Endoscopy Triage & Diagnosis Tool
Clinical Decision Support • R Shiny • Random Forest • Healthcare Analytics
Project Overview
A specialized Shiny web application designed to assist clinicians in diagnosing and triaging patients with gastrointestinal bleeding. The tool uses Random Forest machine learning models trained on clinical data to provide probabilistic predictions for bleeding source identification, urgent endoscopy necessity, and appropriate patient disposition. This clinical decision support system helps healthcare professionals make data-driven decisions while maintaining the primacy of clinical judgment.
Clinical Decision Pipeline
Data Input
Clinical parameters
Demographic data
Vital signs & labs
Model Processing
4 Random Forest models
Simultaneous analysis
Feature evaluation
Prediction Generation
Probabilistic outputs
Confidence intervals
Risk stratification
Visualization
ggplot2 charts
Feature importance
Stacked bar plots
Clinical Decision
Decision support
Physician integration
Patient outcomes
Clinical Prediction Models
Bleeding Source
Classifies GI bleeding as upper, mid, or lower gastrointestinal tract based on clinical presentation
Urgent Endoscopy
Determines the need for immediate endoscopic intervention based on severity indicators
Patient Disposition
Recommends ICU vs. non-ICU placement based on severity and risk assessment
Resuscitation Needs
Evaluates immediate resuscitation requirements for patient stabilization
Application Features
Interactive Interface
Web-based clinical data entry with intuitive Shiny UI design
Real-time Predictions
Probabilistic predictions using Random Forest models with instant results
Custom Visualizations
ggplot2 charts showing prediction confidence and feature importance
Physician Validation
Input fields for physician diagnosis comparison and validation
Clinical Parameters
Comprehensive collection of demographics, vital signs, and lab values
Multi-Patient Analysis
Compute and refresh functionality for multiple patient assessments
Clinical Applications & Important Disclaimers
Clinical Applications
Decision Support
Provides probabilistic guidance for complex GI bleeding cases
Educational Tool
Demonstrates machine learning applications in clinical settings
Research Platform
Framework for developing additional clinical prediction models
⚠️ Important Disclaimers
Clinical Judgment Priority
Intended as demonstration tool, not replacement for clinical judgment
Privacy Compliance
Users must ensure healthcare privacy law compliance with patient data
Medical Device Status
Not approved as medical device - requires clinical validation for deployment