Case-Based Reasoning in the Health Sciences: Foundations and Research Directions - Computational Intelligence in Healthcare: Advanced Methodologies - page 131

Fig. 4. The knowledge spectrum in biomedical informatics [63]
these domains [62]. Recent studies have worked at better formalizing this specific
role. These studies explain that the gold standard for evaluating the quality of
biomedical knowledge relies on the concept of evidence. Pantazi et al. propose
an extension of the definition of biomedical evidence to include knowledge in in-
dividual cases, suggesting that the mere collection of individual case facts should
be regarded as evidence gathering [62]. To support their proposal, they argue
that the traditional, highly abstracted, hypothesis centric type of evidence that
removes factual evidence present in individual cases, implies a strong ontological
commitment to methodological and theoretical approaches, which is the source
of the never-ending need for current and best evidence, while, at the same time,
offering little provision for the reuse of knowledge disposed of as obsolete [62].
By contrast, the incremental factual evidence about individuals creates, once
appropriately collected, a growing body of context-dependent evidence that can
be reinterpreted and reused as many times as possible. See Figure 4, from [63]
for a graphical representation of this knowledge spectrum.
Currently, the concept of evidence most often refers to an abstract proposition
derived from multiple cases obtained in the context of a randomized controlled
trial [62]. Hypothesis forming is the cornerstone of this kind of biomedical re-
search. Hypotheses that pass an appropriately selected statistical test become
evidence. However, the process of hypothesis forming also implies a commitment
to certain purposes (e.g., research, teaching, etc.), and inherently postulates on-
tological and conceptual reductions, orderings and relationships. These result
from the particular conceptualizations of a researcher who is influenced by expe-
rience, native language, background, etc. [62] This reduction process will always
be prone to errors as long as uncertainties are present in our reality. In addition,