Agile Healthcare: Overcoming Hospital Congestion via System Dynamics and DRG Intelligence

TL;DR

Hospital overcrowding isn't just a lack of beds; it's a failure of dynamic responsiveness. This study moves beyond "Lean" efficiency to propose an Agile Healthcare model. By simulating patient flows with System Dynamics and using DRG (Diagnosis Related Group) data to trigger timely discharges, the authors achieved an 8% reduction in gurney usage and slashed ED wait-times by nearly half.

Context: Why Lean is Not Enough

For years, the gold standard in hospital management has been "Lean Healthcare"—the adaptation of the Toyota Production System. However, hospitals are not assembly lines. Unlike a factory, a hospital cannot control the "input" (accidents don't follow a schedule).

The authors argue that the high variability and uncertainty of healthcare demand make the stable rates required for Lean almost impossible to achieve. Instead, they pivot to Agile Logic, a manufacturing paradigm designed for volatile demand, focusing on resource coordination and flexibility rather than just waste removal.

Methodology: The Matrix-Based Simulation

The researchers modeled the Cardarelli Hospital in Naples, Italy, using a System Dynamics (SD) approach. Unlike standard SD models that use scalar values (simple numbers), they utilized a Matrix Structure ($nimesm$). This allows the simulation to track specific features for every single patient ($n$) across various attributes ($m$), preventing information loss as patients move through the "stocks" (wards) and "flows" (admissions/discharges).

Resource Peak Management

By analyzing three years of data, the authors identified that patient arrivals follow a Poisson distribution with a distinct peak between 10 AM and 12 PM.

Fig 1: Identifying the late-morning peak allows for Agile resource shifting.

The Core Innovation: The DRG-Driven Bed Manager

The most "Academic-to-Applied" breakthrough in this paper is the role of the Bed Manager. While many hospitals have this role, the authors provide them with a mathematical "stick": the DRG Threshold.

The Diagnosis Related Group (DRG) is typically a billing tool. Here, the authors use it to calculate a "Trim Point" (upper limit of stay) using the formula: $T_{DRG}=[\sqrt[3]{q_3}+(\sqrt[3]{q_3}-\sqrt[3]{q_1}){]}^3$ (Where $q_1,q_3$ are the 1st and 3rd quartiles of stay duration).

When a patient's Bed Turnaround Time (BTT) exceeds this threshold, they are flagged as an "outlier." The Bed Manager then pressures department heads to prioritize these discharges, creating "Pull" in the system to make room for emergency arrivals.

Experimental Results & Performance

The model was validated using a Student’s t-test ($|t|=1.18<t_{crit}$), proving the simulation's accuracy against real-world hospital data.

Key Improvements:

1. Wait Times: By introducing an "Agile" medical resource during peak hours, the number of patients waiting in the visit room dropped from 23 to an average of 7.
2. Overcrowding (Gurneys): The DRG-driven pressure system reduced the average number of daily patients relegated to gurneys by 8%.

Fig 2: The complex interplay between DRG outliers and the "Level of Pressure" in the hospital.

Critical Insight: The Risk of Forced Discharge

The authors objectively note a "Risk of Discharge" variable. As the Bed Manager pushes for higher turnover, there is a non-zero risk of premature discharge. This highlights the delicate balance between operational throughput and clinical safety.

Conclusion

This work demonstrates that hospital performance is a Dynamic Actor Association. By combining the physics of system flows (SD) with the economics of healthcare (DRGs) and the flexibility of manufacturing (Agile), we can move closer to an "Elastic Hospital" that breathes with its demand rather than breaking under it.