Cardiovascular Epidemiology and Epidemiological Modelling DATA NEEDS FOR
- Slides: 9
Cardiovascular Epidemiology and Epidemiological Modelling DATA NEEDS FOR A MODEL Martin O’Flaherty Simon Capewell Division of Public Health University of Liverpool
Model data needs Depend on: Questions to be addressed Choice of model logic Desired outputs Purpose: Populate the model Validate the model Key issues Availability Format A modeller’s dilemma: Built your model around your data OR Build a model and then gather the data.
A generic model of a chronic disease Healthy Disease Death
A generic model addressing a generic public health question Healthy Primary Prevention Disease Death Secondary Prevention
A generic model of a chronic disease What determine s INCIDENC E Healthy Disease How large are the groups What Determines prognosis Time Death
A summary of IMPACT data needs Number of people in each group In each risk factor In each disease subgroup (eg: AMI, UA, CA, HF) Number of deaths What determines incidence Risk factors effect measures (RR or b) What determines prognosis: Case fatality rates for each disease group (rates) Interventions that reduce case fatality (RRR) Uptake of those interventions (%)
A summary of IMPACT data needs Take into account time Trends of: Number of people in each group Levels of risk factors Levels of uptake of treatments Data to validate the model Observed mortality
Some last (but not least) data need Data to support assumptions: Needed to fill gaps in knowledge They are guesses, but the better the data supporting the guess the better the guess. An example: “English fatality rates: Scottish SLIDE data adjusted using England/Scotland SMR (see http: //www. heartstats. org/temp/Tabsp 1. 6 spweb 06 spup. XLS)” Data will need: Critical appraisal Adaptation Documented (appendices)
The data gathering task This phase of a modelling project is critical: Defines feasibility Defines the quality of the final product (the principle of “garbage IN, garbage OUT” It is time consuming: it will take a large amount of the available resources. But as any activity in a modelling exercise, data gathering is an ITERATION: We start simple, and we add layers of greater complexity.
