Health-Related Quality of Life Modeling of Adolescent & Young (AYA) Adult Cancer Patients

Abstract

Background. Health-related quality of life (HRQoL) in adolescent and young adult (AYA) cancer patients is worse than population norms, with the largest deficits in physical and emotional functioning, and persists for years after diagnosis^{[1, 2]} Clinical and observational studies capture HRQoL at limited time points and cannot represent the joint, evolving influence of patient heterogeneity, social support, and service-system feedback.

Methods. A tri-layer hybrid simulation model couples an agent-based model of individual patients, an evolving social-support network, and a system-dynamics model of population and services. The three layers exchange information through six explicit interfaces, each of which can be disabled independently to measure the contribution of each interface. No prior model is known to couple all three layers for AYA HRQoL^{[16, 17]}

Results. Over 20 simulation seeds and a 24-month horizon, the model reproduces patterns it was not fitted to, including a treatment nadir, persistent emotional deficits, and a durable-remission fraction near 78% among treated agents, within the band anchored to AYA 5-year relative survival^{[1, 5]} Each interface acts in the direction predicted by the governing mechanism, with effect sizes that differ sharply: social support dominates the outcome while the population-feedback loop is near-dormant under baseline parameters. All interface verification checks pass, including a no-leak test, a structural verification method used to ensure that stocks (accumulations) do not gain or lose material spontaneously. It checks that all inflows and outflows are properly accounted for and that no hidden or "leaking" flows exist^[9]

Conclusions. The model stands as an exploratory scaffold for theory and decision support. Predictive validation against a longitudinal AYA cohort remains open. OPEN

Model card

Purpose, anchoring evidence, and stated limitations, each expanded in the sections that follow.

Background

Understanding how HRQoL evolves after an AYA cancer diagnosis is essential to improving survivorship care. Three findings define the burden this work addresses:

• HRQoL is worse than population norms, with the largest deficits in physical and emotional functioning^[1]
• These deficits persist for years after diagnosis, not only during treatment^[2]
• Most computational models of cancer represent tumor biology or care logistics^{[18, 19]}, not the joint effect of individual heterogeneity, an evolving support network, and population/service feedback

Clinical and observational studies capture HRQoL at limited time points and cannot represent these factors as they interact and evolve. Simulation can. Most simulation work, however, couples only one or two paradigms; the difficulty, and the contribution here, is the interfaces between three. This work develops a tri-layer hybrid model and tests whether each layer earns inclusion.

Empirical basis: incidence, age, and diagnosis

The model targets the adolescent and young adult (AYA) population and distinguishes outcomes by diagnosis. The first three figures give the empirical basis from published evidence. The fourth is a model output: cancer type enters the verified simulation only through cited inputs (5-year survival and treatment intensity), and HRQoL emerges by diagnosis. Inputs taken from the project AYA dataset are literature-informed estimates flagged for provenance verification, so these figures are exploratory; comparisons across pediatric and older-adult groups stay empirical, since the simulation spans ages 15-39.

Methods: three autonomous layers, six interfaces

The model comprises three layers, each with a separate state and separate update rules. The layers exchange information through six explicit interfaces, and each interface can be disabled independently, which allows the contribution of each one to be measured.

Conceptual framing: how this differs from the usual hybrid

Hybrid simulation most often nests system dynamics inside agents: each agent carries an internal set of continuous dynamics, and the agents interact, often over a network. Population structure and finite shared services are then implicit in the aggregate of agent states. This model retains per-agent dynamics but also promotes the support network and a finite-capacity services subsystem to peer layers, coupled through six explicit, switchable interfaces, so the contribution of each can be isolated and measured. A fuller treatment is in docs/CONCEPTUAL_FRAMING.md.

• SW_NA: social support buffers psychological wellbeing (network to patient)^[6]
• SW_AN: patient withdrawal dissolves ties (patient to network)
• SW_AS: patient distress raises referral demand (patient to services)
• SW_SA: service access and population stress feed back to wellbeing (services to patient)
• SW_SN: program capacity creates new ties (services to network)
• SW_NS: ties diffuse referrals (network to services)

What each layer owns

Feedback structure (causal loop diagram)

The feedback structure is the essential part of the coupling. The causal loop diagram below renders it in the convention used by system-dynamics articles: named variables, causal links carrying a plus or minus polarity, and labeled balancing (B) and reinforcing (R) loops. A stock-and-flow view of the services layer is in the diagrams section, and a guide to constructing such interfaces is in docs/CONSTRUCTING_INTERFACES.md.

Network model and node definition

What the nodes are

Each node is an AYA patient-agent, the same agents as the ABM layer; an edge is a peer support tie between two patients. Support reaches a patient through the single N→A operator, built from two literature-grounded parts. The first is structural emotional support set by network closure: the local clustering coefficient of a patient sets how much emotional and esteem support that position affords, since closure (densely interconnected contacts) predicts emotional and esteem support, while brokerage predicts informational support^{[59, 60, 61]} The term is weighted by tie strength, since strong, multiplex ties carry emotional aid,^[62] and saturates with connectedness, the diminishing returns implied by the dose-response between network size and outcomes^[6] The second, smaller part is complex-contagion reinforcement: wellbeing lifts only when several neighbors are thriving, so the effect travels along clustered, wide-bridge ties rather than single shortcuts, the form social reinforcement takes for behavioral and psychosocial spread^{[58, 63]} Referral diffusion runs the opposite direction through brokerage: the N→S interface routes help-seeking along non-redundant broker ties, since brokerage carries informational support while closure carries emotional support^{[59, 60]} An isolated patient (degree zero) carries a small penalty.

The simplification is deliberate. Real AYA support systems are heterogeneous and ego-centric, but small: a typical support network holds roughly 2 to 10 members (mean about 6), about 40% family and 60% non-family, alongside 3 to 5 clinicians^{[70, 6, 20]} Each run draws a network size from this distribution (mean 6, SD 2.2, range 2 to 10), so the simulated cohort spans sparsely to well connected patients rather than sharing one fixed degree. The current model collapses family, peer, clinician, navigator, and digital ties into one homogeneous patient-peer tie type. Peer structure is the AYA-distinct, intervention-relevant channel, through peer-mentoring programs,^[21] so it is modeled first; a typed support network is identified as future work.

Why a small-world (Watts-Strogatz) network

The default initial topology is a Watts-Strogatz small-world graph^[55] Real social networks combine high local clustering with short global path lengths, the small-world signature, which neither a regular lattice (clustered but distant) nor an Erdos-Renyi random graph (short but unclustered) reproduces^[57] Because emotional support here is driven by network closure and the wellbeing-contagion term is a complex, reinforcement contagion, both of which reward clustering, the choice of topology now changes the outcome^{[59, 63]} Both mechanisms favor more clustering, so the ring lattice, the most clustered graph, scores highest. The small-world topology is used regardless, because it matches the realistic structure of human social networks, high clustering with short paths^[55] The comparison below holds mean degree constant across topologies to isolate the effect of network structure.

Extension: a typed support network EXPLORATORY

The verified model uses one homogeneous patient-peer tie. A typed-support extension replaces it with an ego-network of providers for which tie strength and support profile are set by relationship type: family and partners are the strongest emotional sources for AYAs, clinicians carry informational support, and social workers and navigators carry instrumental support^{[59, 62]} Provider nodes have no HRQoL and are colored by type; patients are colored by HRQoL. The tie-type weights are literature-informed estimates flagged for provenance, so this extension is exploratory and is not part of the calibrated core.

Data sources feeding the model

Model parameters, the observation layer, and the validation targets are anchored to published evidence and public datasets. The diagram below traces each source to the model component it informs, the table lists what each one contributes, and full citations appear in the references section.

How the evidence was aggregated

Six categories of evidence, spanning public datasets, longitudinal cohort studies, mechanism and theory literature, validated instruments, care guidelines, and reference models, feed the model components they inform. The map below shows the relative contribution of each category. The breakdown that follows adds the provenance-tier composition, the source counts, and a worked example of how several sources reconcile into model terms.

Parameter provenance

Beyond the data sources above, every model parameter and embedded constant is traced to a data source, a paper, or an explicit assumption in the provenance register (docs/PARAMETER_PROVENANCE.md); the model code carries inline [Cited] / [Derived] / [Assumption] tags that point to it. Values fall into three tiers.

The full parameter table, the stock / flow / auxiliary equation listing, and a written justification for every assumption are in the Parameters & equations section below.

About half of the values are calibration or structural assumptions, and several are swept in the sensitivity and stress experiments. The Derived sources were retrieved and checked (author, year, and the cited value or finding confirmed). One parameter is flagged as weakest-supported: the tie-dissolution rate rests on qualitative withdrawal reports,^[31] whereas the strongest longitudinal evidence finds networks net-expand after diagnosis,^[68] a pattern the model in fact reproduces. The equal-weight composite is a known simplification relative to the regression-weighted SF-12 summary scores.

Parameters & equations

The model is specified in the stock / flow / auxiliary form used by published system-dynamics models, so the structure can be read and audited directly. The temporal design is diagnosis-anchored (the clock starts at diagnosis) over a 24-month horizon, with fast patient dynamics integrated every half day and the service queue and network restructuring resolved monthly. Every parameter traces to a source or a flagged assumption, and each assumption carries a written justification. A workbook in a four-sheet layout available: AYA_parameters_and_equations.xlsx.

Run specifications

How the timeline compares with other SD models

The diagnosis-anchored, population-level, hybrid design departs deliberately from an event-anchored single-patient pure-SD model such as Q-PRIMA (Hasgul 2024)^[16] Full discussion and the complete equation listing are in docs/MODEL_TIMELINE_AND_EQUATIONS.md.

State variables (stocks)

ABM patient layer NET network layer SD services layer

Derived (not integrated): HRQoL Q = (P + Ψ)/2; closure Ci (recomputed monthly); in_remission (a latch).

Flows (rates)

The six switchable interfaces are the cross-layer auxiliaries in bold: N→A support, S→A stress, A→S demand, N→S brokerage, S→N formation, A→N dissolution. Setting a switch to 0 removes exactly one of these, which is the basis of the ablation ladder.

Parameters and provenance

Cited value or narrow range from a source Derived mechanism/range cited, value calibrated Assumption no direct anchor; each value carries a written justification

Full table (~50 values) and the inline prov: code tags are in docs/PARAMETER_PROVENANCE.md.

Assumption justifications

An assumption marks a value with no source for the exact number. Each one is justified in writing and classified by the type of support behind it. Several wellbeing-core choices inherit structure directly from the published Q-PRIMA model^[16]

Pattern-oriented means a value is not free but jointly constrained with others to reproduce several observed patterns at once (baseline distribution, nadir timing, recovery plateau, persistent tail), the standard defense in pattern-oriented modeling^[7] Swept parameters carry a further protection: the conclusions hold across a range of values rather than resting on one tuned number. Full write-up: docs/ASSUMPTION_JUSTIFICATIONS.md.

Model implementation

The model is implemented in plain MATLAB (R2024b, no toolboxes). Four excerpts show the design decisions that keep the layers separable and the validation sound.

1. Each interface is a binary switch, so the ablation experiment can turn any one of them off in isolation (coupling_switches.m).

% {name, abm_on, AS, SA, AN, NA, SN, NS, description}
'C1',  true, 0,0,0,0,0,0, 'ABM-only (heterogeneity baseline)'
'C3',  true, 0,0,1,1,0,0, 'ABM + Network'
'C4',  true, 1,1,1,1,1,1, 'Full model (all six interfaces on)'

2. Social support reaches a patient through exactly one channel, so disabling that channel removes the effect completely (AYA_MVP_Coupled.m).

% N->A : social support (THE ONLY support channel; no leak)
support = zeros(N,1);
if sw.SW_NA
    deg = sum(A,2);  WA = A .* W;
    support = baseline_sup + contagion + buffer;   % connection, contagion, buffering
    support(deg==0) = -0.01;                        % isolation penalty
end

3. System dynamics holds a real integrated stock, a referral queue, rather than a simple count of agents in a category.

Wq = Wq + demand;                       % inflow: referrals from distressed agents
served = min(Wq, Cpsy);                  % outflow bounded by service capacity
Wq = max(0, Wq - served);
service_avail = served / max(demand,1); % feeds back to patients via SW_SA

4. Verification: turning a coupling off must change the outcome, proving no hidden proxy leaks (verify_interfaces.m)^[9]

cOff = cOn;  cOff.SW_NA = false;
v9 = (run_mean(cOn,1:15) - run_mean(cOff,1:15)) > 1e-3;   % PASS = separable

Model diagrams

Three UML views document the implementation at the level a reader needs to reproduce or audit it: an activity diagram for control flow within one run, a sequence diagram for the order in which the six interfaces fire over a monthly cycle, and a component diagram for the static structure of the code. A fourth view, a stock-and-flow diagram, documents the internal mechanics of the system-dynamics layer. All are rendered from editable sources in website/diagrams/ (provided as both .drawio and .xml) and use a fixed layer color key: blue for the ABM (patient) layer, green for the network (support) layer, and orange for the system-dynamics (services) layer.

Results

The model was run over 20 simulation seeds and a 24-month horizon. Inference operates at the level of the simulation run rather than the individual agent.

Headline results

• 8 of 8 macro trend bands in range, with bands derived from the AYA HRQoL literature^{[1, 2, 6]}
• 4 of 4 held-out instrument targets in range across QLQ-C30, SF-12, and PROMIS;^{[1, 3, 4]} the PROMIS band rests on a low-confidence anchor flagged for source verification, so that one target is the weakest of the four
• About 78% of treated agents reach durable remission by 24 months (20-seed mean), near the lower bound of the band anchored to AYA 5-year relative survival (~86%)^[5]
• The social-support interface dominates the outcome (about 0.020 HRQoL); the population-feedback interface is dormant at baseline but accounts for up to 0.35 HRQoL under population stress and constrained capacity
• All interface verification checks pass, including the no-leak test^[9]

Simulation dynamics (animated)

The animations are rendered directly from a single full-model run (cell C4, all six interfaces on, seed 42) over the 24-month horizon. The first depicts the whole simulation; the rest isolate each layer and the interfaces.

The coupled tri-layer model

The three layers over time

The six interfaces over time

Validation and credibility

With sparse, non-equivalent data, a model like this cannot be validated the way a prediction model is. The argument separates four claims and states which are established and which remain open.

Iterative validation: an informative failure

An early held-out instrument test failed on emotional functioning, passing 1 of 4 targets. The failure was diagnostic: it located a missing response-shift adaptation mechanism^{[10, 11]} Adding that one theory-grounded mechanism brought the test to 4 of 4 across a wide parameter range. Because the mechanism was added in response to the failure, those bands now serve as development targets rather than independent validation. Genuine predictive validation requires new external data.

Objections and responses

Full argument, all objections, and references: MODEL_CREDIBILITY_AND_VALIDATION.md.

Comparison to pre-existing models

The exercise below is comparative validation by docking: the novel hybrid is placed alongside pre-existing system-dynamics and agent-based models from the AYA HRQoL literature and reduced to the scope of those models. It is a confidence check, not a change to the model. The model code and every result elsewhere on this page are unchanged; the benchmark only reads the model and compares it.

Five arms were run over the same diagnosis-anchored 24-month horizon, each at a separately documented initialization: an SD-only model (with a common-initialization variant), an ABM-only ablation, an older ABM + static-network hybrid, and the full hybrid. Two literature models are structural comparators only: Q-PRIMA (the published implementation is incomplete and will not run as shipped) and Beaulieu (2022), which is a causal loop diagram rather than a runnable model.

Where this sits in the field. A 2025 PRISMA systematic review of system-dynamics cancer models (Chen et al., Journal of Medical Economics) identified sixteen studies. All are single-paradigm system dynamics, most addressing screening, prevention, or treatment operations at the population level, and produce cases, deaths, quality-adjusted life years, and costs rather than an individual health-related quality-of-life trajectory. None couples an agent-based patient layer, an evolving support network, and a finite-service subsystem, and none tracks HRQoL across survivorship. This architecture addresses that gap. Because those models compute different quantities, they serve as structural comparators only; the graph above does not plot them as additional lines.

Reproducible as a standalone folder: benchmark_comparison/ with run_benchmark(20). Details in benchmark_comparison/README.md, the anchor trace in ANCHOR_RECONCILIATION.md, and the structural comparison in tables/structural_docking.csv.

Future work

The model is a verified prototype with an open predictive-validation claim and several scoped limitations. The priorities below would close those gaps, ordered from the most decisive (external validation) to incremental refinements.

Predictive validation

• Secure a longitudinal, multidomain, multi-instrument AYA HRQoL cohort, or a harmonized AYA HOPE to SURVAYA crosswalk, for a pre-registered held-out predictive test^{[1, 3]}
• Test measurement invariance across age, sex, tumor group, country, and survivorship phase, and split calibration from validation by source, domain, and horizon^[4]
• Report interval coverage and distributional metrics alongside point trajectories

Diagnosis and age

• Replace the flagged project-dataset estimates (per-type survival, treatment mix, age-group HRQoL) with certified SEER and EORTC extractions, which would remove the exploratory label from the diagnosis figure^{[5, 15]}
• Encode diagnosis-specific late-effect burden, for example brain/CNS neurocognitive effects and anthracycline cardiotoxicity, so that high-late-effect cancers separate from survival alone
• Add pediatric (under 15) and older-adult comparator arms, so the AYA paradox becomes a model output rather than an empirical comparison

Model structure and scope

• Restore long-term between-person dispersion, which is currently narrower than the SURVAYA long-term distributions^[3]
• Add a dual-network extension: an intrapersonal HRQoL-domain network alongside the interpersonal support network
• Ground the typed support network (an exploratory version is shown above) in data: replace the literature-informed tie-type weights for family, peer, clinician, social worker, navigator, and digital ties with values estimated from longitudinal AYA ego-network data, and integrate it into the calibrated core^{[6, 20, 62]}
• Implement NCCN-concordant counterfactuals (distress screening, financial navigation) as capacity-constrained service scenarios^[30]
• Extend the horizon from 24 months to 5 to 10 years of survivorship

Verification, docking, and uncertainty

• Add extreme-condition and conservation tests to the verification harness^[9]
• Conduct formal quantitative docking against a published system-dynamics HRQoL model and a published network result^{[8, 16, 17]}
• Run global sensitivity and identifiability analysis on the coupling parameters^[12]
• Resolve the flagged provenance parameters and expand the provenance register

Collaboration

A model of this scope needs data and domain expertise that no single project holds. The project seeks:

• Longitudinal AYA HRQoL data, any instrument, to enable genuine predictive validation
• Domain expertise in AYA psycho-oncology, survivorship care, social work, and financial toxicity, to refine mechanisms and face validity
• Measurement expertise for harmonizing United States and EORTC instruments through crosswalks and invariance testing
• Health-systems partners to ground the NCCN distress-screening and financial-navigation scenarios in real capacity constraints

Studies traced, and the specific data requested

The figure traces the studies and datasets this project already relies on (due diligence, each tied to the parameter-provenance register) and identifies the one dataset requested to close the open validation question: individual-level EORTC QLQ-C30 longitudinal data in AYA patients. An EORTC data-request has been placed to compare individual-level, longitudinal quality-of-life data against the tri-layer model.

Questions? Contact: jblochb2@jhu.edu

References

1. A. W. Smith et al., "Health-related quality of life of adolescent and young adult patients with cancer in the United States: the Adolescent and Young Adult Health Outcomes and Patient Experience study," Journal of Clinical Oncology, vol. 31, no. 17, pp. 2136–2145, 2013
2. O. Husson et al., "Health-related quality of life in adolescent and young adult patients with cancer: a longitudinal study," Journal of Clinical Oncology, vol. 35, no. 6, pp. 652–659, 2017
3. Janssen et al., "Health-related quality of life of long-term adolescent and young adult cancer survivors: the SURVAYA study," Journal of Cancer Survivorship, 2025
4. A. Hinz et al., "European reference values for the quality of life questionnaire EORTC QLQ-C30," Acta Oncologica, vol. 53, no. 7, pp. 958–965, 2014
5. Surveillance, Epidemiology, and End Results (SEER) Program, National Cancer Institute, "Cancer stat facts: adolescent and young adult (AYA) cancers (ages 15–39)," https://seer.cancer.gov/ (5-year relative survival)
6. M. Pinquart and P. R. Duberstein, "Associations of social networks with cancer mortality: a meta-analysis," Critical Reviews in Oncology/Hematology, vol. 75, no. 2, pp. 122–137, 2010
7. V. Grimm et al., "Pattern-oriented modeling of agent-based complex systems: lessons from ecology," Science, vol. 310, no. 5750, pp. 987–991, 2005
8. R. Axtell, R. Axelrod, J. M. Epstein, and M. D. Cohen, "Aligning simulation models: a case study and results," Computational and Mathematical Organization Theory, vol. 1, no. 2, pp. 123–141, 1996
9. R. G. Sargent, "Verification and validation of simulation models," Journal of Simulation, vol. 7, no. 1, pp. 12–24, 2013
10. M. A. G. Sprangers and C. E. Schwartz, "Integrating response shift into health-related quality of life research: a theoretical model," Social Science & Medicine, vol. 48, no. 11, pp. 1507–1515, 1999
11. K. M. Bellizzi et al., "Positive and negative psychosocial impact of being diagnosed with cancer as an adolescent or young adult," Cancer, vol. 118, no. 20, pp. 5155–5162, 2012
12. S. Bankes, "Exploratory modeling for policy analysis," Operations Research, vol. 41, no. 3, pp. 435–449, 1993
13. Y. Barlas, "Formal aspects of model validity and validation in system dynamics," System Dynamics Review, vol. 12, no. 3, pp. 183–210, 1996
14. G. E. P. Box, "Science and statistics," Journal of the American Statistical Association, vol. 71, no. 356, pp. 791–799, 1976
15. N. K. Aaronson, S. Ahmedzai, B. Bergman, et al., "The European Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for use in international clinical trials in oncology," Journal of the National Cancer Institute, vol. 85, no. 5, pp. 365–376, 1993
16. Z. Hasgul, A. Spanjaart, S. Javed, A. Akhavan, M. J. Kersten, and M. S. Jalali, "Health-related quality of life dynamics: modeling insights from immunotherapy," Quality of Life Research, vol. 34, no. 1, pp. 273–286, 2025
17. E. Beaulieu, A. Spanjaart, A. Roes, B. Rachet, S. Dalle, M. J. Kersten, D. Maucort-Boulch, and M. S. Jalali, "Health-related quality of life in cancer immunotherapy: a systematic perspective, using causal loop diagrams," Quality of Life Research, vol. 31, pp. 2357–2366, 2022
18. E. S. Kenzie, M. Seater, W. Wakeland, G. D. Coronado, and M. M. Davis, "System dynamics modeling for cancer prevention and control: a systematic review," PLOS ONE, vol. 18, no. 12, art. e0294912, 2023
19. A. Gjerloev, S. Crowe, C. Pagel, Y. Jani, and L. Grieco, "A systematic review of simulation methods applied to cancer care services," Health Systems, vol. 13, no. 4, pp. 274–294, 2024

Data sources and model-building bibliography

References grounding the parameters, instruments, methodology, and validation of the model, drawn from the project literature review and the source papers used to build the model.

AYA cancer HRQoL: field, outcomes, and policy

• S. C. Sodergren et al., "Systematic review of the health-related quality of life issues facing adolescents and young adults with cancer," Quality of Life Research, vol. 26, no. 7, pp. 1659–1672, 2017
• B. Zebrack et al., "Quality cancer care for adolescents and young adults: a position statement," Journal of Clinical Oncology, vol. 28, no. 32, pp. 4862–4867, 2010
• J. M. Yeh et al., "Life expectancy of adult survivors of childhood cancer over 3 decades," JAMA Oncology, 2020
• A. W. Smith et al., "Next steps for adolescent and young adult oncology workshop: an update on progress and recommendations for the future," Cancer, vol. 122, no. 7, pp. 988–999, 2016
• S. J. Nass et al., "Identifying and addressing the needs of adolescents and young adults with cancer: summary of an Institute of Medicine workshop," The Oncologist, vol. 20, no. 2, pp. 186–195, 2015
• Adolescent and Young Adult Oncology Progress Review Group, "Closing the gap: research and care imperatives for adolescents and young adults with cancer," U.S. DHHS, NIH, NCI, NIH Pub. No. 06-6067, 2006
• A. Bleyer et al., Eds., Cancer Epidemiology in Older Adolescents and Young Adults 15 to 29 Years of Age, Including SEER Incidence and Survival, NCI, NIH Pub. No. 06-5767, 2006

HRQoL measurement instruments and care guidelines

• S. Bhatia et al., "The Minneapolis-Manchester Quality of Life Instrument: reliability and validity of the Adolescent Form," Journal of Clinical Oncology, vol. 20, no. 24, pp. 4692–4698, 2002
• E. L. Goodman et al., "The Minneapolis-Manchester Quality of Life Instrument: reliability and validity of the Adult Form in cancer survivors," Quality of Life Research, vol. 27, no. 1, pp. 93–102, 2018
• E. J. Siembida et al., "Measuring health-related quality of life in adolescent and young adult cancer survivors with the NIH PROMIS," Psycho-Oncology, vol. 30, no. 3, pp. 303–311, 2021
• S. Bhatia, A. S. Pappo, et al., "Adolescent and young adult (AYA) oncology, NCCN clinical practice guidelines in oncology," Journal of the National Comprehensive Cancer Network, vol. 21, no. 8, pp. 851–880, 2023

Social, psychological, and survivorship mechanisms

• Zhao et al., "Loneliness and mortality risk among cancer survivors," Journal of the National Comprehensive Cancer Network, 2024

System-dynamics and hybrid simulation of cancer and health

• J. B. Homer and G. B. Hirsch, "System dynamics modeling for public health: background and opportunities," American Journal of Public Health, vol. 96, no. 3, pp. 452–458, 2006
• M. Hosking et al., "Investigating interventions for increasing colorectal cancer screening: insights from a simulation model," Socio-Economic Planning Sciences, vol. 47, no. 2, pp. 142–155, 2013
• L. W. Kivuti-Bitok et al., "System dynamics model of cervical cancer vaccination and screening interventions in Kenya," Cost Effectiveness and Resource Allocation, vol. 12, art. 26, 2014
• O. İrsoy et al., "Dynamic trade-offs in granulocyte colony-stimulating factor administration during chemotherapy," System Dynamics Review, vol. 36, pp. 397–446, 2020
• M. Heshmat and A. Eltawil, "A system dynamics-based decision support model for chemotherapy planning," Journal of Simulation, 2018
• F. Williams et al., "Combining community-engaged research with group model building to address racial disparities in breast cancer mortality and treatment," Journal of Health Disparities Research and Practice, vol. 11, pp. 160–178, 2018
• Ö. Karanfil and J. Sterman, "Saving lives or harming the healthy? Overuse and fluctuations in routine medical screening," System Dynamics Review, vol. 36, pp. 294–329, 2020
• C. Swinerd and K. R. McNaught, "Design classes for hybrid simulations involving agent-based and system dynamics models," Simulation Modelling Practice and Theory, vol. 25, pp. 118–133, 2012
• D. A. Marshall et al., "Selecting a dynamic simulation modeling method for health care delivery research, Part 2," Value in Health, vol. 18, no. 2, pp. 147–160, 2015
• "Applying dynamic simulation modeling methods in health care delivery research: the SIMULATE checklist," Value in Health, vol. 18, no. 1, pp. 5–16, 2015
• A. Jalali et al., "Strengthening a weak link: transparency of causal loop diagrams," System Dynamics Review, 2024

Agent-based and computational oncology

• N. Cogno et al., "Agent-based modeling in cancer biomedicine: applications and tools for calibration and validation," Cancer Biology & Therapy, vol. 25, no. 1, 2024
• J. Metzcar et al., "A review of cell-based computational modeling in cancer biology," JCO Clinical Cancer Informatics, vol. 3, pp. 1–13, 2019
• J. West et al., "Agent-based methods facilitate integrative science in cancer," Trends in Cell Biology, 2022
• L. K. Leykum et al., "Use of an agent-based model to understand clinical systems," Journal of Artificial Societies and Social Simulation, vol. 15, no. 3, art. 2, 2012
• A. Moskalewicz et al., "Development of a microsimulation model to project the future prevalence of childhood cancer in Ontario, Canada," Medical Decision Making, 2025
• S. F. Railsback and V. Grimm, Agent-Based and Individual-Based Modeling, 2nd ed. Princeton University Press, 2019

Simulation verification and validation methodology

• J. W. Forrester and P. M. Senge, "Tests for building confidence in system dynamics models," TIMS Studies in the Management Sciences, vol. 14, pp. 209–228, 1980
• E. J. Rykiel, "Testing ecological models: the meaning of validation," Ecological Modelling, vol. 90, pp. 229–244, 1996
• J. S. Hodges and J. A. Dewar, Is It You or Your Model Talking? A Framework for Model Validation, RAND, R-4114, 1992
• J. P. C. Kleijnen, "Verification and validation of simulation models," European Journal of Operational Research, vol. 82, pp. 145–162, 1995
• V. Grimm et al., "Towards better modelling and decision support: documenting model development, testing, and analysis using TRACE," Ecological Modelling, vol. 280, pp. 129–139, 2014

Network science and social contagion

• D. J. Watts and S. H. Strogatz, "Collective dynamics of 'small-world' networks," Nature, vol. 393, no. 6684, pp. 440–442, 1998
• A.-L. Barabási and R. Albert, "Emergence of scaling in random networks," Science, vol. 286, no. 5439, pp. 509–512, 1999
• M. E. J. Newman, "The structure and function of complex networks," SIAM Review, vol. 45, no. 2, pp. 167–256, 2003
• J. H. Fowler and N. A. Christakis, "Dynamic spread of happiness in a large social network: longitudinal analysis over 20 years in the Framingham Heart Study," BMJ, vol. 337, art. a2338, 2008
• J. Meng, "Linking network structure to support messages: effects of brokerage and closure on received social support," Journal of Communication, vol. 66, no. 6, 2016
• R. S. Burt, Structural Holes: The Social Structure of Competition. Cambridge, MA: Harvard University Press, 1992
• J. S. Coleman, "Social capital in the creation of human capital," American Journal of Sociology, vol. 94, pp. S95–S120, 1988
• B. Wellman and S. Wortley, "Different strokes from different folks: community ties and social support," American Journal of Sociology, vol. 96, no. 3, pp. 558–588, 1990
• D. Centola and M. Macy, "Complex contagions and the weakness of long ties," American Journal of Sociology, vol. 113, no. 3, pp. 702–734, 2007

Parameter-provenance sources (verified)

• B. Thom and C. Benedict, "The impact of financial toxicity on psychological well-being, coping self-efficacy, and cost-coping behaviors in young adults with cancer," Journal of Adolescent and Young Adult Oncology, vol. 8, no. 3, pp. 236–242, 2019
• L. V. Ghazal, R. Terrell, D. Rodriguez, et al., "Adolescent and young adult cancer survivor experiences of financial toxicity: a scoping review," Journal of Cancer Survivorship, 2026
• D. Kirk, I. Kabdebo, and L. Whitehead, "Prevalence of distress, its associated factors and referral to support services in people with cancer," Journal of Clinical Nursing, vol. 30, no. 19–20, pp. 2873–2885, 2021
• L. Melton, D. Krause, and J. Sugalski, "Psychology staffing at cancer centers: data from National Comprehensive Cancer Network member institutions," JCO Oncology Practice, 2020
• U.-S. Lehto, M. Ojanen, S. Lääperi, et al., "Cancer patients' social relationships during 3 years after diagnosis: generic and cancer-specific social networks," International Journal of Behavioral Medicine, vol. 32, no. 3, pp. 325–340, 2024
• S. Ameringer, R. K. Elswick Jr., D. P. Shockey, and R. Dillon, "A pilot exploration of symptom trajectories in adolescents with cancer during chemotherapy," Cancer Nursing, vol. 36, no. 1, pp. 60–71, 2013
• K. G. Cloyes, J.-W. Guo, K. E. Tennant, et al., "Exploring social support networks and interactions of young adult and LGBTQIA+ cancer survivors and care partners," Frontiers in Oncology, vol. 12, art. 1058411, 2022. (AYA support-network size: mean 6, SD 2.22, range 2–10.)

Public datasets

• National Center for Health Statistics, "National Health and Nutrition Examination Survey (NHANES)," Centers for Disease Control and Prevention, https://www.cdc.gov/nchs/nhanes/