PhD Research

Health-Related Quality of Life in Adolescent & Young Adult Cancer

A Hybrid System Dynamics, Agent-Based, and Network Modeling Approach

944 Records Triaged
237 Relevant Papers
9 Prior Works w/ Bidirectional Coupling

Research Overview

The Problem

AYA cancer patients (ages 15–39) report health-related quality of life below general-population norms for years after treatment, with the largest and longest-lasting shortfalls in physical and emotional functioning (Smith 2013, AYA HOPE; Husson 2017). Observational studies sample only a few time points and cannot represent how heterogeneity, support, and services evolve together.

The Gap

A structured triage of 944 records (237 relevant) confirms established use of system-dynamics, agent-based, network, and complex-systems methods in oncology (for example, Kenzie 2023; Stephan 2024). Yet only 9 works show bidirectional cross-method coupling, and none combine an agent-based patient model, a social-support network, and system dynamics for AYA HRQoL.

The Solution

A tri-layer hybrid architecture: System Dynamics (population and services), Agent-Based Modeling (individual heterogeneity), and an evolving social Network (peer support), coupled through six formally specified, independently switchable interfaces.

Tri-Layer Hybrid Architecture

MACRO: System Dynamics

Population feedback loops, service queue, resource constraints

dX/dt = f(X, αmacro)
SW_SA / SW_AS interfaces

MESO: Agent-Based Model

Individual patients, heterogeneous attributes, behavioral rules

Agenti = {age, dx, HRQoLi(t)}
SW_NA / SW_AN interfaces

MICRO: Network Layer

Evolving peer-support ties; per-run network size drawn from AYA evidence (Cloyes 2022)

G = (V, E, wij)

The methodological contribution is the formally specified, verified, bidirectionally coupled architecture of six switchable interfaces (SW_NA, SW_SA, SW_AS, SW_NS, SW_SN, SW_AN), plus an ablation protocol showing each layer earns inclusion. A "no-leak" verification proves each off switch removes exactly one effect.

Key Results

7.2%
Full three-layer model docking error versus single-method models (v6)
R² > 0.99
Q-PRIMA VenSim → MATLAB translation accuracy
6 / 6
Coupling interfaces pass no-leak verification
2–10
Support-network size drawn per run (mean 6, Cloyes 2022)

Note: Verification is complete. Behavioral validity is pattern-oriented. Predictive validity against an external cohort remains open under a pre-registered EORTC data request (see Publications). The model is a theory and decision-support scaffold rather than a tool for individual prognosis.

Literature Corpus

The gap claim rests on a structured, SHA-256–verified triage of the working corpus: every count below is drawn from the triage audit, and every audited source path matches the hash frozen in the corpus record.

944
Bibliographic records triaged (902 canonical after de-duplication)
237 / 234
Relevant / Marginal (473 out of scope)
245
Source paths deep-audited (SHA-256 verified)
9
Records with bidirectional cross-method coupling

Nearest prior art (cited)

The closest partial precedents from the triage, each using one or more of the same methods yet missing the AYA-HRQoL domain or confirmed bidirectional coupling.

Relevant • SD review

System dynamics modeling for cancer prevention and control: A systematic review

Kenzie E, et al.

PLoS ONE, 2023

Relevant • ABM in oncology

Agent-based approaches for biological modeling in oncology: A literature review

Stephan D, et al.

Artificial Intelligence in Medicine, 2024

Relevant • Complex systems

An Introduction to Complex Systems Science and Its Applications

Siegenfeld AF, Bar-Yam Y

Complexity, 2020

Marginal • ABM in epidemiology

Formalizing the role of agent-based modeling in causal inference and epidemiology

Marshall BDL, Galea S

American Journal of Epidemiology, 2014

Marginal • Systems thinking

Systems science and systems thinking for public health

Carey G, et al.

BMJ Open, 2015

Scope note: Counts reflect the working corpus triage as of July 2026 and describe the current state of the literature. The full corpus and the triage records live in the literature-review folder of the project.