The STAPM Ecosystem

A living infrastructure for public health economic modelling.

The “Sociable Weaver” Philosophy

STAPM is more than a collection of 40+ repositories; it is a Modelling Platform. Inspired by the concept of “Platform Trials,” we have built a shared environment of data, code, and expertise that makes our research more efficient, flexible, and robust.

We take our inspiration from the Sociable Weaver, a bird that builds massive communal nests.

  • The Communal Roof: Our core platform—the standardised data pipelines, version-controlled engines, and governance procedures that everyone shares.
  • Individual Nest Chambers: These are specific projects (e.g., a report for Public Health Wales or a PhD study). Each project builds its own specific logic but relies on the strength and protection of the shared roof.
Illustration of a Sociable Weaver bird and its large communal nest
Figure 1: The Sociable Weaver: A metaphor for communal platform modelling. Individual efforts (projects) contribute to and benefit from a shared infrastructure (the roof).

🔄 The Organic Growth Cycle

The platform is an evolving entity. Every new project, no matter how specific, plays a role in reinforcing the communal roof.

graph LR
    A[Project Goals] --> B[Data & Governance]
    B --> C[Modular Code Tools]
    C --> D[People & Capacity]
    D --> E[Stakeholder Engagement]
    E --> F[Platform Evolution]
    F --> A

  1. Goals over Mega-Models: We avoid “one model to rule them all.” Instead, we use specific project goals to refine small, reusable parts of the ecosystem.
  2. Adaptability: If we build a tool for England, we design it to be extensible to Scotland, Wales, or international jurisdictions.
  3. People-First (Capacity Building): A platform is nothing without the people. Our “Watch one, Do one, Teach one” approach ensures technical resilience and team autonomy.
  4. Efficiency: By leveraging the existing “roof,” we deliver complex projects that would be infeasible to build from scratch.

🏗️ Governance & Version Control

To manage innovation while maintaining a stable “Gold Standard,” we follow a Fork-and-Reintegrate lifecycle. This allows us to innovate on the edges without destabilising the core.

graph TD
    CORE[CORE MODEL v1.0] --> FORK[Project Fork / Innovation]
    FORK --> DEV[New Policy Logic / Data Linkage]
    DEV --> DELIVER[Project Delivery]
    DELIVER --> REVIEW{Review & QA}
    REVIEW -->|Selected Features| REINTEGRATE[CORE MODEL v2.0]
    REVIEW -->|Bespoke Logic| ARCHIVE[Project Archive]

  • The Core Model: Our validated, reproducible “Gold Standard.”
  • Project Forks: When a project requires a new mechanism (e.g., a novel tobacco tax), researchers fork the main repository for rapid, isolated development.
  • Reintegration: Following project delivery and peer review, generalisable improvements are merged back into the Core Model.

🧩 The Modular Toolbox (Systems View)

Our 50+ repositories are “gathered up” as needed to build bespoke simulation pipelines tailored to specific policy questions.

Standardises raw national surveys and mortality microdata into a “common language”. * Tools: hseclean, mort.tools. * Consistency: Ensures definitions of health behaviours remain stable across decades of data.

Links multiple datasets to create the initial state for our simulations. * Tools: SynthTobAlc, SynthSmoke, SynthNoLo. * Logic: For example, linking health surveys with detailed behavioural studies to model individual motivations and consumption behaviours.

The mathematical relationships between consumption, health, and wellbeing. * Tools: smktrans, tobalcepi, alc.tools, qalyr. * Pedigree: Provides the transition probabilities and relative risk estimates used across the platform.

Where we launch simulations to appraise specific interventions. * Tools: TAX-sim, SAPM-R, STPM, stapmr. * Innovation: Regularly updated with new fiscal mechanisms and economic outcome metrics.

Our Modelling Paradigms

We utilise a multi-paradigm approach, allowing us to select the methodology that best fits the policy problem at hand.

Approach Lead Expertise Best For…
Microsimulation Gillespie / Angus / Morris Long-term health outcomes, fiscal impacts, and population epidemiology.
Agent-Based Modelling Purshouse Social networks, “tipping points”, and individual-level interactions.
Behavioural Complexity Squires Designing models that reflect real-world psychology and habit formation.

The Alcohol Frontier: While our alcohol work is currently driven by microsimulation, we are actively developing new ABM methods to simulate the social determinants of drinking. Our goal is to achieve the same “front-end/back-end” integration currently being deployed in our tobacco research.

🛡️ Access & Transparency

We are committed to Open Science while maintaining the high standards required for government-level evidence.

  • Internal Development: Active iteration happens in private GitHub environments to allow for rigorous Quality Assurance.
  • Public Release: Validated tools and R packages are released publicly via GitHub, with DOIs provided for formal academic citation.
  • Collaborative Culture: We encourage external analysts to “dip into” our methodology, fostering a more coherent and transparent public health modelling landscape.
NoteJoining the Nest

The platform grows through your contributions. As you develop a new project, look for the “win-win”—the small fix or feature that strengthens the “communal roof” for everyone.