🔢 mathematics

A Novel Compartmental Model for the Mathematical Epidemiology of Addiction: Backward Bifurcation, Bistability, Persistence, and Treatment-Capacity Policy Implications

This paper develops a novel four-compartment epidemiological model for addiction dynamics that demonstrates how treatment-capacity saturation, rather than incidence saturation, uniquely drives backward bifurcation and bistability, creating a critical policy gap where the basic reproduction number fails to capture the risk of long-term persistence even below traditional thresholds.

Original authors: Milad tahavor, Reza Memarbashi, Khashayar Ashkboos, Mohammad Mehdipour Zarrinkamar

Published 2026-08-25

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Milad tahavor, Reza Memarbashi, Khashayar Ashkboos, Mohammad Mehdipour Zarrinkamar

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

For decades, scientists have used a specific kind of mathematical map to understand how diseases spread through a population. These maps divide people into groups: those who can catch an illness, those who are currently sick, and those who have recovered. By tracking how people move between these groups, researchers can calculate a single number that predicts whether an outbreak will die out or grow. If this number is below one, the standard rule has been that the disease will eventually vanish. This logic has guided public health decisions for everything from flu to measles. However, when researchers apply this same logic to addiction, the rules change. Addiction is not a virus passed from person to person in the same way; it is a behavior that can be influenced by social pressure, limited access to help, and the tendency to return to old habits even after a period of recovery. Because of these unique factors, the simple rule that "below one means safe" may not hold true for addiction epidemics.

A team of researchers from Iran has built a new, more detailed map to explore exactly how these factors interact. Their work focuses on a four-part system: people who have never used drugs, those actively struggling with addiction, those currently in formal treatment, and those who have recovered but are not currently using. The researchers introduced two critical real-world complications into their model that previous studies often ignored. First, they accounted for the fact that treatment centers have limited space. When too many people seek help at once, the system becomes overwhelmed, and the rate at which people can enter treatment slows down, regardless of how many people are waiting. Second, they included a direct path for relapse. In many models, a person who relapses must first return to being a "new" user before becoming addicted again. In reality, a person who relapses often returns immediately to active addiction, skipping the intermediate stage.

By combining these realistic constraints with a model of how social influence spreads addiction, the researchers discovered a startling phenomenon known as backward bifurcation. In a standard epidemic, if you lower the transmission rate enough to get the reproduction number below one, the disease disappears. In this new model, however, the system can get stuck in a high-addiction state even when that number is below one. The researchers proved mathematically that if the capacity for treatment is too tight, the system creates a "trap." In this trap, two stable outcomes exist at the same time: a world with almost no addiction and a world with a high, persistent level of addiction. Which world a society finds itself in depends not just on the current number of users, but on the history of the system. If the number of active users is high enough to push the system over a certain invisible threshold, it will remain in the high-addiction state even if the conditions for spreading addiction improve.

The study identifies the bottleneck in treatment capacity as the sole cause of this dangerous trap. The researchers found that the standard number used to measure risk is completely blind to this problem. You could have a treatment system that is severely overcrowded and prone to collapse, yet the standard risk number would remain unchanged because it is calculated based on how the system behaves when few people are sick. This means that a public health official could look at the data, see that the risk number is low, and believe the epidemic is under control, while the system is actually teetering on the edge of a permanent crisis. The only way to detect this danger is to look directly at the saturation of treatment centers, not just the overall spread rate.

To test how these findings play out in the real world, the team calibrated their model using data on heroin use in the United States. They adjusted the model to match the number of users reported in national surveys from 2013 and 2022. The results revealed a "ghost" of a crisis. Even though the model predicted that the addiction rate should eventually drop to zero, the decline is so incredibly slow that it would take nearly two centuries for the number of users to fall by half. On the scale of a human lifetime or a government budget cycle, the epidemic appears to be permanent and growing, even though the math says it is slowly fading. This suggests that the current situation is not a stable epidemic, but a long, slow-motion transition that feels like a permanent state to anyone living through it.

The researchers also explored what happens when you add the randomness of real life, such as the chance that a few people might relapse or recover unexpectedly. In a perfect mathematical world, there is a sharp line separating the path to recovery from the path to addiction. In the real world, this line blurs into a smooth gradient. A population sitting near the edge of this boundary might be pushed toward recovery by a small change, or pushed back into addiction by a minor shock. This means that policy decisions cannot rely on a simple "yes or no" calculation of where a population stands; they must account for the probability of shifting between states.

Finally, the team used their model to design the most efficient way to spend money on prevention and treatment. They found that the best strategy involves a burst of intense action at the beginning. Pouring resources into expanding treatment capacity and prevention efforts immediately can push the system out of the dangerous high-addiction trap. Once the number of users drops below a critical point, the intensity of the intervention can be reduced. However, the study warns that simply telling more people to seek treatment without actually expanding the number of available beds or counselors will not work. If the system is already full, increasing the demand for help only makes the bottleneck worse. The only way to break the cycle is to increase the actual capacity of the treatment system, removing the saturation that creates the trap in the first place. This research provides a clear, mathematical reason why some addiction epidemics seem impossible to solve with standard methods and offers a specific path forward that focuses on the physical limits of the help available.

Technical Summary: A Novel Compartmental Model for the Mathematical Epidemiology of Addiction

Problem Statement
Existing mathematical models of addiction epidemiology often rely on linear transfer rates and fail to capture two empirically critical features of real-world treatment systems: (1) Limited treatment capacity, where the rate of entry into treatment saturates as the addicted population grows due to waiting lists and fixed budgets; and (2) Direct relapse, where recovered individuals return directly to active addiction rather than reverting to a fully susceptible state. The absence of these mechanisms in standard models obscures the potential for backward bifurcation, a phenomenon where reducing the basic reproduction number ( $R_0$ ) below unity is insufficient to eliminate addiction, as the system may remain trapped in a stable, high-prevalence endemic state.

Methodology
The authors develop a four-compartment deterministic cohort model ( $S, A, T, R$ ) representing Susceptible, Actively Addicted, In Treatment, and Recovered populations. The model incorporates three novel nonlinear/structural ingredients:

Saturating incidence: $\frac{\beta SA}{1+\alpha A}$ , modeling bounded exposure capacity.
Capacity-limited treatment entry: $\frac{\sigma A}{1+\gamma A}$ , where $\gamma$ measures the severity of capacity constraints.
Direct relapse: A pathway from $R$ directly to $A$ at rate $\psi$ .

The study employs a rigorous mathematical toolkit:

Equilibrium Analysis: Derivation of the Addiction-Free Equilibrium (AFE) and endemic equilibria, reducing the system to a quadratic equation to analyze root existence.
Stability Theory: Local stability analysis via the next-generation matrix method ( $R_0$ ) and global stability proofs using a Metzler-matrix comparison argument and Lyapunov functions.
Bifurcation Theory: Application of the Castillo-Chavez-Song center-manifold theorem to derive an exact, closed-form criterion for backward bifurcation.
Sensitivity Analysis: Both local (normalized sensitivity indices) and global (Sobol indices and PRCC) analyses to determine parameter influence on $R_0$ and the bifurcation coefficient.
Extensions: Stochastic diffusion approximation (Itô SDE) to assess demographic noise effects, and an optimal control formulation (Pontryagin's Maximum Principle) for treatment capacity and prevention investment.
Empirical Calibration: Fitting the model to U.S. NSDUH heroin-use data (2013–2022) to test the relevance of the theoretical findings.

Key Contributions and Results

Closed-Form Bifurcation Criterion: The paper derives an exact expression for the bifurcation coefficient $a$ . The analysis reveals that treatment-capacity saturation ( $\gamma$ ) is the sole source of bistability (backward bifurcation), while incidence saturation ( $\alpha$ ) is strictly stabilizing. The coefficient decomposes as:
$a = 2\beta^*v_2w_1w_2 - \frac{2\Lambda\alpha\beta^*}{\mu}v_2w_2^2 + 2\gamma\sigma w_2^2(v_2 - v_3)$
This proves that backward bifurcation occurs if and only if $\gamma > 0$ and the relapse pressure is sufficient ( $v_2 > v_3$ ).
The "Invisibility" of Capacity to $R_0$ : A central theoretical finding is that the basic reproduction number $R_0$ is independent of the treatment capacity parameter $\gamma$ ( $\partial R_0 / \partial \gamma \equiv 0$ ). Consequently, standard surveillance monitoring $R_0$ cannot detect the risk of a capacity-driven bistability trap. $R_0$ is governed by transmission and demographic rates, while the type of bifurcation (forward vs. backward) is governed by $\gamma$ and $\psi$ .
Global Stability and Persistence:
- A sufficient condition for the global stability of the AFE is established via a Metzler-matrix comparison argument, showing that linear bounds fail precisely in the regime where bistability occurs.
- A uniform persistence theorem is proven for $R_0 > 1$ , ensuring addiction cannot die out from interior initial conditions.
- For the forward bifurcation regime ( $a < 0$ ), a global stability result for the unique endemic equilibrium is provided, supported by a hand-proven Lyapunov function for the incidence component and computational certification for the treatment/relapse cycle.
Saddle-Node "Ghost" Mechanism: Calibration to U.S. heroin data reveals a scenario where $R_0 < 1$ but $R_0 > R_0^c$ (the saddle-node threshold). In this regime, the endemic equilibria have vanished, yet the system exhibits a "ghost" of the saddle-node bifurcation. This results in centuries-long quasi-endemic transients (e.g., a 20% reduction in prevalence over 150 years), making the population appear stable and endemic on any realistic policy horizon, despite the theoretical attractor being the disease-free state.
Stochastic Robustness: The deterministic basin boundary between the disease-free and endemic states is shown to become a smooth probability gradient under demographic noise. In finite populations, the probability of reaching the endemic branch is a continuous function of initial conditions, rather than a hard threshold.
Optimal Control: An optimal control problem is formulated to minimize active addiction and intervention costs. Results indicate that while optimal control can reduce the burden of an outbreak, it does not alter the bifurcation regime itself unless the control explicitly targets the saturation parameter $\gamma$ (e.g., by removing queueing frictions rather than just adding slots).

Significance and Policy Implications
The paper argues that the standard policy metric of reducing $R_0 < 1$ is insufficient for addiction control when treatment capacity is saturated. The authors claim:

$R_0 < 1$ is not sufficient: In the bistable window ( $R_0^c < R_0 < 1$ ), the system can remain trapped in a high-addiction state. Elimination requires either a massive one-time reduction in prevalence (pushing the system below the basin boundary) or structural changes that reduce $R_0$ below the critical threshold $R_0^c$ .
Capacity Policy is Invisible to $R_0$ : Expanding treatment capacity (reducing $\gamma$ ) does not lower $R_0$ ; its effect is entirely on the shape of the bifurcation diagram. Therefore, surveillance systems tracking only $R_0$ will fail to detect approaching capacity-driven collapse.
Dual Benefits of Awareness: Increasing awareness (increasing $\alpha$ ) provides a dual benefit: it lowers $R_0$ and simultaneously stabilizes the system against backward bifurcation.
Structural vs. Quantitative Interventions: Merely increasing the rate of treatment seeking ( $\sigma$ ) without expanding actual capacity ( $\gamma$ ) moves the system along the backward branch but does not eliminate the bistability trap. Only reducing $\gamma$ can cross the boundary to a forward-bifurcation regime.

The study concludes that effective addiction policy must monitor the safety margin $R_0/R_0^c$ and address structural capacity constraints directly, rather than relying solely on transmission reduction or treatment-seeking promotion.

Technical Summary: A Novel Compartmental Model for the Mathematical Epidemiology of Addiction

More like this