Potential problems
physiological/clinical
data collection
physiological variation
For example if your GP took your blood pressure
three times in a short period of time,
first reading is on average higher than the subsequent readings
Take this into account taking the
measurements more than once and at different
times so that an average reading is calculated
and recorded.
variation in the measuring instrument
e.g. the blood test procedure or taking blood pressure
BP variation =app of cuff before inflation and this can
change the final reading.
so must be calibrated before, during and after
each study and data may need to be adjusted on
the basis of subsequent calibration.
observer variation
People taking blood pressure measurement may vary
in their ability to hear the sounds (if using the manual
device),
might read the measurement incorrectly,
might write down the incorrect result onto the data collection form.
even experienced clinicians can vary in their ability to read an X-ray
counteract this
2 exp ppl
read X-rays
other test results
before study standardised criteria and
interpretation
should be developed and included in the study protocol
Experienced observers should practice with
more experienced observers
all observers should be tested and corrected until
reliability is obtained both between observers and
within all observations taken by a single observer
bias from interactions
involving the person being measured, the measuring
instrument, and the observer.
participant may learn over time how to use an instrument
better and so it might appear that the test results are
improving
can happen with respiratory testing for asthma
Interactions can occur between the
observer conducting the test and the
participant that introduces bias.
For example a person's pulse rate decreases as
they become less anxious, or blood pressure can
increase when the reading is taken by a person of
the opposite gender
even tho clinical tests used in health have
already been rigorously tested for reliability
and validity.
Variation still occurs
Records= measurement of interest
may not have been conducted or recorded
the information may have been
recorded by many different people in an
incomplete and non-standardised way
Limited value
establish if a participant has had the
disease of interest, particularly in
case-control and cohort studie
exposure and confounding variables are not usually obtained