Pages within the "support" section:
17. Treatment of dependencies in Qalibra framework
The Qalibra framework allows the user to represent of some types of dependencies in the risk-benefit calculation:
· Dependencies between intakes of different foods, contaminants or nutrients (e.g. individuals who eat less meat are likely to eat more vegetables or dairy products).
· Dependency of effects on intakes, including the familiar dose-response relationships for the probability of quantal effects and the magnitude of continuous effects, and also any dose-dependency that may exist for other aspects of effects (e.g. age of onset, duration, severity, recovery rate, mortality rate).
· Dependencies between uncertainty distributions for the same parameter in different subpopulations (e.g. different age/gender groups, see section 15.3).
Users of Qalibra are responsible for correctly representing these dependencies within the input formats accepted by the software.
Some types of dependency that may be relevant when considering health risks and benefits cannot readily be accommodated within the directly attributable health impacts model adopted for Qalibra. These include:
· Dependencies or interactions between different health effects (e.g. the severity of one disease, or the probability of recovery or mortality from it, may depend on whether the individual suffers another disease concurrently).
· Dependencies in the between-individual variation of different parameters, except for intakes. This is because the directly attributable health impact calculations uses averages across individuals (of the same subpopulation) for all parameters except intakes, and because the result of a calculation based on averages may differ from the average result of the same calculation repeated for different individuals.
The potential impact of such dependencies is difficult to evaluate qualitatively, but should be considered as part of the overall evaluation of unquantified uncertainties affecting the assessment (see section 15.1). In cases where such dependencies might have a material effect, it may be possible to explore them quantitatively, either by using a different modelling approach or by representing them very approximately in Qalibra, e.g. by using the functionality for individual attributes to define separate population subgroups in different quantiles for parameters where individual variation may be important. Any assessment of this type would require very careful planning and interpretation.