Cochrane Project

Download Report

Transcript Cochrane Project

A ‘best match’ for common search
commands: a comparison of
searching (CDSR and CENTRAL) via
Wiley and Ovid
HLG Conference, July 2012
Jenny Craven, Information Specialist, NICE
Jenny Kendrick, Information Specialist, NICE
Advice on new
and existing
treatments
Clinical
guidelines,
outcome
frameworks and
quality
standards
NICE and NHS
Evidence
Evidence – guidance – shared learning
Health
promotion and
disease
prevention
Comprehensive
evidence
service
Associate
Director, 23
Information
Professionals,
and an
Administrator
Institute-wide
corporate
information
services
Guidance
Information Services
Tailored
information
support for the
different Centres
and Directorates
within NICE
Professional
expertise to
enable access to
high quality
information to
support the
development of
guidance
Purpose of the project
• As a result of experiencing some issues with the
Cochrane Library via Wiley in 2010/11 a project group
was set up to look into the differences of searching:
– Cochrane Database of Systematic Reviews (CDSR)
– Cochrane Central Register of Controlled Trials
(CENTRAL)
– between the Wiley and Ovid search interfaces
Aims of the project
• To compare search commands and functionality of
Wiley and Ovid, including identifying the ‘best match’ for
common search commands between the different
interfaces.
• To identify any discrepancies in the way comparative
search commands are executed across the two
interfaces.
• To investigate and explain any discrepancies identified.
• To compare the performance.
• To investigate any other issues identified during the
project.
Evidence in Literature
• Inconsistent search commands across various
interfaces...
– Differences in the way stop words are searched and the use of
adjacency (Feinglos, 1983)
– Limits and Field searches can differ between interfaces (De
Groote, 2000, Allison, 2006)
– Text word searches (e.g. ‘All text’ and TX) can produce
different numbers of results. (Younger & Boddy, 2008)
– Some interfaces assume truncation, some require a wildcard (*)
(Younger & Boddy, 2008)
• Learning the ‘intricacies of individual interfaces and
databases is time-consuming’ (Boddy & Younger, 2009)
Methods
• Detailed cross comparisons were undertaken on the
following search functions:
-
Truncation
Proximity Operators
MeSH terms including exploded MeSH terms;
Free text/All fields, including free text (title, abstract,
keyword search) and free text (title, abstract only)
Methods
• Appropriate terms were selected to explore each of the
search functions.
• Terms were selected from:
– existing strategies from Interventional Procedures
(IP) and Guideline Review (GR) searches
– and, for practical reasons, terms were selected that
did not retrieve a large number of results (ideally less
than 200 results, although this was not always
possible)
Methods
• Searches were recorded using a Word table, listing the
test terms, interface and database, syntax applied, and
the number of results.
• If a large number of results were retrieved, these were
imported into Reference Manager to identify unique
records for closer inspection in the appropriate
database.
• Where the numbers retrieved were low, records could
be inspected directly from within the database.
Search syntax and functionality
Database
Truncation
Proximity
operators
MeSH
Exp MeSH
Free text All fields
Free text ti, ab,
kw/hw
Free text –
ti,ab
OVID
CENTRAL
*
adj3
MeSH/
Exp MeSH/
-
ti, ab, hw
ti, ab or tw1
OVID
CDSR
*
adj3
kw
Include all
terms as kw
and Or/
together
tw2
ti, ab, kw
ti,ab2
Wiley Cochrane
Library
*
Near or
lower
proximity
e.g. Near/4
MeSH/4
MesH
explode all
trees
Search All3
text (i.e.
term with no
suffix)
ti, ab, kw –
includes
‘Plain
Language
Summary’
ti,ab
1 Preferred free text searching mode for Ovid CENTRAL
2 Preferred free text searching mode for Ovid CDSR (note that .tw is equivalent to full text searching in CDSR)
3 Preferred free text searching mode for Wiley
4 Note: in Wiley check tags (e.g. adult, child, male, female) need to be searched for using .kw
Truncation & Proximity Operators
• Truncation - Use * in both Ovid and Wiley
• Proximity Operators
– The proximity operators perform differently in Ovid
and Wiley.
– In Ovid Adj3 refers to three words between search
terms and stop words are not included in the count.
– In Wiley Near/3 refers to the first search term plus
the next two words, and all words are counted
(including stop words).
– Either use Near in Wiley, or if too many records are
retrieved, use a lower proximity e.g. Near/4.
MeSH terms
• MeSH terms - CDSR
– In Ovid use .kw
– in Wiley use MeSH
• MeSH Terms - CENTRAL
– Use MeSH/ in Ovid and Wiley
Exploded MeSH
• Exploded MeSH – CDSR
- In Wiley use ‘MeSH explode all trees
- In Ovid use .kw and include the main MeSH heading plus
the narrower terms Or’d together, for example:
MeSH descriptor Labor Onset
explode all trees
1.
2.
3.
4.
5.
Labor Onset.kw
Labor Stage, First.kw
Labor Stage, Second.kw
Labor Stage, Third.kw
OR/1-4
• Exploded MeSH – CENTRAL
- Use exp MeSH/ in Ovid and Wiley.
Free text/All fields
• Free text/ All fields - CDSR
– Use .tw in OVID
– no suffix (all text) in Wiley.
– Small differences in retrieval as ‘evidence
tables’ and ‘reference list’ text included in a
Wiley ‘all text’ search but not included in an
Ovid .tw search.
Free text/All fields
• Free text/ All fields - CENTRAL
– In Wiley use no suffix (all text)
– There is no option for performing an ‘all text’
search in Ovid CENTRAL.
Free text (ti,ab,kw)
• Free text (title, abstract, keyword search) CDSR
– Use .ti, ab, kw in both OVID and Wiley.
– Small differences in retrieval as ‘plain
language summary’ field included in a Wiley
.ti,ab,kw search, but not in Ovid.
Free text (ti,ab,kw)
• Free text (title, abstract, keyword search) CENTRAL
– Use .ti, ab, kw in Wiley.
– Use .ti, ab, hw in OVID
– The qualifier .hw is used in OVID to ensure retrieval
of CENTRAL records indexed with MeSH and nonMESH subject headings (i.e. EMTREE terms). A .kw
search in Ovid will only retrieve records indexed with
non-MeSH headings.
Free text (ti,ab)
• Free text (title, abstract) – CDSR
– In Ovid and Wiley use .ti,ab.
• Free text (title, abstract) – CENTRAL
– in Wiley use .ti,ab.
– In Ovid use .ti,ab or .tw
Other Issues
•
•
•
•
Limits (date and language)
Exporting records
Multifile searching
Currency
Summary and conclusions
• In summary, it is possible to successfully search CDSR
and CENTRAL via Ovid and Wiley.
• Advantages of Ovid:
– Cross-search the CDSR and CENTRAL databases with
MEDLINE and EMBASE (multifile searching), and being able to
de-duplicate within Ovid.
• Disadvantages of Ovid:
– The study experienced a time delay between the updating of
CDSR and CENTRAL in Ovid.
Summary and conclusions
• Advantages of Wiley:
– Cross-search the CDSR and CENTRAL databases with the
CRD databases
– Only having to create a single strategy to search CDSR and
CENTRAL instead of multiple strategies.
• Disadvantage of Wiley:
– Some issues are still being experienced – such as timing out
during searches, loss of searches, and displaying incorrect
search results
Final comment
• It should also be noted that there will always be
differences in the number of records retrieved by the
two interfaces
– execution of MeSH headings (CDSR)
– fields included in an ‘all text ‘search (CDSR)
– proximity operators (CDSR and CENTRAL)
• The search syntax presented here and in the report
simply represent the ‘best match’ between the Ovid and
the Wiley interfaces.
Free/Quick plug!
• Final Report - Cochrane project: a comparison of
searching CDSR and CENTRAL via Wiley and Ovid,
copies available on request
• Craven, J., Kendrick, J., and Boynton, J. (2011)
Cochrane project: a comparison of searching CDSR
and CENTRAL via Wiley and Ovid. Health libraries
Group Newsletter, December
http://www.cilip.org.uk/get-involved/special-interestgroups/health/pages/newsletter.aspx
• Detailed article is planned for the near future
References
• Allison, M. (2006) Comparison of CINAHL via EBSCOhost,
Ovid, and Pro Quest. Journal of Electronic Resources in
Medial Libraries, vol 3(1) pp31-50.
• Craven, J., Kendrick, J., and Boynton, J. (2011) Cochrane
project: a comparison of searching CDSR and CENTRAL via
Wiley and Ovid. Health libraries Group Newsletter,
December http://www.cilip.org.uk/get-involved/specialinterest-groups/health/pages/newsletter.aspx
• De Groote, S. L (2000) PubMed, Internet Grateful Med, and
Ovid: a comparison of Three MEDLINE Internet Interfaces.
Medical Reference Services Quarterly, Vol 19(4) pp1-13.
References
• Feinglos, S.J. (1983) MEDLINE at BRS, DIALOG, and NLM:
Is There a Choice? Bulletin of the Medical Library
Association. 71(1) January.
• Younger, P. and Boddy, K. (2008) When is a search not a
search? A comparison of searching the AMED
complementary health database via EBSCOhost, OVID and
DIALOG. Health Information and Libraries Journal, 26,
pp.126–135
• Younger, P. and Boddy, K. (2009) What a difference an
interface makes: just how reliable are your search results.
Focus on alternative and complimentary therapies Vol 14(1)
March, pp5-7.
Thank You
Any Questions?
[email protected]
[email protected],uk