Biomedical Engineering Reference
In-Depth Information
became aware of a preventive behavior they are already performing, they are likely
to underestimate their risk still further. Similarly, a consumer might overestimate
their risk when they identify a risk behavior (e.g., visiting wooded areas for Lyme
disease), if they fail to consider a preventive behavior that they already perform
(e.g., using tick repellent).
To summarize, the routes to de-bias bottom-up estimates of risk center on
increasing the accessibility and calibrating the diagnosticity of the individual behav-
iors and symptoms that relate to it. An integrative framework for behaviors and
symptoms that need to be identified and aggregated as part of a bottom-up process
is discussed next.
10.6
Sources of Bias in Identification and Integration
of Risk Factors
Consumers use the presence or absence of various symptoms and behaviors to
assess whether or not they are at risk. In fact, a large percentage of DTC advertising
highlights symptoms hoping that the mere awareness that a state is a symptom of a
disease will persuade consumers to assess their risk. However, symptoms vary along
many dimensions. Some symptoms are common across many conditions (e.g.,
fatigue), whereas others are more specific to a certain malady (e.g., joint pain).
Some appear extreme (e.g., thoughts of suicide or death in the context of depres-
sion), whereas others appear more “normal” (e.g., feeling unusually sad or irritable
over a 2-week period). Some occur with a high frequency for one individual (e.g.,
daily drinking in the context of alcoholism), and others with a lower frequency for
another individual who may be at equal risk (e.g., binge drinking).
Signal detection theory is a useful theoretical framework to think of how people
use behaviors or symptoms to make risk judgments. It postulates that there are six
characteristics of signals (in this case, symptoms or behaviors) that affect their use
(Sperling and Dosher 1986 ): their degree of existence, their actual expected consis-
tency, the frequency of occurrence of the signal, their causal clarity, measurement
error associated with them, and the expected gain from the signal detection task.
Signals that have a higher threshold of detectability (i.e., are less ambiguous), have
low measurement error (i.e., are difficult to reinterpret due to their ambiguity), have
lower actual expected consistency (i.e., are expected to occur only some of the
time), are infrequent , have higher causal clarity (i.e., if every time a symptom
occurs, the disease also occurs), and provide a high expected gain from signal detec-
tion task (i.e., have extreme consequences) are likely to be weighted to a greater
extent in the information integration task (Raghubir and Menon 2005 ).
In many situations (such as depression, thyroid imbalance, lupus, diabetes, and
heart disease), the symptoms vary in terms of many of the above characteristics. For
example, in the case of depression the symptom of “thoughts of suicide/death” is
different from the remaining eight symptoms (loss of interest or pleasure in activi-
ties, being sad or irritable, sleep disturbances, decreased ability to concentrate,
Search WWH ::




Custom Search