Predictive Analytics Agent

Agent Description:

The Predictive Analytics Agent is a healthcare AI system designed to forecast patient outcomes and health risks using historical and clinical data. It automates the end-to-end pipeline from data extraction to classification and reporting. By generating both clinician-focused summaries and simplified patient-facing reports, it supports proactive healthcare management and enables timely, informed decision-making for all stakeholders.

Purpose and Components
  • Purpose: To automate the prediction of patient health risks and generate easy-to-understand reports for both clinicians and patients.
  • Components:
    • A data extractor for pulling clinical and historical patient data.
    • A prediction engine to classify patients into risk levels.
    • A dual-format reporter to produce both clinical summaries and patient-friendly outputs.
Supported Capabilities
  • Extracting comprehensive data from clinical databases, including patient history, vitals, and lab results.
  • Classifying patients into Low, Moderate, or High risk categories.
  • Identifying clinical indicators driving the patient’s risk status.
  • Generating clinical summaries showing risk distribution and predictive patterns.
  • Producing readable, patient-facing reports with suggested next steps in non-technical language.
LLM Used
  • Google Vertex
    Note: To learn more about the LLM and to modify its behavior, refer to the Configuring LLM settings section.

Sub-Agents

1. Clinical Data Extractor

  • Role:Data Extractor.
  • Scope:Identifies and extracts all relevant tables from the clinical database for forecasting tasks.
  • Description: This sub-agent connects to the clinical database and retrieves information from tables containing patient history, vitals, and model metadata. It formats the data into a structured format for downstream analysis, without performing any predictive operations.
  • LLM Used: Google Vertex (inherits from parent).
  • Tool Used: MCP Client Tool

2. Health Outcome Predictor

  • Role:Health Predictor.
  • Scope:Analyzes structured data to assess patient-level risk of readmission, mortality, or disease onset.
  • Description: This sub-agent applies classification models on clinical indicators like blood glucose, blood pressure, age, and comorbidities to categorize patients into Low, Moderate, or High risk. It identifies key contributing indicators and flags high-risk cases.
  • LLM Used: Google Vertex (inherits from parent).

3. Risk Insight Reporter

  • Role:Risk Analyzer.
  • Scope:Analyzes prediction results and produces both clinical and patient-friendly outputs.
  • Description: This sub-agent processes the results from the HealthOutcomePredictor to generate two tailored report types:

    • A Clinical Summary for healthcare providers, including risk breakdowns and driver metrics.

    • A Patient-Friendly Report that explains the risk level and recommends actions in accessible language.

  • LLM Used: Google Vertex (inherits from parent).
Tools Used:
  • MCP Client Tool : Used by the Clinical Data Extractor for extracting and structuring data from clinical sources.
Note: For details on modifying the Tools, refer Tools Library section.
Agent Workflow Behavior Summary
  • The Clinical Data Extractor begins by connecting to the clinical database and retrieving all relevant historical and diagnostic data for each patient.
  • The structured data is passed to the HealthOutcomePredictor, which processes it using analytical models to classify patients into Low, Moderate, or High risk categories. It also highlights key clinical indicators contributing to the prediction.
  • The RiskInsightReporter then takes the prediction output and generates:
    • A Clinical Summary Report for medical professionals showing aggregated risk levels and common contributing factors.
    • A Patient-Friendly Report for individuals, written in plain language, outlining their personal risk category and suggested health actions.
Sample Questions:
  • What is the predicted risk score for patient ID 2?
  • Generate health risk predictions and show me the top contributing factors.