Letters to the Editor

Dear Sir,

Ecker and Skelly advise us how to conduct a winning literature search (Evidence-Based Spine-Care Journal 2010; vol 1 (issue 1): 9–14). The search example they provide, however, contains several flaws, major and minor ones. These flaws are:

incorrect use of Medical Subject Headings in PubMed (MeSH); incorrect use of Boolean logic; incorrect use of truncation; exclusion of relevant freetext subject terms; no use of term-weighing; no warning for the adverse effects of PubMed-limits.

 

I would like to demonstrate these flaws by providing more details. First of all let me show the search chart which the authors provide:

The major flaw in all four search variants is the MeSH-term „Spinal Fracture”. This MeSH-term does not exist at all. The plural version, „Spinal Fractures”, however does exist. If one types the singular version „Spinal fracture”[MeSH] in the PubMed search bar, zero references will be retrieved. So if we adjust the first query, an improved version goes like this:

„Spinal fractures”[MeSH] OR vertebral compression fracture

If we look at the second query, we see another flaw: the illogic use of boolean operators. In this query, no brackets were used in conjunction with the OR- and AND-operators. The correct use of such logic operators is imperative for a winning search. Brackets improve the semantic relationship between the keywords:

(„Spinal fractures”[mesh] OR vertebral compression fracture) AND „osteoporosis”[MeSH]

In the third query, a new, third concept is introduced: surgery. However, the authors only use surg*. As such, this query results on the 12th of July 2010 in 1.972.843 references. However, due to the truncation of this term, the PubMed-function of automatic translation of a word in relevant MeSH-headings or subheadings will be lost – the third flaw in this query. My version, surg* OR surgery, retrieves on the 12th of july 2010 2.894.391 references. The word surgery will be automatically translated to „surgery”[Subheading] and „surgical procedures, operative”[MeSH], and will better the results significantly (an increase of 30 %). Also, and here is the fourth flaw, only the MeSH „Osteoporosis” is used: the combined use of both MeSH and free-text words is essential, otherwise, among others, the most recent references, which are not fully indexed yet, will be missed. My suggested version looks like this:

(„Spinal fractures”[MeSH] OR „vertebral compression fracture” OR „vertebral compression fractures” OR „spine fracture” OR „spinal fracture” OR „spine fractures” OR „spinal fractures” OR „vertebral fracture” OR „vertebral fractures”) AND („osteoporosis”[MeSH] OR osteoporosis[tw] OR osteoporotic[tw]) AND (surg* OR surgery)

Furthermore, the search could be improved by the inclusion of freetext versions of the phrase osteoporotic spine fracture:

(„osteoporotic spine fracture” OR „osteoporotic spine fractures” OR „osteoporotic spinal fracture” OR „osteoporotic spinal fractures” OR „osteoporotic vertebral fracture” OR „osteoporotic vertebral fractures” OR ((„Spinal fractures”[mesh] OR „vertebral compression fracture” OR „vertebral compression fractures” OR „spine fracture” OR „spinal fracture” OR „spine fractures” OR „spinal fractures” OR „vertebral fracture” OR „vertebral fractures”) AND („osteoporosis”[mesh] OR osteoporosis[tw] OR osteoporotic[tw]))) AND (surg* OR surgery)

Subsequently, these results could be slimmed down by weighing the core component of the question: osteoporotic spine fracture as the major topic. This can be accomplished by the use of Major Subheadings and the use of title words:

(„osteoporotic spine fracture”[ti] OR „osteoporotic spine fractures”[ti] OR „osteoporotic spinal fracture”[ti] OR „osteoporotic spinal fractures”[ti] OR „osteoporotic vertebral fracture”[ti] OR „osteoporotic vertebral fractures”[ti] OR ((„Spinal fractures”[majr] OR „vertebral compression fracture”[ti] OR „vertebral compression fractures”[ti] OR „spine fracture”[ti] OR „spinal fracture”[ti] OR „spine fractures”[ti] OR „spinal fractures”[ti] OR „vertebral fracture”[ti] OR „vertebral fractures”[ti]) AND („osteoporosis”[majr] OR osteoporosis[ti] OR osteoporotic[ti]))) AND (surg* OR surgery)

Finally, the authors limit results to several more or less formal PubMed-limitations, eg, humans. These limits have to be discouraged, because by using such limits, the most recent references, which have not been enriched by check tags such as human or clinical trial, will be lost, although these items are or could be in actual fact human studies or clinical trials.

So, the authors final statement, I feel, is rather poignant: „Use of personnel with specialized expertise in conducting such searches may provide the best results and be the most resource effective.”

Jan W Schoones, MA

Walaeus Library

Leiden University Medical Center

C1-Q

PO Box 9600

2300 RC Leiden

The Netherlands

31–71–5 262 182

j. [e-mail]