Sidelines 16 (1)

Centre for Reviews and Dissemination

University of York

Web: http://www.york.ac.uk/inst/crd

The Sidelines abstracts in this issue focus on effective searching via search filters, the continuing discussion around clinical trial registration, assessments of health management information sources, how to present information to health care professionals and guidance on preparing grant proposals.

Jenkins M, Johnson F. (2004) ‘Awareness, use and opinions of methodological search filters used for the retrieval of evidence-based medical literature – a questionnaire survey’, Health Information and Libraries Journal, vol. 21, pp. 33-43.

The development of methodological search filters, or search strategies combining search terms relating to research methodology, has been seen as one way that health librarians can contribute to evidence-based practice. Filters designed to retrieve specific types of study such as systematic reviews, randomised controlled trials or diagnostic studies, are now available. This survey found that there was generally a high level of awareness of filters among health librarians, but this did not necessarily correspond with a high level of use. Confusion had been created by the variety of terms used to describe methodological search filters, such as clinical queries, hedges, quality filters, and optimal search strategies. In addition, there was no one web site where all the available filters were listed. Reported difficulties in the use of filters revealed some lack of understanding. For example, are filters an “inclusion” or an “exclusion” tool? should they be amended? is there a need to quality assess the papers retrieved when a filter has been used? The findings of the survey lead the authors to conclude that there needs to be greater promotion of the concept of filters and more guidance on how they can be used effectively.

Jenuwine ES, Floyd JA. (2004) ‘Comparison of Medical Subject Headings and text-word searches in MEDLINE to retrieve studies on sleep in healthy individuals’, Journal of the Medical Library Association, vol. 92, no. 3, pp. 349-53.

This article compares the performance of two search strategies in the retrieval of primary research papers on the sleep of healthy people, from MEDLINE. The first strategy used Medical Subject Headings (MeSH) only, and the second used text-words only.

A 'reference standard' set of relevant articles was obtained by hand searching the journal Sleep from 1996 - 2001. This provided a set of 137 papers to be retrieved.

The study found that the subject heading search provided higher specificity than the text-word search (i.e. fewer irrelevant records were retrieved by the subject heading search), but resulted in lower sensitivity (i.e. fewer of the relevant articles in the reference standard were retrieved than by the text-word search). Each strategy retrieved some records that the other did not.

The authors concluded that for optimum sensitivity and specificity, a combination of MeSH subject headings and text-words produces the best results. Subject heading searching provides a more precise search, but text-word searching allows a large number of terms to be used, to account for variations in the terminology used by authors.

The study authors suggest that the information professional can play an important role in informing authors of scientific publications of the value of consistent use of terminology in the titles and abstracts of their publications, thus maximising retrieval of research from online databases.

Montori VM, Wilczynski NL, Morgan D, Haynes RB for the Hedges Team. (2005) ‘Optimal search strategies for retrieving systematic reviews from Medline: analytical survey’, BMJ, vol. 330, pp. 68.

This is one of a series of papers produced by the Hedges Team at the McMaster University on developing optimal search strategies. This paper focuses specifically on retrieving systematic reviews from MEDLINE.

In addition to new strategies the authors test three previously published search strategies, one previously proposed by the Hedges Team, one from the Centre for Reviews of Dissemination (CRD) developed to populate the Database of Abstracts of Reviews of Effects (DARE) and one from Shojania and Bero who developed a search for PubMed, as a clinical query, and for OVID as a limit.

The methods used in this paper are similar to those used in the previous research by the Hedges Team on developing optimal search strategies. 161 journal titles indexed in MEDLINE and published in 2000 were hand searched for systematic reviews. These journals were for the fields of general internal medicine, family practice, nursing, and mental health, and indexed on MEDLINE.

4862 unique terms compiled from clinicians, librarians, published strategies and others were tested in MEDLINE (Ovid interface). Those individual search terms with a sensitivity of more than 50% or a specificity of more than 75% were selected. These terms were then combined, with the Boolean OR operator, up to a maximum of five term strategies and tested for sensitivity, specificity and precision. This procedure yielded an impressive 782,485 unique strings of one to five terms for analysis.

The results indicate the best single performing search terms, the best multiple term strategies maximising sensitivity and the best multiple term strategies maximising precision.

The authors, however, do appreciate the limitations of their study and acknowledge that searching the Cochrane Database of Systematic Reviews (CDSR) and DARE may be more appropriate in many circumstances to identify systematic reviews quickly.

Kyffin RGE, Goldacre MJ, Gill M. (2004) ‘Mortality rates and self-reported health: database analysis by English local authority area’, BMJ, vol. 329, pp. 887-8.

This article compares the use of mortality rates and surveys of self-reported health in the assessment of the health of a population.

Mortality rates are commonly used as measures of population health, but are often criticised because although they are readily available and objective, they are an extreme measure of ill health. However, self-reported health, such as the data gathered from questions in the UK Census, tends to be regarded as flawed because of its subjectivity.

For each local authority in England, the authors gathered the age standardised mortality rates for the major causes of death for 1999 and 2001 from the Compendium of Clinical and Health Indicators. This was then compared with the age standardised rates of self reported health status for the same areas using data from the 2001 Census.

The study found that the mortality rates were highly correlated with the census measures of self-reported ill health, although this was not true for all local authorities. This suggests that despite the conceptual concerns about each method, the two different measures both result in a similar health profile for a given population.

Inouye, SK, Fiellin, DA. (2005) ‘An evidence-based guide to writing grant proposals for clinical research’, Annals of Internal Medicine, vol. 142, pp. 274-282.

This article gives advice to physicians applying for grants, particularly those intending to carry out patient-oriented research.  Rather than merely dispensing advice from on high, the authors have attempted a systematic approach, basing their recommendations and suggestions on a critique of previous grant applications.  The evidence for the advice is culled from an examination of 66 grant applications made to the Clinical Research Study Section of the National Institutes of Health.  Some of the comments and criticisms made by the grant reviewers on each key section of the 66 proposals are given.  These are followed by the authors’ suggestions, which respond to these criticisms, about how each section might be addressed in a more focused, clear and methodical way.  Criticisms range from proposals not clearly explaining the importance of the proposed research to poor presentation and spelling.  The authors stress the importance of writing a proposal that engages the reviewer from the start, making it stand alone without the need for appendices and other supporting material, and ensuring the writing is grammatically correct and not sloppy or careless.  The article contains graphic timeline examples as well as a useful ‘Checklist for the grant-writing process’.  Although the paper is written primarily as a guide for US grant applicants, the advice given would be applicable to any grant proposal.

Vans T, Gữlmezoglu M, Pang T. (2004) ‘Registering clinical trials: an essential role for WHO’, Lancet, vol. 363, pp. 1413-1414.

On behalf of WHO, the authors of this commentary highlight the need for a comprehensive and international system of trial registration that aims to either build one single register or to link all the existing registers. The ISRCTN (International Standard Randomised Controlled Trial Number) scheme is viewed as an encouraging start to achieving this goal. It is argued that such an initiative would increase international access to health-related knowledge, improve decision making by practitioners, researchers, policy makers and funders.

Haun MT. (2004) ‘Information to go: publishing organisation-specific information for personal digital assistants’, Joint Commission Journal on Quality and Safety, vol. 30, no. 5, pp. 286-289.

This article suggests how to package information and other resources for physicians to access using personal digital assistants (PDAs) . The advantages of PDAs over print resources include quick and easy accessibility to information, portability and the ability to provide files that will “self-destruct” when they become out of date.

Stevens AJ, Raftery J, Roderick, P. (2005) ‘Can heath technologies be assessed using routine data?’, International Journal of Technology Assessment in Health Care, vol. 21, no. 1, pp. 96-103.

This study explores the potential of routine data to provide an alternative, or complement, to the use of randomised controlled trials (RCTs) in health technology assessment. Routine data is defined as information that is required to be collected regularly at national or regional level and uses standard definitions.

The authors identified 270 databases that met these criteria but were disappointed by the small number of databases that were suitable for their needs. Problems included the type of information collected. For example, many datasets contain information about health technologies but not about the healthcare state for which they were used.  There are also difficulties with the way information is presented. For instance, the International Classification of Diseases (ICD) is widely used to classify information, but lacks the kind of detail required to inform health technology assessments. Issues around confidentiality also make it difficult to access information on a range of topics (for example, sexual health).

The article ends with examples of how routine data has been successfully used in other countries but warns that the issues outlined above must be urgently addressed if we are to make use of routine data.     

Back to Inform 16 (1)

This page was last updated on: 19 September 2005