Webcursus.ubvu.vu.nl

PITFALLS AND TRAPS
Things to watch out for when searching literature databases.
1. Multiple relationships between concepts:
Examples: Depression in patients as a result of cardiovascular complaints v. Depression as a risk factor for cardiovascular complaints Depression as a side-effect of medicaton v. Medication in the treatment of depression The Boolean ‘AND’ makes no distinction. Distinctions can be made with keyword combinations but that usually makes the search less sensitive. Take, for example, the MeSH-SH combination: Depression/aetiology v. Depression/drug therapy. 2. Different meanings for 1 word/acronym
Homonyms, words or acronyms with different meanings corrupt the results. Example: IC can stand for ‘intensive care’, ‘internal control’, ‘immune complex’, ‘interstitial cystitis’, etc. 3. Negative examples of term-mapping in PubMed / MeSH difficult to find
It isn’t too difficult to find controlled terms. There are some examples of good ‘mapping’ in PubMed both in the search (see ‘Details’) and the MeSH Database: Adalat, effexor, heart attack, nsaid, nsclc, tens, itp, alpha blocker. But that isn’t always the case. Here are some other examples from PubMed: No mapping, no (usable) reference in the MeSH Database: • ascal (cf. ‘adalat’ and ‘effexor’); • IBD (for ‘inflammatory bowel disease’; cf. ‘ITP’); • ic, icu (for ‘intensive care (unit)’); • ms (for ‘multiple sclerosis’), a low 74th place in MeSH; • alpha blocking agent (cf. ‘alpha blocker’). No mapping, but usable reference in the MeSH Database: • tocolytic drugs (cf. ‘tocolytic drug’ and ‘tocolytic agent’); • prophylactic (cf. ‘prophylaxis’); • heart failure (not expanded to ‘Heart failure, congestive’[mesh], so the • antiphospholipid antibodies syndrome (meanings APS, • ct (if ‘Computer Tomography’ is meant).
No searches in the intended fields
Search machines have standard settings for the search fields.
These can interfere with the aim of the search.
Example: A specific reference needs to be looked up in PubMed for this article:
Burn J, Dennis M, Bamford J, Sandercock P, Wade D, Warlow C. Long-term risk of
recurrent stroke after a first-ever stroke. The Oxford Community Stroke Project.
Stroke 1994 Feb:25(2):333-7.
The search query ‘burn dennis bamford 333’ yields no results. That is because
‘burn’ is interpreted as a subject term and no searches were performed in the
author field (see ‘Details’).
The query needs to be changed to: ‘burn[au] dennis[au] bamford[au] 333’
Example: You need to do a search with a non-English term such as ‘platzbauch’ or
‘wiegedood’.
The original titles of non-English articles are registered in PubMed, but the field –
the ‘translated title’ field – is not included in a standard search.
The field must be explicitly tagged in the search query: platzbauch [tt] wiegedood [tt] 4. Adjacency not carried out as expected
A space can be interpreted as a Boolean AND operator, while the searcher has merely intended it as a space. For example, in PubMed: • cingulate cortex (though ‘cingulate gyrus’ is recognised); • positional preference; • preferred position; • mechanical neck disorders. 5. Notations are often sensitive:
There are cases where a comma or quotation marks can make a big difference, for example: In PubMed: aneurysm, dissecting[mh] and aortic aneurysm, thoracic[mh] aneurysm, dissecting[mh] and "aortic aneurysm, thoracic"[mh] in PubMed: aneurysm, dissecting[mh] and aortic aneurysm, thoracic[mh] aneurysm, dissecting[mh] AND aortic aneurysm, thoracic[mh] 6. Case-insensitivity is a hindrance
You can lose a lot of specificity if the search engine doesn’t distinguish between lower and upper case letters. Here are two classic examples: who (WHO) aids (AIDS) 7. Stop words etc.
There are stop words for most databases. They are not searchable but they are in the records. For example, without, not in PubMed Search non-demented elderly 8. No controlled keyword
Don’t expect all the elements in a question to translate into controlled keywords. Examples: Delay (in diagnosis, for example) Ototoxicity, nephrotoxicity (not in MeSH, but in Emtree) Corresponding MeSH: Ear diseases/chemically induced. Duration, e.g., 24 hours (pH measurement), 3 days v. 10 days (course of antibiotics) Age, e.g., 75 years and older 9. Complexity of the keyword system; different controlled keywords are
feasible
The keyword systems of the large databases are so big that for a sensitive search you often need to select more than one keyword for each core element in the question: • Medication groups in MEDLINE: use both standard Mesh and • Closely connected terms are assigned different places in the MeSH system, e.g. ‘clinical trial’ is ‘Publication type’, ‘cohort study’ is a real MeSH. • MeSH terms can be applied in totally different ways from similar • Choose 1 combined MeSH (e.g. Cerebellar ataxia[mesh]) or 2 simple MeSH (e.g. Cerebellum[mesh] AND Ataxia[mesh]). Or don’t make a choice and incorporate both options in the search profile? o Another example: Back pain[mesh] v. Back[mesh] AND Pain[mesh] • Choose 1 combined MeSH (e.g. Appendectomy) or a MeSH/SH tag (e.g. • All keywords that represent closely connected concepts should be included o Fertility – Infertility o Cardiomegaly - Heart Failure. o Steroids - (‘corticosteroids’ >) ‘Adrenal cortex hormones’ o Intensive Care[mesh] – Intensive Care Units[mesh] – Critical 10. Misleading controlled keywords
Example: the MeSH Rcts is not used for descriptions of RCTs but for articles about (the execution of) RCTs, including meta-analyses of several RCTs. The index term for articles that report an RCT is ‘publication type’ rct Another example: included in the MeSH Animals is the MeSH Humans. In a standard setting for an ‘exploded’ search, the query NOT Animals[mesh] could deliver totally unintended results. 11. Rarely used keywords
In the past: ‘Mammary neoplasms’ fell under ‘Breast neoplasms’ but the indexers almost invariably used ‘Breast neoplasms’ for female patients. Now ‘Mammary neoplasms’ has been scrapped. Under ‘Breast neoplasms’ you will still find ‘Breast neoplasms, male’. ‘Femoral neck fractures’ is fine, but in practice the most commonly used term is still ‘Femoral fractures’ or even ‘Hip fractures’. Positron Emission Spectrometry > since 2005, but scarcely used since then; indexers prefer the BT (= too wide). 12. Author’s language (‘free terms’)
There are immense variations in the way authors use language. For example: • UK English v. US English (oesophagus - esophagus, anaemia - anemia); • Singular, plural, adjective, noun; • Synonyms, antonyms; • Abbreviations v. full terms. 13. Truncation too short or too long
If words are truncated too soon, they will generate a lot of unwanted terms. • Rat* will generate not only rat and rats, but also ratio, rational, rate. • Mole* will generate not only molecule, molecules etc. but also mole • Analysi* will not find ‘analyses’, ‘analysed’ etc. • Premature* will not find ‘prematurity’. 14. Effects of ‘Limits’
Some searchers do not realise the effect of ‘limit’ settings. For example: the ‘subsets’ cancer, nursing etc. are not limited with subject terms but with journal titles. 15. Uncertainty about the study format
The range of ‘filters’ and ‘limits’ for types of studies is unequally balanced. For example, in MEDLINE and EMBASE you can filter with ‘rct’, ‘ct’, ‘cct’, but not with ‘cohort’, ‘transectional study’, etc. Because the filter/limit function works so easily, people tend to use study formats that fall short of the requirements. Example: • ‘harm’ (side-effects): there is not enough information in RCTs. You need to add e.g., longer-running cohort studies and case reports for a full picture. • Research into the value of diagnostic procedures is never referred to as ‘ct’ (clinical trial) in MEDLINE, but it is in EMBASE. In MEDLINE a ‘limit’ on clinical trial does not work, whereas in EMBASE it works reasonably well 16. Missing subject elements
Elements of research questions that are underrepresented in titles, abstracts and keywords: population, ‘setting’, outcomes • Frontline care • General population • Healthy people • Employees

Source: http://webcursus.ubvu.vu.nl/docs/PITFALLS_AND_TRAPS.pdf