ICU Resource Optimization and Simulation Project

Authors

• Kenny Chan - kennychan@berkeley.edu
• Sizhe Li - sizheli@berkeley.edu
• Siqi Yao - siqiyao2024@berkeley.edu
• Yiyao Li - yiyaoli@berkeley.edu

---

📖 Project Overview

This project addresses the challenge of optimizing ICU (Intensive Care Unit) resources, including patient beds and caregivers, using a simulation framework. The goal is to minimize patient waiting times and operational penalties under varying demand conditions.

Key components:

1. ICU Queue Simulation: Models patient arrivals, prioritization strategies, and bed allocations.
2. Caregiver Workflow Simulation: Simulates caregiver-patient interactions and service processes.
3. Optimization Strategies: Implements exhaustive, heuristic, and Pareto front-based approaches to resource allocation.

---

📂 File Structure

.
├── Part_1_IcuQueue
│   ├── ArrivalProcess.py
│   ├── DepartureProcessWithFIFO.py
│   ├── DepartureProcessWithDPQandReservedBeds.py
│   ├── DepartureProcessWithDPQ.py
│   ├── DepartureProcessWithPQ.py
│   ├── DepartureProcessWithPQandReservedBeds.py
│   ├── DepartureProcessWithReservedBeds.py
│   ├── figures/ (output plots)
│   ├── penalties/ (output penalty results)
│   └── __init__.py
│
├── Part_2_CareGiver
│   ├── CareGiverMain.py
│   ├── CareRequest.py
│   ├── figures/ (output plots)
│   ├── penalties/ (output penalty results)
│   └── __init__.py
│
├── Sensitivity_Analysis
│   ├── SensitivityAnalysis.py
│   └── figures/ (output Tornado charts)
│
├── Simulation_Optimization
│   ├── ExhaustiveSearch.py
│   ├── ParetoFrontSearch.py
│   ├── TabuSearch.py
│   └── __init__.py
│
├── README.md
├── requirements.txt
└── __pycache__

---

⚙️ Requirements

Prerequisites

• Python 3.8+
• Required Python libraries:
  • numpy
  • pandas
  • matplotlib
  • scipy

Install Dependencies

To install all dependencies, run:

pip install -r requirements.txt

---

🚀 Running the Code

1. ICU Queue Simulation

Simulates patient arrival, prioritization, and ICU bed allocations using different queuing strategies.

Steps:

1. Navigate to the Part_1_IcuQueue directory:

   cd Part_1_IcuQueue

2. Run the main script:

   python IcuQueueMain.py

Customizing Strategies:

• The queuing strategy is determined by the implementation in Part_1_IcuQueue.DepartureProcessWithFIFO.
• To generate results under different strategies, modify or replace the function simulate_departure_process_FIFO. For example:

To Implement Priority Queue:

 from Part_1_IcuQueue.DepartureProcessWithPQ import simultaneously_return

• The script dynamically imports the strategy through simultaneously_return.

3. Outputs:
   • Plots: Saved in Part_1_IcuQueue/figures/.
   • Penalty Results: Saved in Part_1_IcuQueue/penalties/.

---

2. Caregiver Simulation

Simulates caregiver workflows and patient service delivery, including penalties for delays.

Steps:

1. Navigate to the Part_2_CareGiver directory:

   cd Part_2_CareGiver

2. Run the main script:

   python CareGiverMain.py

Customizing Strategies:

• The queuing strategy is determined by the implementation in Part_1_IcuQueue.DepartureProcessWithFIFO.
• To generate results under different strategies, modify or replace the function simulate_departure_process_FIFO. For example:

To Implement Priority Queue:

 from Part_1_IcuQueue.DepartureProcessWithPQ import simultaneously_return

• The script dynamically imports the strategy through simultaneously_return.

3. Outputs:
   • Plots: Saved in Part_2_CareGiver/figures/.
   • Penalty Results: Saved in Part_2_CareGiver/penalties/.

---

3. Sensitivity Analysis

Analyzes the sensitivity of key parameters such as arrival rates, length of stays, and service times.

Steps:

1. Navigate to the Sensitivity_Analysis directory:

   cd Sensitivity_Analysis

2. Run the sensitivity analysis script:

   python SensitivityAnalysis.py

3. Outputs:
   • Tornado Chart: Visualizes the sensitivity indices (saved in the figures/ folder).
   • Results: Sensitivity indices and baseline penalties are displayed in the console.

---

4. Optimization Strategies

Explores resource allocation optimization using exhaustive, Pareto front, and Tabu search methods.

Steps:

1. Navigate to the Simulation_Optimization directory:

   cd Simulation_Optimization

2. Run an optimization script:

   python ExhaustiveSearch.py
   python ParetoFrontSearch.py
   python TabuSearch.py

3. Outputs:
   • Optimized resource allocation results (printed in the console).

---

📊 Key Features

Simulation Strategies

• FIFO (First-In, First-Out): Patients are served in the order of arrival.
• Priority Queue: Patients with higher severity are prioritized.
• Reserved Beds: Dedicated beds for severe cases.
• Dynamic Priority Queue (DPQ): Combines severity and waiting times for dynamic prioritization.

Sensitivity Analysis

• Parameters analyzed:
  • Patient arrival rates
  • Length of patient stays
  • Request frequencies
  • Mean service times
• Results visualized using Tornado charts.

Optimization Techniques

• Exhaustive Search: Tests all parameter combinations for global optimization.
• Pareto Front Search: Optimizes trade-offs between resource constraints (beds, caregivers).
• Tabu Search: Heuristic approach for fast, near-optimal solutions.

---

📄 Outputs

Plots

• Aggregated waiting times by severity level.
• Tornado charts for sensitivity indices.

Results

• Average penalties with confidence intervals.
• Optimized resource allocation strategies.

---

📚 References

• MIMIC-IV Dataset: PhysioNet

---

📝 License

This project is licensed under the MIT License.