Integrated assessments are complex things.  They typically relate to issues that have many (and often remote) causes, which operate via different pathways and processes, and lead to a wide range of health and associated impacts.  If the assessments are to deal effectively with these complexities, they often need to bring together different forms of  information from a variety of different sources and to analyse and supplement these by different methods.  Great care is therefore needed to ensure that the assessment remains:

• Relevant - i.e. is faithful to, and targetted at, the issue that needs to be assessed;
• Scientifically credible - i.e. provides results that are accurate and reliable;
• Transparent - i.e. can be evaluated and understood by those not directly involved (and if necessary can be replicated or validated).

None of this is likely to occur unless the assessment is carefully designed.  And the focus of design must be on identifying and eliminating (or at least minimising) the uncertainties that might arise.  In this context, the most important rules for assessment design can be summarised as follows:

1. All the most important hazards and benefits of relevance to the issue (and their interactions) must be included in the analysis, and must be properly represented by the data and models being used.
2. Variations in the study population (especially in their differing susceptibility to potential health effects) must be allowed for, so that results of the assessment are not biased towards particular, non-representative groups.
3. The value systems applied in converting the health outcomes to overall measures of impact (and in selecting the indicators to reflect these) must be explicit and appropiate, and should not unacceptably bias the results.
4. The temporal and spatial 'dimensions' of the assessment must be valid and appropriate - i.e. the assessment should cover an appropriate geographical area and time scale, and analysis must be done at  an appropriate resolution, so that key impacts are not omitted or blurred in the analysis;
5. The overall analysis must be coherent and consistent - i.e. there should be no illogicalities or gaps in the way the causal chain has been represented or analysed.

Information on how to achieve each of these is available through the links to the left.