per advisor roles.
• Training Specialist & Enterprise Steward: Run workshops for advising staff, document pipelines, and example dashboards.

Integration Across Domains

These scenarios illustrate:

• Federated Governance Workflow: Domain committees approve access and policies, aligned centrally via the University Data Governance Committee (UDGC).
• Shared Platform Enforcement: The ICS Platform Engineering Team provides infrastructure, access controls, data ingestion pipelines, runtime updates, and auditing that support all domains.
• Stewardship and Literacy: Domain stewards interpret business definitions; enterprise stewardship trains users across the university.

Summary Table: Use Cases and Governance Roles

Published Precedents

• Stanford’s STARR‑OMOP: Supports research workflows and phenotyping in a governed, scalable analytics platform (Taylor & Francis Online, starr.stanford.edu, arXiv, starr.stanford.edu).
• Hospital Network Outcome Prediction Deployment: ML-based pipelines driving operational improvement across inpatient populations (arXiv).
• Broader Health Analytics Use Case Literature: Shows typical models for fraud detection, financial anomaly identification, and predictive operations in academic medical centers (TechTarget).