|
Investigators
can take several steps to strengthen etiologic classification.
Explicit
criteria for assigning symptoms to specific categories can
be developed by review of the literature and by expert opinion.
Explicit criteria for broad categories (physical, psychological,
idiopathic) applicable to a variety of symptoms have been developed
(Khan
et al., 2000). There are also published examples of
more precisely classifying the cause of a specific symptom, such
as dizziness (Kroenke
et al., 1992).
Since such
criteria rely heavily on signs and symptoms obtained by patient
interview and physical examination, softer data should be solidified
by using:
- Structured
interviews;
- Validated
questionnaires; and
- Standardized
physical examinations.
Measurement
bias can be further mitigated by training, calibration, and reliability
checks of all study personnel responsible for obtaining this clinical
data. Even then, explicit criteria are sometimes at best a "bronze
standard", an operational or working definition to facilitate
more uniform classification and reproducibility in subsequent
studies (Drossman,
1999; Fukuda
et al., 1994).
A second way
to "harden" this relatively soft process of etiologic
classification is to have multiple raters review the data
and independently assign a cause, with any disagreements arbitrated
by consensus. Because even experts differ substantially in their
diagnostic preferences (Kroenke
et al., 1992), etiologic classification that depends on one
person reviewing the data may be unduly biased. Interobserver
agreement can be calculated to provide an estimate of diagnostic
reliability.
A third technique
for fortifying classification is to develop a uniform abstract
of all salient historical, physical, and laboratory data, so that
each rater reviews the identical information. In fact, sequential
review of these elements of the data with raters assigning a probable
cause after each step may be useful in determining the incremental
and independent contributions of historical, physical, and laboratory
data toward establishing a diagnosis (Kroenke
et al.,1992).
|