Dear all,
Based on my experience with retrieval and selection of records in embase of pubmed for systematic reviews and practice guidelines, after an initial learning curve, up to 300 records per hour can been screened for inclusion. That is when in advance words and phrases in title and abstract that lead to selection are clear. The quality of the search query determines the selection yield and the number needed to read. But information retrieval based on title and abstract of records in bibliographic databases, is only possible when their text includes useful phrases and words in natural language that accurately reflect the content of the article.
In most electronic databases only titles and abstracts of published articles are indexed. Unfortunately, the information contained in many abstracts differs markedly from that in the text (BMC Bioinformatics 2010;11:492). Therefore, authors need to be careful that titles and abstracts are both rich of information while accurately reflecting the article.
Huth (Ann Intern Med 1987;106:626) adopted the use of structured abstracts to improve informativeness of published articles. Mulrow and colleagues (Ann Intern Med 1988;108:613) proposed guidelines for more informative abstracts to promote accurate and efficient of literature searches. The EQUATOR Network (www.equator-network.org) promotes transparent and accurate reporting of research studies. Their reporting guidelines concern specific study designs and these also concern the abstracts. In 2003 the editors of Journal of clinical Epidemiology decided to implement “informative titles” to help readers distil the content of articles more quickly.
Informative titles (see http://pubs.chla-absc.ca/doi/pdfplus/10.5596/c05-029)
• include the answer to the research question (the results or main conclusion as message)
• are in the past tense for individual studies (whose results might be overruled by later studies or meta-analyses) and in the present tense for systematic reviews (whose results are unlikely to be overruled by later studies)
• include the design or architecture of the study
• are no longer than 15 words
The use of natural language in more informative titles and abstracts will reduce confusion and misclassification during indexing and retrieval of published articles. At the same time informative titles and abstracts will further improve the accuracy and efficiency of literature searches.
Geert JMG van der Heijden, PhD
Associate Professor of Clinical Epidemiology
Department of Clinical Epidemiology - Divison Julius Center for Health Sciences and Primary Care
Department of Otorhinolaryngology - Division Surgical Specialisms
University Medical Center Utrecht
The Netherlands
|