Abstract
Traditional epidemiological models often struggle to balance computational efficiency with the granular detail required for precise policy-making. This paper introduces a novel predictive framework using high-dimensional partial differential equations (PDEs), specifically Population Balance Equations (PBEs), to track the evolution of infectious diseases. By defining the infected population across dimensions of time, geography, infection severity, duration, and age, the model provides a continuous view of an outbreak's progression. Applied to the COVID-19 pandemic in India, the framework demonstrates high prognostic accuracy and offers a robust tool for evaluating the impact of various societal interventions.
Introduction: Beyond Compartmental Models
Epidemiologists typically choose between simple compartmental models (like SIR or SEIR) and complex Agent-Based Models (ABMs). While compartmental models are computationally "cheap," they treat populations as discrete blocks, losing crucial nuances such as how age or the specific day of an infection affects transmission. ABMs are highly detailed but computationally massive, making them difficult to scale for real-time decisions.
This research proposes a "middle path" using a six-dimensional PBE framework. Originally a tool in chemical engineering for modeling particle distributions, PBEs are adapted here to treat infection as a continuum. This allows for tracking not just the number of infected individuals, but also their specific clinical needs (severity) and expected recovery times (duration).
Methodology: The 6D Population Balance Equation
The core innovation is a scalar field representing the infected population across six dimensions:
- Time ($t$)
- 2D Geography ($x, y$)
- Infection Severity ($l_v$): Ranging from asymptomatic to critical.
- Duration of Infection ($l_d$): Time since the individual was first exposed.
- Population Age ($l_a$): Crucial for COVID-19 due to age-varying mortality rates.
The model incorporates "nucleation" for new infections and "growth factors" for disease progression. To maintain efficiency, an operator-splitting finite element scheme is employed, breaking the high-dimensional problem into simpler sub-problems solved sequentially.
Key Results and Scenario Analysis
The framework was validated using the COVID-19 outbreak in India, incorporating national demographics and clinical parameters. A primary strength of the model is its ability to simulate various policy "knobs," such as interactivity, hygiene, and social distancing.
Flattening the Curve
The model compared a "Current Trend" scenario against various lockdown strategies. As shown in the projections for active cases, specific interventions were predicted to significantly flatten the curve. For instance, certain strategies could reduce peak active cases from nearly 1 million to roughly 200,000.
Prognostic Accuracy
The model's early forecasts closely aligned with actual reported data for total infections and recoveries. By tracking the "internal space" of the infection, the model provides a detailed outlook on the total burden on the healthcare system over time.
Discussion and Future Directions
While the model demonstrates significant predictive power, its accuracy is inherently linked to the quality of reported epidemiological data. Future work aims to integrate real-time data assimilation to update the model dynamically. Additionally, incorporating more granular coordinates like mobility history could further refine predictions for reopening specific sectors of the economy.
By bridging the gap between simple continuum models and complex discrete simulations, this PBE-based framework provides a scalable, science-informed tool for public health policy planning.
Summary updated