Infrastructure
We develop accessible modelling infrastructure to enable collaborators to employ advanced simulation methods within their own research
Modelling Infrastructure
SHINE modelling is guided by a number of epidemiological, economic, and data science principles. To learn more, see our research protocol. This protocol describes the data pipelines and standardised methodologies used by SHINE, in order to support clear and practical processes for current and future modelling.
Proportional Multistate Lifetable Modelling. At the heart of SHINE is a proportional multistate life table (PMSLT) model. This model has evolved from Excel-based models developed initially by ACE-Prevention and evolved in the BODE3 program. It includes the following features:
- Data pipelines from GBD data, to automatically derive input business-as-usual disease incidence, morbidity and case fatality rates (and future projections), and all-cause mortality and morbidity. For any country.
- Data processing tools to disaggregate any country’s data by heterogeneity (socioeconomic position, indigeneity, disease risk strata) given user inputs of relative risk differences in disease parameters by strata of heterogeneity.
- Disease-related health expenditure for Australia, disaggregated by disease phase (year of diagnosis, last year of life, prevalent) using NZ estimates. We plan to also include such disease-related health expenditure for many other countries, using algorithms under development. This data – when integrated with the PMSLT – automatically outputs the change in health expenditure by time into the future for any intervention.
- Disease-related income loss for NZ , rescaled to Australia. This data – when integrated with the PMSLT – automatically outputs the change in future income earnings due to preventing death and morbidity. That is, productivity impacts of health interventions, an output of interest to decision making considering the societal impacts of health interventions.
- A range of ‘interventions models’, for example our smoking-vaping life history, risk factor (e.g. BMI) intervention models linked to changing disease incidence, and our COVID-19 agent-based model. Over time, a growing suite of such intervention models will be built that can be re-used in future similar projects.