Timeframe in assessment design
- The text on this page is taken from an equivalent page of the IEHIAS-project.
If its results are to be meaningful, any assessment must relate to a clearly specified timeframe. The way in which this is defined will vary depending on the type of assessment. Diagnostic assessments often relate to current conditions; summative assessments to the period from when the policy (or other development) of concern was first introduced up to the present time; prognostic assessments to future impacts.
In the case of prognostic (and to a somewhat lesser extent, summative) assessments, defining a relevant timeframe for the assessment can pose substantial challenges. Problems arise not only because the effects of policies and technologies vary over their lifetime (as they are developed, established, operated and finally withdrawn), but also because many leave a much longer lasting legacy. Chemical works and old military sites, for example, remain contaminated scores of years after they have been decommissioned; landfill sites may continue to release pollutants for hundreds of years after they have been closed and infilled; modern cities and transport networks (and their attendant health implications) have, to a large extent, been shaped by historic planning decisions. In addition, many health effects persist long after the first symptoms emerge, and some (e.g. reproductive effects) have inter-generational effects.
In all these cases, ignoring these longer term consequences by confining attention to a single snapshot in time (e.g. one year) or to immediate impacts can seriously bias the assessment. Instead, if we wish to consider the total impacts of a policy or technology, we should try to sum the impacts over the full timespan of its influence.
Doing so has numerous implications for the way the assessment is designed. It means, for example, that we need data on human populations, hazards and exposures that cover many years, and in some situations extend far into the future. As a consequence, we have to rely on models not only for the analysis itself, but also to help us work out what the real timeframe for analysis should be (i.e. how long impacts might persist). These models may have to be extended far beyond the period over which they can be fully validated, which inevitably increases the uncertainties involved - though because the total health burden also tends to increase, the relative error (as a proportion of the total impact) may not change greatly.
It also means that we have to be able to accumulate impacts over time. This in turn requires that we often need to consider (and combine) very different types of effect, including acute and chronic outcomes, and thus to make allowance for the duration, as well as severity, of effect. By the same token, we may have to make judgements about the relative importance of more immediate compared with more remote effects (including, in some instances, those on generations as yet unborn). Standard measures of health outcome (e.g. life expectancy or total mortality) give a very limited perspective in these situations, so usually we are driven to use more synthetic indicators, such as disease- or quality-adjusted life years (DALYs and QALYs) or monetary evaluation.