| |
A useful model has
to approximate the relevant features characteristic of the clinical
phenomenon under study. The validity of the model must be demonstrated
to convince others that it represents a reasonable research tool.
Model validation is straightforward in situations in which the
clinical case definition is established by validated laboratory
measures. However, in the absence of a definitive standard, such
as in the case of "complex" or "multi-factorial"
diseases, the evaluation of model validity requires multi-level
investigations. These include:
| Table
12.1: Levels of Model Validity Evaluation
|
|---|
| Level
1:
| Description
of clinical phenomenon, including internal validation
of observations |
|---|
| Level
2:
| Study
of model attributes, including internal validation of
observations |
|---|
| Level
3:
| Cross-validation
of experimental model |
|---|
|
This multi-level validation
approach is necessary to arrive at valid clinical descriptions
that will then be used as contrasts to model observations.
Level
1: Characterization of the Clinical Phenomenon
Using research designs and measurements principles customary in
sound epidemiological work, the clinical phenomenon in question
is characterized. The resulting data serve as a reference standard
needed for the comparison of the clinical phenomenon and the model
system. If universal agreement does not exist on the definition
of the disease in question, reasonable taxonomic constructs (A,
B, C refer to different taxonomic constructs) reported in the
literature should be explored. Taxonomic constructs used to define
a case may differ with respect to what components are mandatory
and/or the level of sensitivity by which these mandatory features
are assessed. If the measurement cutoff point for a mandatory
feature is different among taxonomies, differences result with
respect to what constitutes a clinical case. Differences can also
be due to the fact that a particular taxonomy includes a certain
component as a mandatory feature while another treats the disease
attribute in question as a co-morbid disease expression. In sum,
if the case definition is uncertain (which happens to be an inherent
problem for complex diseases) it may be prudent to include all
necessary measurements in support of the diagnostic assignments
using the prevailing taxonomic systems.
|