Glossary - Evaluation Methods in Biomedical Informatics - page 362

Glossary

The glossary in the first edition of this topic was adapted from an earlier
version produced by the authors for a conference on evaluation of
knowledge-based information resources, sponsored by the National Library
of Medicine and held in 1995. Some terms defined here are not explicitly
used in this topic but may be encountered elsewhere by readers as they
read evaluation reports or the methodological literature. We thank Bruce
Buchanan, Gregory Cooper, Brian Haynes, Harold Schoolman, Edward
Shortliffe, Mark Roberts and Bonnie Webber for their suggestions for terms
and definitions. This second edition glossary has been revised and expanded
to reflect revisions in the text itself.

Accuracy: (1) Extent to which the measured value of some attribute of an
information resource, or other object, agrees with the accepted value
for that attribute or “gold standard” (qv. † ); (2) extent to which a measure-
ment in fact assesses what it is designed to measure (roughly equivalent to
“validity”).

Action research: A disciplined method for intentional learning from expe-
rience characterised by intervention in real world systems followed by close
scrutiny of the effects. The aim of Action Research is to improve practice
and it is typically conducted by a combined team of practitioners and
researchers. Originally formulated by social psychologist Kurt Lewin.
[Adapted from Wikipedia definition, www.wikipedia.org].

Alerting resource: Resource that monitors a continuous signal or stream
of data and generates a message (an alert) in response to patterns or items
that may require action on the part of the care provider.

Allocation concealment: Ensuring that those recruiting participants to a
randomized trial have no knowledge of the group to which each participant
will be allocated. Failure to conceal allocation has been shown (using meta
regression techniques [qv]) to be a major cause of bias in such studies, and

†

qv. = see also

362