Modelling health, environment and transport: tools for a world of shocks

The methods we use share one thing: they represent individual people, not population averages. That changes what we can ask.

We can move from a momentary assessment — what someone is doing, where, with whom, breathing what, at every minute of the day — through cumulative dose across a day, to chronic exposure across a year, and finally to disease incidence, mortality and healthy life expectancy across a life course. We can model trade-offs: 24 hours are fixed, so more sleep means less work, a longer commute means less time to cook a healthy meal. And because every agent has a postcode and a household, we can stratify every result by age, sex, area, ethnicity and deprivation — making inequality a first-class output, not an afterthought.

That capability is what brings us into questions that pure epidemiology cannot answer on its own: who absorbs the cost of a clean-air zone, who benefits from a retrofit programme, who is left exposed when a heatwave or a supply shock hits.

This suite, originated by Prof Rolf Moeckel at TU Munich, was the first tightly integrated land use–travel behaviour–network assignment model that follows the same synthetic agents throughout. The whole stack is in Java and open source.

Working with him on the JIBE and UBD Policy projects, we incorporated health into the framework, and then the built environment alongside it ^[1].

The latest implementation, for Greater Manchester, runs on a highly detailed network — including for walking and cycling — that we built specifically to support active-travel research ^[2]. Using this network and travel survey data from Transport for Greater Manchester, we have built a behavioural model in which mode choice and route choice are sensitive to traffic stress, elevation and greenness — the things people actually weigh up when deciding whether to walk or cycle.

Integrated microsimulation behavioural framework for Greater Manchester

We then use MATSim to simulate travel along that network across a full week, and use the resulting trajectories to estimate person-centric exposure to PM2.5, NO₂, noise, green space, traffic injury risk and physical activity. A health impact model takes those exposures, behaviours and demographics and simulates the risk of developing disease and of dying. Code is in the and the .

In the DARe Hub, we are extending the suite to represent how behaviour adapts to extremes of heat and rainfall — modelling both the disruption that flooding causes to the system and the protective effects of nature-based solutions.

The Integrated Transport and Health Impact Model uses quasi-microsimulation comparative risk assessment to evaluate the health impacts of changes in travel behaviour, through air pollution, traffic injuries and physical activity ^[3]. It includes Monte Carlo simulation, with code available for value-of-information analysis ^[4].

The current implementation is open source in R. ITHIM was developed in Cambridge, with further versions built in California; a 2023 review found it to be the most widely used model worldwide for quantitative health impact assessment of transport ^[5].

We are now developing a streamlined version that requires far less input data, generating week-long synthetic travel survey data from a mix of sources and sub-models (travelBehaviour repo). It currently covers air pollution and physical activity, with traffic injuries to follow. The aim is to make rigorous transport health impact assessment feasible for any city, anywhere in the world — not only those with rich travel survey data.

ITHIM Global 2 - proposed workflow

A streamlined transport health impact assessment that needs far less input date

Open source in R - Monte Carlo simulation - stratifiable by age, sex, area, deprivation

Where this is heading

Three directions matter most to us right now.

Climate-cost trade-offs. Health, household budgets and carbon are rarely modelled together, but households experience them together. A shock in the energy system propagates through transport, food, housing and employment all at once — and the cost is rarely shared equally. We are building modelling frameworks that treat this as a single problem.

Beyond transport. Activity-based modelling is fundamentally a model of how people spend a day. That makes it a natural tool for questions about housing tenure, gig-economy working patterns, time poverty and diet — questions where the absence of a person-day-level representation is a serious limitation of current evidence.

Scaling up. Detailed agent-based models are powerful but expensive to build. Streamlined ITHIM and synthetic-population pipelines aim to make the same kind of analysis routine — for many cities, many policies, and at speed. We are particularly keen to expand work in the UK Midlands and to build deeper case studies in cities such as Singapore through the HD4 Programme and Bogotá.