The filter was developed by researchers with ample experience in the retrieval of literature in biomedical databases and the development of search strategies.
Our goal was to develop a search filter that would facilitate the retrieval of high quality research describing sex-specific data on clinical questions related to conditions that can occur in both sexes. We excluded conditions that occur only in one sex, such as heavy menstrual bleeding, whereas literature on those conditions can be located simply by using disease-specific search terms (MeSH).
Jenkins has described a number of possible methodologies for filter development [
27]. Given the absence of previous work on the selection of sex-specific search terms and the composition of a gold standard of studies for the evaluation and validation of those terms, we have followed Jenkins' recommendation to develop a so called "second generation filter". A characteristic of second generation filters is that search terms are subjectively derived and tested against a gold standard. A novelty of our approach was that we wanted to ensure that the papers that were selected for developing and validating the filter would also be considered as relevant by clinicians. To achieve this goal we composed two sets of published papers that had been critically reviewed for methodological soundness and clinical relevance by experts, prior to the study. The first set, that was used to develop the search filter, was selected from clinical guidelines that were developed by organizations that belong to the Appraisal of Guidelines for Research and Evaluation (AGREE) collaboration. These guideline organizations use the internationally developed AGREE instrument for testing the quality of the guidelines they produce [
28]. Some of the AGREE quality criteria pertain to the selected evidence, which should be clearly evaluated for methodological soundness and clinical relevance. The second set of papers, that was used to validate the filter, was selected from a core set of leading clinical journals. The publication policy of these journals requires that published papers are carefully reviewed for level of clinical interest and methodological quality.
Development of the filter
Our aim was to identify 80 papers. Firstly, we selected 56 guidelines from a larger set of clinical guidelines published by four guideline organizations: the Dutch College of General Practitioners (NHG), the Dutch Institute for Healthcare Improvement (CBO), the Scottish Intercollegiate Guidelines Network (SIGN) and the National Institute for Health and Clinical Excellence (NICE, United Kingdom). Each of these organizations use a similar evidence-based methodology for the critical appraisal of the quality and the clinical relevance of research reports, as well as the aforementioned AGREE instrument for evaluating the quality of the final guideline documents [
29‐
31]. To be selected, a guideline should have been published in 2006 or 2007, provide an answer to clinical questions and deal with a condition that could occur in both men and women.
Secondly, these guidelines were screened for statements about men (boys), women (girls) or differences between them. Statements were included if they referred to any of the following topics: risk factors, the natural course of the disease, diagnostics (including disease manifestation and test performance), treatment or prognosis. Twenty-two of the 56 selected guidelines included such statements. Subsequently, for each of the included statements, we selected one reference to the underlying literature, based on the following criteria: the article must be written in English, published between 1996 and 2007, indexed in MEDLINE and contain an abstract. We excluded references to consensus papers or systematic reviews, as our aim was to identify papers on original studies.
Using this process we identified 75 research papers. Topics of these papers included: cancer (colon cancer), heart disease (chronic heart failure, familial hypercholesterolemia, secondary prevention after myocardial infarction, stable angina, stroke), other chronic disease (asthma in children and adults, Type 2 diabetes, osteoarthritis of the hip and knee, rheumatoid arthritis, thyroid disorders), infectious disease (hepatitis C, tuberculosis), mental health conditions (alcohol dependence, bipolar disease, dementia, attention-deficit hyperactivity disorders, eating disorders), neurologic disease (Parkinson) and others (enuresis nocturna). To complete our target of 80 papers we added five other papers to this set: two (on heart failure and Type 2 diabetes) were identified through the website of the US Agency for Healthcare Research and Quality
http://www.ahrq.gov/research/womenh1.htm[
32,
33], and three through a recent report on sex differences in rheumatoid arthritis [
34]. We checked if these papers were indexed with one of the sex-specific MeSH as listed in Table
1. This was the case for 41 of the 80 papers (51%).
To identify potentially relevant search terms by which the papers in this set could be located in MEDLINE, the citations were downloaded and the OvidSP™ interface was used to screen the title, abstract and the MeSH of the individual articles. This interface was selected because it is commonly available in medical institutions and guideline organizations [
35]. Moreover, we assumed that the more often a word is used in the abstract, the more important the topic will be (frequency). Likewise, we also assumed that the more closely two words are put together in the abstract, the more likely it is that their meaning is connected (adjacency). In contrast to other interfaces, such as PubMed, OvidSP™ offers operators for searching for the adjacency and frequency of words. This was another reason for selecting this interface.
For each of the 80 papers we registered all the words referring to male or female (either children or adults), sex and gender. Secondly, we registered the various combinations and frequencies by which these words appeared in title, abstract and MeSH as well as how closely they were located together. RD chose four as the minimum criterion for frequency and eight as the maximum criterion for adjacency. This choice was based on her prior experience with filter development and some tests. Finally, these data were examined to identify common patterns of terms by which a substantial number of the articles could be located. This led to the formulation of the sex-specific search filter (SSS filter). (Table
3)
Table 3
Sex-specific search filter (SSS filter) for MEDLINE for use with the OvidSP™ interface
#1 | (gender$ or sex$).af. |
#2 | (boys or girls).tw. |
#3 | (women or men).ti. |
#4 | (male$1 or female$1).ti. |
#5 | (women or men).ab./freq=4 |
#6 | (male$1 or female$1).ab./freq=4 |
#7 | (women adj8 men).ab. |
#8 | (female$1 adj8 male$1).ab. |
#9 | or/1–8 |
| The filter can be combined with a disease or other topic by adding search commands for the disease or topic and combining these commands with the SSS filter by using the Boolean operator 'AND'. |
| We recommend to do a search in the leading clinical journals on women's health in addition to a search with the SSS filter as described above. The following Ovid search terms for journals can be used: gender medicine.jn., journal of womens health.jn., journal of womens health & gender based medicine.jn. in combination with the disease or topic in question. |
The SSS filter consists of nine command lines to search for free text words in the fields containing information about the title, abstract and MeSH of individual articles. The first eight lines include one or more text words, followed by an affix, indicating the fields in which the words in question should be located (e.g., all fields (including title, abstract and MeSH) or a selection of those fields). In addition, lines #5 and #6 include a frequency operator. This operator is used to indicate the minimum frequency with which the terms male or female, or men or women, should appear in the abstract. Lines #7 and #8 include an adjacency operator. This operator is used to be able to identify statements in the abstract in which the terms women and men or male and female are used in combination with each other. Line #9 combines lines #1 to #8 by using the Boolean operator 'OR'.
The SSS filter was able to identify 74 of the 80 articles (92.5%) from the set it was derived from. Four of the six papers that could not be identified mentioned sex-specific information in the body of the text, but not in the fields that were searched by the filter (title, abstract and MeSH). The two other papers mentioned sex-specific information in the abstract, but the way in which this was phrased could not be recognized by the filter.
Validation of the filter
To validate the SSS filter we composed a reference set of papers through a search in MEDLINE [
36]. To be included in the reference set, a paper must report recent primary research on Alzheimer's disease or on asthma in humans, be published in core clinical journals and contain sex-specific evidence relevant to answer clinical questions. We chose asthma as a topic because the disease occurs in all age groups, including children. Alzheimer's disease was added as a random choice. We limited our search to core clinical journals because those journals are selected by the NLM as being of immediate interest to the practicing physician.
As a first step we searched all articles on Alzheimer's disease and on asthma that were published in core clinical journals in 2007 and 2008 and included in the MEDLINE database as of 13-06-2008. To this end we used the MeSH for the two diseases (exp Alzheimer Disease/or exp asthma/) and corresponding free text words in titles and abstracts (Alzheimer?.mp or asthma.mp). Only articles in the English language that contain an abstract were included. Studies involving animals were excluded. In order to obtain reports of original studies, papers of clinical conferences or consensus development conferences, congresses, (practice) guidelines, meta-analyses, reviews, and technical reports were also excluded, using MEDLINE's categorization by publication type.
As a second step we made a selection within this set by singling out the papers containing potentially relevant sex-specific information. To this end we screened the titles and the abstracts of the identified papers for the words (wo)man, (wo)men, (fe)male, widow(er), boy(s), girl(s), mother, father, sex or gender and MeSH including the words sex or gender. The papers that met these criteria were downloaded and their content was critically reviewed using the following criteria: a paper obtained a positive score if it reported data on men or women (or boys or girls) or the differences between them; if it evaluated the role of sex/gender as an independent variable or predictor for the outcome of the study or if it evaluated the role of sex/gender as an effect modifier for the relationship under study (see Table
4 for the criteria). We first evaluated information in the title and the abstract sections of the paper. If titles and abstracts did not provide sufficient information to decide whether relevant sex-specific evidence was present or not, we also evaluated information in the methods and results sections of the paper. The initial assessment of the papers was performed by the first author (CM). In case of doubt the papers were discussed with a second author (JH) until agreement was reached. The papers with a positive score formed the reference set.
Table 4
Criteria for the presence of sex-specific evidence relevant to clinical questions in primary research papers
Statements in title or abstract concerning: |
Men (boys) |
Women (girls) |
Differences between the sexes |
(Evaluation of) Independent effect of sex/gender on the relationship under study |
(Evaluation of) Sex/gender as predictor for the outcome of the study |
(Evaluation of) Sex/gender as an effect modifier of the relationship under study |
When the above-mentioned statements in title or abstract are absent:
|
Screening full text for statements regarding methods of analysis:
|
Comparison between the sexes |
Comparison between groups of a single sex |
Stratification by sex or subgroup analysis by sex |
Evaluation of sex/gender as effect modifier |
Evaluation of the independent effect of sex/gender on the relationship under study |
Evaluation of sex/gender as an independent predictor for the study outcome |
Statements regarding methods must be followed by a report about the outcome of this particular analysis. |
Screening full text for statements regarding reporting of results:
|
Presentation of separate risk estimates for men (boys) and women (girls) |
Presentation estimating the differences between (groups within) the sexes |
Outcome of a subgroup analysis by sex |
We evaluated the performance of the SSS filter on the set of original papers on Alzheimer's disease and on asthma. We used recall (the number of papers containing relevant sex-specific evidence on clinical questions retrieved by the SSS filter as a proportion of the total number of papers in the reference set) and precision (the number of papers containing relevant sex-specific evidence on clinical questions retrieved as a proportion of the total number of papers retrieved) as measures of performance [
27].
As a second test, we compared the performance of our filter to that of two other filters: a combination of previously selected sex-specific MeSH (Table
1) (SS MeSH filter) and the previously published filter for retrieving information on women's health (M&S filter) (see Table
2[
26]). To this end we applied the other two filters to the set of original papers on Alzheimer's disease and on asthma and compared the yields with those of the SSS filter. For this comparison we did not search in women's health journals, as is recommended by the M&S filter, because those journals are not represented in the set of core clinical journals, which was a selection criterion for the reference set.