Dear Adrian, When I think about database searches I see them as relatively
simple filters trying to capture, in a pragmatic and sometimes unnaturally
structured way, concepts expressed in records of research that may not be
well expressed nor used consistently. Search strategies are relatively
simple devices designed to catch something complex. Being relatively simple,
database search strategies may fail to retrieve all relevant records in a
database for a variety of reasons:
1. The key words used in the strategy may not capture all the concepts
required (as expressed by authors in titles and abstracts)
2. the limited words available to and used by the authors for a database
record (title and abstract terms) may not adequately describe the research
they conducted, or all aspects of the research, even though the full
research paper itself may contain the concepts in which the searcher is
interested.
3. the quality of the indexing applied by the database of interest may not
capture all the aspects of the research we want to find, or may not be
applied consistently.
So even very thorough strategies with lots of synonyms, using both textwords
and indexing, may miss highly relevant studies because of one or more of
these factors. There are many research papers recorded on the Cochrane
Methodology Register which describe the mismatch between searching and
available studies retrieved.
With handsearching you have so much more to work with - you have a whole
article to scan to look for relevance and you are not approaching it with an
inevitably limited set of keywords - you are using all your knowledge of the
topic to assess the relevance of much detailed material, rather than hoping
a (possibly large) set of carefully chosen keywords will coincide with the
terms that are available to be searched in database records.
Having said that, database searches are essential because project resources
do not usually allow the luxury of lots of handsearching vast numbers of
journals. The massive handsearching exercise that the Cochrane Collaboration
undertakes to find RCTs in journals to feed into the CENTRAL database is an
indication of how much work is involved to identify records that meet just
one concept.
Databases allow us to do reasonably thorough searches of millions of records
while being aware of the potential limitations of what we can claim.
Handsearching has to be carefully targeted - perhaps at high yielding
journals (identified from database searches) in the hope of finding more
records than the database search revealed or in highly relevant journals
unindexed on major databases to pick up studies that will not otherwise be
found by database searches.
The challenge and the enjoyment for searchers can be trying to find a
pragmatic mix of approaches to identify as much relevant research evidence
to answer a specific question, within the resources available. An integral
part of this approach to me seems to be an element of unavoidable redundancy
- searching a range of databases because of the added retrieval value
different indexing and search interfaces may offer, and handsearching as
well as database searching because both approaches can offer results.
Regards
Julie
Julie Glanville
Associate Director
Centre for Reviews and Dissemination
University of York
York
-----Original Message-----
From: Evidence based health (EBH)
[mailto:[log in to unmask]] On Behalf Of Adrian Sayers
Sent: 19 April 2006 10:19
To: [log in to unmask]
Subject: Re: Searching the literature - Electronic vs Hand
Hi interesting point on hand searching, I would love to hear more points of
view on this.
I would have to say in a lot of cases hand searching is a logistical
nightmare, and actually where do you hand search?
I think you only know where to hand search after finishing a good electronic
search, which then highlights the hotspots.
In addition a hand search of a large journal (BMJ, NEJM, LANCET etc etc.)
would take an extraordinary length of time, and do you only hand search
journals which you know aren't indexed. If a journal is indexed, what's the
point in hand searching articles that your keywords don't appear in?
On reviews similar to the one I am currently working on I have seen
individuals state that hand searching was an effective way to find articles
that would otherwise have been missed by electronic searching. Does this not
actually mean my electronic search strategy was not effective?
Surely the days of hand searching are numbered. The more research is
published the more we will need to devise better ways of searching the
literature electronically. By hand searching are we not just covering up our
limitations in electronic searching?
I pose some of these questions to entice debate.
Adrian
____________________________________
Adrian Sayers
Systematic Reviewer
Room 306H
Department of General Practice
School of Medicine
Cardiff University
3rd Floor, Neuadd Meirionnydd
University Hospital Wales
Heath Park
Cardiff
CF14 4XN
Direct Tel: 029 2068 7168
Dept. Tel: 029 2068 7159
Fax. 029 2068 7219
|