SCAR dynamics of adolescent substance use: peer influence, dropout, and bifurcation structure in a school-based model

Technical Summary: SCAR Dynamics of Adolescent Substance Use

Problem Statement
Adolescent substance use is a complex public health issue driven by peer influence, protective social environments, and school-related transitions such as disengagement and re-entry. While mathematical models exist for adult populations and general addiction dynamics, comparatively few are tailored to adolescent, school-based populations where peer effects are potent and educational transitions (dropout, graduation, return) directly alter the effective population. Existing models often fail to simultaneously account for behavioral contagion (initiation, escalation) and school-specific turnover mechanisms. This paper addresses the need for a mathematically tractable framework that integrates peer-driven behavioral transitions with school disengagement and re-entry dynamics to understand how these interacting mechanisms generate qualitatively different school-level outcomes.

Methodology
The authors develop a four-compartment Susceptible–Casual–Addicted–Resistant (SCAR) model using ordinary differential equations (ODEs) adapted to a high-school setting. The population is divided into:

• $S(t)$: Susceptible students (non-users, behaviorally vulnerable).
• $C(t)$: Casual or experimental users.
• $A(t)$: Students with sustained, problematic, or Substance Use Disorder (SUD)-level involvement.
• $R(t)$: Resistant students (non-users in strong protective environments).

The model incorporates:

1. Demographic Turnover: Constant entry ($\mu$) and exit (graduation/departure) rates, with new entrants classified as susceptible ($1-\rho$) or resistant ($\rho$).
2. Peer Influence: Frequency-dependent transmission terms ($\beta XY/N$) representing risk-promoting influence (initiation from $C$ or $A$ to $S$; escalation from $C$ to $A$) and protective influence (de-escalation from $A$ to $C$ or $C$ to $S$; resistance building from $S$ to $R$).
3. Disengagement and Re-entry: A rate $\delta$ at which problematic users ($A$) disengage from school. A fraction $\phi$ of these students re-enter the school population (returning to $S$), while the remainder ($1-\phi$) leave the modeled system permanently.

The analysis employs qualitative dynamical systems theory, including:

• Proof of positivity and boundedness of solutions.
• Derivation of a scaled system describing the evolution of proportions ($x, y, z, w$) independent of total population size in certain regimes.
• Equilibrium analysis for substance-free, SUD-free, and interior (coexistence) states.
• Derivation of invasion thresholds ($R_y$ for casual use, $R_z^{(c)}$ for problematic use).
• One-parameter bifurcation analysis to examine structural changes in equilibrium stability and multistability.

Key Contributions and Results

1. Structural Regimes Defined by Re-entry ($\phi$):
   The model identifies a critical bifurcation based on the return parameter $\phi$.

   • $\phi = 1$ (Conserved Population): The total school population is constant. Positive equilibria in the scaled system correspond to true endemic equilibria in the original variables, allowing for the long-term coexistence of all behavioral classes.
   • $\phi < 1$ (Population Loss): Problematic use induces net population loss ($dN/dt < 0$). In this regime, positive equilibria of the scaled system represent "shape equilibria" (proportions) but do not correspond to true endemic equilibria in the original variables, as the total population eventually declines.

2. Distinct Threshold Mechanisms for Initiation and Escalation:
   The analysis reveals that first use and progression to problematic use are governed by separate dynamical thresholds:

   • $R_y$ (Casual Use Invasion): Determines if casual users can invade a substance-free environment. It depends on initiation pressure ($\beta_{CS}$), protective return ($\beta_{RC}$), and the baseline resistant population ($w_f$).
   • $R_z^{(c)}$ (Problematic Use Invasion): Determines if problematic users can invade an environment where casual use is already established. It depends on escalation pressure ($\beta_{AC}$), removal rates ($\mu + \delta$), and protective de-escalation ($\beta_{RA}$).
   • Implication: Preventing initiation and preventing escalation are distinct intervention problems; suppressing one does not automatically suppress the other.

3. Multistability and Bistability:
   The model exhibits multistability, specifically bistability between a substance-free equilibrium and a stable interior high-use equilibrium.

   • SUD-Free Subsystem: Even without problematic use, the system can converge to a substance-free state, a stable casual-use state, or exhibit bistability between them, depending on the balance between initiation and protective influence.
   • Full System: Numerical examples demonstrate that the full system can support bistability between a substance-free state and a stable interior state where all classes coexist. The final outcome depends on initial conditions (basin of attraction) as well as parameter values.

4. Bifurcation Analysis of Policy Levers:
   One-parameter bifurcation diagrams (for $\phi=1$) categorize parameters into three functional classes:

   • Risk-Promoting: Increasing $\beta_{AC}$ (escalation), $\beta_{AS}$ (recruitment by problematic users), or $\beta_{CS}$ (recruitment by casual users) expands the region of stable high-use equilibria. Notably, $\beta_{CS}$ acts as a structural organizer, as casual users both recruit susceptibles and supply the pool for future problematic users.
   • Protective/Recovery: Increasing $\beta_{RA}$, $\beta_{RC}$, $\beta_{RS}$, or $\rho$ (resistant entry) shrinks or eliminates the stable interior high-use branch, enlarging the basin of attraction for the substance-free state.
   • Contextual/Structural: The disengagement rate $\delta$ and turnover $\mu$ have complex effects. While increasing $\delta$ mathematically weakens the high-use branch, the authors caution that this may reflect dropout rather than recovery, highlighting the model's ability to distinguish between removal via intervention versus removal via disengagement.

Significance and Claims
The paper claims that its primary contribution is a qualitative dynamical characterization of adolescent substance use, rather than empirical estimation. The significance lies in demonstrating that school-level outcomes are governed not only by invasion thresholds but also by:

• Escalation Dynamics: The distinct mechanism of moving from casual to problematic use.
• Multistability: The existence of multiple stable states (low-use vs. high-use) for identical parameter sets, implying that initial conditions and historical context (e.g., early-use clusters) determine long-term outcomes.
• Re-entry Structure: The critical role of whether disengagement is reversible ($\phi$).

The authors argue that effective intervention requires a layered strategy addressing initiation, escalation, protective influence, and successful re-entry. They emphasize that policies must aim not only to reduce average prevalence but to keep the system away from parameter regimes and state-space regions that support sustained high-use outcomes. The framework is presented as applicable to various substances (cannabis, alcohol, nicotine) where peer influence and school transitions are central, though the specific application motivating the model is adolescent cannabis use. The paper acknowledges limitations, including the assumption of homogeneous mixing and the simplification of re-entry dynamics, suggesting future work should incorporate network structures and longitudinal data calibration.