HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Appendix E1 Supplemental Tables All Versions of this Article: 2472070101v1 247/2/365    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Astin, M. P. Articles by Grimshaw, J. M. Search for Related Content PubMed PubMed Citation Articles by Astin, M. P. Articles by Grimshaw, J. M.

Developing a Sensitive Search Strategy in MEDLINE to Retrieve Studies on Assessment of the Diagnostic Performance of Imaging Techniques¹

¹ From the Health Services Research Unit, Health Sciences Building, School of Medicine, University of Aberdeen, Foresterhill, Aberdeen AB25 2ZD, Scotland. Received April 19, 2007; revision requested June 11; revision received August 23; accepted September 27; final version accepted October 18. Supported by a grant from the European Union Directorate of the Environment. Address correspondence to M.P.A. (e-mail: m.p.astin{at}abdn.ac.uk).

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCES IN KNOWLEDGE IMPLICATIONS FOR PATIENT CARE References

 
Purpose: To prospectively develop a search strategy in MEDLINE for identifying studies on the diagnostic performance of any imaging modality, with maximized and minimized retrieval of relevant and irrelevant studies, respectively.

Materials and Methods: Predefined inclusion criteria were used to conduct a hand search of two sets of radiologic journal articles for studies on assessment of the diagnostic performance of imaging techniques. These two sets of articles formed independent derivation and validation record sets for developing and evaluating the search strategy. The sensitivity and positive predictive values (PPVs) of search terms from the derivation reference-standard set of records were used to select terms and develop two components of the search strategy. The first component was used to identify any study (from the imaging literature) in which diagnostic test performance was assessed. The second component was used to identify studies of any imaging modality. The two components were combined in the final search strategy. The sensitivity, specificity, and PPV of the search strategy in the derivation and validation record sets were calculated.

Results: The final search strategy had a sensitivity of 92.8%, a specificity of 58.5%, and a PPV of 25.1% for retrieval of the derivation set of records. Validation with an independent set of records gave a sensitivity of 91.9% (95% confidence interval [CI]: 87.1%, 95.1%), a specificity of 52.2% (95% CI: 49.2%, 55.2%), and a PPV of 25.1% (95% CI: 22.0%, 28.5%). Removal of irrelevant publication types further improved specificity and PPV in the validation set: to 77.6% (95% CI: 75.0%, 80.0%) and 40.9% (95% CI: 36.2%, 45.7%), respectively. The volume of imaging literature retrieved from MEDLINE by using the described search strategy has tripled since 1975.

Conclusion: A sensitive search strategy to identify studies of the diagnostic performance of any imaging test was developed and validated. The retrieval estimates of this strategy in MEDLINE are adequate to develop a register of studies.

Supplemental material:
http://radiology.rsnajnls.org/cgi/content/full/2472070101/DC1
http://radiology.rsnajnls.org/cgi/content/full/2472070101/DC2

© RSNA, 2008

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCES IN KNOWLEDGE IMPLICATIONS FOR PATIENT CARE References

 
Radiology is a progressive clinical discipline that encompasses a wide and increasing range of imaging modalities in health care, including diagnostic, prognostic, and interventional procedures. Given the large volume of published radiology literature and the increasing role of systematic reviews in the assessment of diagnostic test performance, there is a particular need to identify diagnostic imaging studies accurately.

A number of search strategies, or search filters, for identifying studies of diagnostic tests (1–9) and of imaging modalities (10–14) have been published. Other search strategies can be found at various Internet sites (15–19). The majority of search strategies for identifying diagnostic studies do not discriminate between laboratory disciplines, imaging disciplines, and clinical studies on the differential diagnosis of specific conditions. Also, the sensitivities and specificities of search strategies for identifying diagnostic imaging studies are not always reported. Therefore, the purpose of our study was to prospectively develop a search strategy in MEDLINE to identify studies on the diagnostic performance of any imaging modality, with maximized retrieval of relevant studies and minimized retrieval of irrelevant studies.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCES IN KNOWLEDGE IMPLICATIONS FOR PATIENT CARE References

 
The search strategy was developed in two stages: In the first stage, a derivation set of journal articles (Appendix E1, http://radiology.rsnajnls.org/cgi/content/full/2472070101/DC1) was used to derive the diagnostic and imaging components of the search strategy. In the second stage, these two components were validated by using an independent set of journal articles (validation set) (Appendix E1, http://radiology.rsnajnls.org/cgi/content/full/2472070101/DC1) (Fig 1).

View larger version (16K): [in this window] [in a new window] [Download PPT slide]   Figure 1: Stages in development of search strategy. MeSH = medical subject headings, PPV = positive predictive value.

Inclusion Criteria for Diagnostic Imaging Studies
The following inclusion criteria for selecting studies were defined by using specific guidelines for evaluating the performance of diagnostic tests (5,22–27): (a) The investigation had to be a primary study, (b) one or more imaging tests had to be used for diagnosis or screening, (c) a reference standard (eg, histologic analysis, surgery, or follow-up) had to be used to confirm the presence or absence of the target condition, (d) both an index test and a reference-standard examination had to be performed in most of the participants, and (e) the sensitivity, specificity, and/or predictive values of the test were reported, or the raw data from which these values could be derived were reported.

The exclusion criteria for selecting studies were nonsystematic literature reviews, meta-analyses, editorials, commentaries, letters, pictorial essays, case reports of single cases, opinion articles, conference proceedings, and other types of abstracts. We also did not include therapeutic, prognostic, interventional, volunteer, or phantom and animal studies. Studies about technical developments of instrumentation were excluded as well. These hand searches were conducted prior to the publication of the STARD (Standards for Reporting of Diagnostic Accuracy) guidelines (28) for reporting studies of diagnostic test accuracy.

Hand Search of Derivation and Validation Article Sets
Derivation set.—In 1985 and 1995, each journal was hand searched for a 6-month period to identify the derivation reference-standard (DRS) articles. Exceptions were Clinical Radiology, which was hand searched for the equivalent of 1 year, and the American Journal of Neuroradiology, six issues of which were searched in 1988 (the first issue available locally). These years were chosen to ensure a time span that was adequate to accommodate changes in MEDLINE indexing, study designs, and imaging modalities. Articles were screened for possible inclusion, but they were not critically appraised. The title, abstract, and body of the articles in each issue were screened independently by two authors (M.P.A., M.G.B.) and one researcher from the Royal College of Radiologists; the articles were then double-checked for inclusion eligibility by one researcher (M.P.A.). Any uncertainties were resolved by means of discussion between two researchers (M.P.A., M.G.B.).

Validation set.—The same hand search methods were applied to the validation set of journals to identify the validation reference-standard (VRS) articles. A more recent year (2000), with complete MEDLINE indexing, was chosen to accommodate changes in article indexing and imaging technologies. In this set, fewer diagnostic studies were identified, so one researcher (M.P.A.) screened these journals for the entire year by using the inclusion criteria.

Types of included studies.—The studies in the DRS and VRS articles were subclassified as type 1 or type 2 studies on the basis of the diagnostic performance statistics that could be calculated (M.P.A., M.G.B.). Type 1 studies described a broad population of both subjects who were suspected and subjects who were not suspected of having a clinical syndrome, so sensitivity and specificity values could be calculated. Type 2 studies described a narrow study population that included only patients confirmed to have a given clinical condition, so only sensitivity could be calculated. The DRS and VRS article sets were divided into two subsets representing journals with type 1 and those with type 2 studies: DRS1 and DRS2 and VRS1 and VRS2, respectively. Type 1 studies were methodologically stronger than type 2 studies.

Definition of Terms
We used the same formulas that previous researchers used to calculate the retrieval statistics (2) (Fig 2, Appendix E1 [http://radiology.rsnajnls.org/cgi/content/full/2472070101/DC1]). In this context, the PPV served as an indication of the proportion of records that needed to be screened for the identification of relevant articles. For example, a search strategy with a PPV of 30% would imply that 100 records needed to be screened to identify 30 relevant articles.

View larger version (18K): [in this window] [in a new window] [Download PPT slide]   Figure 2: Contingency table for diagnostic test evaluation applied to information retrieval. The formulas used to calculate test performance are listed. DS = derivation article set, RS = reference standard, VS = validation article set.

Development of the Search Strategy
The search strategy was developed by entering search terms for two independent components in MEDLINE: terms to identify any article describing assessment of the diagnostic performance of a test and terms to identify any article on an imaging modality. The terms were added sequentially to the diagnostic or imaging components. Initially, the terms with the highest PPV were added, and at each subsequent step the next term that enabled retrieval of the largest proportion of additional DRS records was added. Some terms with a high PPV that were initially included were subsequently excluded as the strategy was developed because the addition of terms with a lower PPV made them redundant. This process was repeated until the sensitivity was maximized for each component by two researchers (M.P.A., C.M.F.). Finally, the diagnostic and imaging components were combined by using the Boolean operator "AND" (Appendix E1, http://radiology.rsnajnls.org/cgi/content/full/2472070101/DC1). Each line of the search strategy was numbered sequentially by using the OVID program (OVID Technologies, New York, NY).

Performance of Diagnostic and Imaging Components in Retrieving Derivation and Validation Sets
The performances of the individual diagnostic and imaging components of the search strategy, of the full search strategy (with combined diagnostic and imaging components), and of the final adjusted strategy (after removal of the animal studies and irrelevant publication types [ie, case reports, letters, editorials, pictorial essays, review articles, and historical articles]) were assessed by running the searches in MEDLINE and combining the individual terms being tested with the derivation, DRS, validation, and VRS article sets. One researcher (M.P.A.) used the records retrieved to calculate the sensitivity, specificity, and PPV.

The DRS and VRS records that were not retrieved by using the diagnostic and/or imaging components at each stage of the strategy were reviewed by comparing the titles, abstracts, and MeSH or other MEDLINE indexing terms with the search terms of the strategy (M.P.A.). The adjusted search strategy was also run in MEDLINE and limited to the period between 1975 and 2005 to assess the volume of imaging literature (M.P.A.).

Statistical Analyses
The sensitivity, specificity, and PPV of the search strategy components to retrieve the derivation and validation article sets, with corresponding 95% confidence intervals (CIs), for the identification of diagnostic studies were calculated by using the Confidence Interval Analysis program, 2000, version 2.1.1 (Trevor Bryant, University of Southampton, Southampton, England). The recommended Wilson method was applied to calculate 95% CIs. The differences in proportions between the search strategy components were calculated with the ² test to obtain P values by using the computer program Epitable for Proportions in Epi Info 2001, version 6.04d (World Health Organization, Geneva, Switzerland). One author (M.P.A.) performed all data analyses and generated all tables and figures, with final approval from all other authors.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCES IN KNOWLEDGE IMPLICATIONS FOR PATIENT CARE References

 
Hand Searches
More type 2 studies (involving patients with a confirmed clinical condition) than type 1 studies (involving subjects suspected and subjects not suspected of having a clinical syndrome) were identified among the DRS and VRS articles (Tables E1 and E2, http://radiology.rsnajnls.org/cgi/content/full/2472070101/DC2). The proportions of each study type did not change over this time span (1985–2000), with approximately 44% of the type 1 studies (148/333 for DRS, 82/186 for VRS) and 56% of the type 2 studies (185/333 for DRS, 104/186 for VRS) identified.

Performance of Search Strategy Components in Derivation Set
The sensitivities of individual search terms in the DRS article set ranged from 2.1% to 60.1%, and the PPVs ranged from 17.0% to 100% (Table E3, http://radiology.rsnajnls.org/cgi/content/full/2472070101/DC2). The sensitivities of the diagnostic and imaging search components decreased as the components were combined (by using the Boolean operator "AND") (Table 1; Fig 3, line 22). The decrease was not significant for the diagnostic component (P = .09), but it was for the imaging component (P = .01). After irrelevant publication types were removed (Fig 3, line 28), the sensitivities of the diagnostic (P = .002) and imaging (P < .001) components were significantly lower. The largest losses of articles not retrieved occurred among the type 2 studies. The increases in specificity were significant when the diagnostic (P = .002) and imaging (P < .001) components were combined (Table 1; Fig 3, line 22). The removal of irrelevant publication types in the adjusted search strategy also led to significantly improved PPVs (Fig 3, line 28; P < .001 for diagnostic and imaging components).

View larger version (14K): [in this window] [in a new window] [Download PPT slide]   Figure 3: Search terms used in diagnostic and imaging components of search strategy. To improve sensitivity in the validation set, the italicized search terms were later added to the imaging component by using the "OR" operator. Exp = explode, MR = magnetic resonance, MRS = magnetic resonance spectroscopy, PET = positron emission tomography, SPECT = single photon emission computed tomography, SPET = single photon emission tomography.

Four DRS1 studies were not retrieved by using the imaging component. Two of these studies involved assessment of MR imaging technologies, another involved assessment of both MR imaging and computed tomography (CT), and the fourth involved assessment of CT. Although CT may be indexed with the floating subheading "ra/fs" (with "ra" standing for radiography), this was not used for the article. MEDLINE lacks a floating subheading term for MR imaging. Three of the missed articles on the assessment of MR imaging were indexed correctly with the MeSH term magnetic resonance imaging/; similarly, the articles on the assessment of CT were indexed with the MeSH term tomography, x-ray computed/. All four articles could be retrieved by using the exploded MeSH term exp diagnostic imaging/. However, because none was indexed as a main focus of the article by using the asterisk symbol (*magnetic resonance imaging/, or *tomography, x-ray computed/), they were not retrieved by using our search strategy. After the search strategy was adjusted to exclude irrelevant publication types, an additional DRS1 study was missed. It was indexed incorrectly as a case report, although there were more than 400 participants in the study.

Larger numbers of DRS2 studies were missed by using the search strategies. Twelve studies were not retrieved by using the diagnostic search component because the relevant terms were not listed in the MeSH, article titles, or abstracts. Six DRS2 studies were not retrieved by using the imaging search component. All six of them were indexed as "magnetic resonance imaging/" but not as the main focus of the article. After the search strategy was adjusted to exclude irrelevant publication types, an additional 10 DRS2 studies were missed: Nine of these were indexed as case reports, and one was indexed as a review.

Validation of Diagnostic and Imaging Search Components
The sensitivity of the diagnostic search component in the validation article set (Table 2) was comparable to that in the derivation article set (Table 1), with a small increase in sensitivity (96.8% vs 95.8%). However, specificity was significantly lower in the validation set (43.9% vs 52.3%, P < .001). The sensitivity of the imaging component in the validation set was not significantly different from that in the derivation set (P = .29). However, the specificity of the imaging component was significantly higher (26.2% vs 22.1%, P < .01). After the diagnostic and imaging components were combined (Fig 3, line 22), there was a significant reduction in specificity in the validation set compared with that in the derivation set (52.2% vs 58.5%, P < .001). There were no significant differences in sensitivity between the derivation and validation article sets when the diagnostic and imaging components were combined (Fig 3, line 22; P = .7) or after the irrelevant publication types were removed (Fig 3, line 28; P = .9).

As in the cases of missed DRS records, larger numbers of VRS2 records than VRS1 records were missed by using the search strategies. Five VRS2 studies were not retrieved by using the diagnostic component: Two of these studies were indexed with the subheading "x/di" for diagnosis of a clinical condition, and all five of them lacked diagnostic component search terms in the MeSH, article titles, or abstracts. Eight VRS2 records were not retrieved by using the imaging component. Four of these records had no imaging MeSH terms; however, CT and MR imaging; and CT, MR imaging, and ultrasonography were cited in one abstract each, while PET or scintigraphy was listed in the remaining two abstracts. The other four records were indexed with imaging MeSH terms (magnetic resonance imaging or magnetic resonance spectroscopy) but not as the focus of the article, and, consequently, these records were not retrieved.

A larger proportion of type 2 records in the VRS set than in the DRS set were not retrieved by using the imaging component. After the search strategy was adjusted to remove irrelevant publication types, an additional three VRS2 studies were missed. All three of these studies were indexed as case reports. The frequency of hits retrieved with the adjusted search strategy increased from 2989 in 1975 to 8124 in 1995 and then to 11 128 in 2005, suggesting that the volume of imaging articles indexed in MEDLINE had more than tripled (Fig 4).

View larger version (8K): [in this window] [in a new window] [Download PPT slide]   Figure 4: Graph shows frequencies of MEDLINE hits according to year for the adjusted diagnostic imaging search strategy. Number of hits refers to total number of records retrieved by using the strategy, with the search limited to individual years.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCES IN KNOWLEDGE IMPLICATIONS FOR PATIENT CARE References

 
We identified two types of imaging studies: type 1 studies, in which both the sensitivity and the specificity of the given imaging test could be calculated, and type 2 studies, which were methodologically weaker and in which only sensitivity values could be calculated. Using our search strategy enabled us to identify both study design types. Retrieval of type 1 studies was greater than retrieval of type 2 studies, among both the derivation and the validation article sets.

There were several challenges in developing a sensitive and specific MEDLINE search strategy, most of which were generic to the development of any search strategy. There is always a trade-off between sensitivity (identifying as many relevant articles as possible) and specificity (not identifying as many irrelevant articles as possible). As sensitivity increases, specificity decreases. There are also issues regarding inconsistencies in indexing and unstructured abstracts of articles in electronic bibliographic databases. We noted inconsistencies in the assignment of case report as a publication type when we sought to exclude this type of article for improved specificity. One of the studies indexed as a case report included more than 400 patients. However, there did appear to be some improvement over time: A smaller proportion of records in the VRS set (2000) than in the DRS set (1985 or 1988, 1995) were inappropriately excluded as case reports.

One issue specific to the imaging studies was the possibility that the indexing terms may not have remained in pace with the development of new imaging technologies. This was demonstrated by the fact that the sensitivity of our validated imaging component of the search strategy could be further improved by adding free text terms to identify CT, MR imaging, and spectroscopy studies because there were no existing index subheading terms for these major imaging modalities. Some of the studies that were missed would have been retrieved by using the MEDLINE MeSH terms magnetic resonance imaging/ and computed tomography, x-ray computed/. However, if we had not restricted the use of these terms to the focus of the article, the specificity would have been considerably lower owing to increased numbers of irrelevant hits and increased numbers of articles to be screened. Retrieval of imaging studies could, therefore, be improved by the development of subheading terms for MR imaging and CT or by indexing CT studies with the "ra/fs" floating subheading. Imaging terms used as floating subheadings have other uses when they are combined with appropriate clinical terms—for example, to retrieve imaging studies on particular clinical conditions. However, the indexing of new imaging technologies may be an ongoing problem because newer imaging techniques are continually being developed (29).

We recognize that there were a number of limitations in our study. The derivation and validation sets were identified from specific imaging journals and thus were not representative of all the journals that publish imaging studies. We chose imaging journals to obtain a high yield and wide range of diagnostic imaging articles for the DRS and VRS sets because we did not have the resources to hand search a broader set of journals. Although this might have limited the generalizability of our search strategy, we are not aware of any evidence that the quality of indexing imaging journals is different from the quality of indexing other types of journals. The higher prevalence of relevant articles in the imaging journals does, however, explain why the PPVs in the derivation (33.9%) and validation (40.9%) sets were higher than those achieved when the original search strategy was applied to MEDLINE (22.4%) in a pilot study (30). We also did not independently validate the final revised search strategy, for which modifications were made on the basis of studies from the VRS set that were missed.

In conclusion, we believe that we have developed a search strategy to identify studies of the diagnostic performance of imaging techniques that has a good balance between high sensitivity (to ensure that as many relevant studies as possible are retrieved) and reasonable specificity (to ensure that the number of articles to be reviewed is feasible and that there are fewer irrelevant articles). Our final search strategy may be improved further by including additional free text terms for certain imaging modalities; however, further external validation of this revised strategy is required. We believe that our strategy could be used to develop a provisional broad register of studies that assess the performance of any diagnostic imaging test and that such a register would be timely because the volume of diagnostic imaging studies is increasing rapidly; in our study, it more than tripled between 1975 and 2005. Such a register might need to be complemented with hand searches of key journals to identify the studies missed with electronic searching. Our search strategy could also be used for particular systematic reviews in which the addition of the diagnostic search component would reduce the number of articles that needed to be screened with use of a search strategy of imaging and clinical condition terms only (31). The imaging component of the search strategy also has the potential to be used in combination with the appropriate clinical terms to identify any imaging modality for the detection or diagnosis of a particular clinical condition.

	ADVANCES IN KNOWLEDGE

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCES IN KNOWLEDGE IMPLICATIONS FOR PATIENT CARE References

 

• We have developed a search strategy for MEDLINE that has retrieval characteristics adequate for the detection of studies on assessment of the diagnostic performance of imaging techniques.
  
• The diagnostic and imaging components of the search strategy may be used to retrieve any diagnostic study or a study of any imaging modality for the diagnosis or detection of a particular condition when the components are combined with clinical search terms related to the condition of interest.
  

	IMPLICATIONS FOR PATIENT CARE

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCES IN KNOWLEDGE IMPLICATIONS FOR PATIENT CARE References

 

• Identification of studies of diagnostic imaging tests is necessary for evidence-based radiologic practice and the development of radiology guidelines.
  
• Radiologists could use the described search strategy to identify studies relevant to their clinical practice.
  

	ACKNOWLEDGMENTS

 
We thank the Royal College of Radiologists, London, England, for its support and contribution to the hand searching of journals.

	FOOTNOTES

 

Abbreviations: CI = confidence interval • DRS = derivation reference standard • MeSH = medical subject headings • PPV = positive predictive value • VRS = validation reference standard

Author contributions: Guarantor of integrity of entire study, M.P.A.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, M.P.A., M.G.B.; statistical analysis, M.P.A., M.G.B., C.E.C.; and manuscript editing, all authors

Authors stated no financial relationship to disclose.

	References

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCES IN KNOWLEDGE IMPLICATIONS FOR PATIENT CARE References

 

1. Wilczynski NL, Walker CJ, McKibbon KA, Haynes RB. Assessment of methodologic search filters in MEDLINE. Proc Annu Symp Comput Appl Med Care 1993;:601–605.
2. Haynes RB, Wilczynski N, McKibbon KA, Walker CJ, Sinclair JC. Developing optimal search strategies for detecting clinically sound studies in MEDLINE. J Am Med Inform Assoc 1994;1(6):447–458.[Abstract/Free Full Text]
3. Haynes RB, Wilczynski NL. Optimal search strategies for retrieving scientifically strong studies of diagnosis from Medline: analytical survey. BMJ 2004;328(7447):1040.[Abstract/Free Full Text]
4. McKibbon KA, Walker-Dilks CJ. Beyond ACP Journal Club: how to harness MEDLINE for diagnostic problems. ACP J Club 1994;121(suppl 2):A10–A12.[Medline]
5. Deeks JJ, Morris JM. Evaluating diagnostic tests. Baillieres Clin Obstet Gynaecol 1996;10(4):613–630.
6. van der Weijden T, IJzermans CJ, Dinant GJ, van Duijn NP, de Vet R, Buntinx F. Identifying relevant diagnostic studies in MEDLINE: the diagnostic value of the erythrocyte sedimentation rate (ESR) and dipstick as an example. Fam Pract 1997;14(3):204–208.[Abstract/Free Full Text]
7. Deville WL, Bezemer PD, Bouter LM, Deville WL, Bezemer PD, Bouter LM. Publications on diagnostic test evaluation in family medicine journals: an optimal search strategy. J Clin Epidemiol 2000;53(1):65–69.[CrossRef][Medline]
8. Bachmann LM, Coray R, Estermann P, Ter Riet G. Identifying diagnostic studies in MEDLINE: reducing the number needed to read. J Am Med Inform Assoc 2002;9(6):653–658.[Abstract/Free Full Text]
9. Vincent S, Greenley S, Beaven O. Clinical evidence diagnosis: developing a sensitive search strategy to retrieve diagnostic studies on deep vein thrombosis—a pragmatic approach. Health Info Libr J 2003;20(3):150–159.[CrossRef][Medline]
10. Harris KM, Kelly S, Berry E, et al. Systemic review of endoscopic ultrasound in gastro-oesophageal cancer. Health Technol Assess 1998;2(18):i–iv, 1–134.[Medline]
11. Berry E, Kelly S, Hutton J, Harris KM, Smith MA. Identifying studies for systematic reviews: an example from medical imaging. Int J Technol Assess Health Care 2000;16(2):668–672.[CrossRef][Medline]
12. Mijnhout GS, Hooft L, van Tulder MW, Deville WL, Teule GJ, Hoekstra OS. How to perform a comprehensive search for FDG-PET literature. Eur J Nucl Med 2000;27(1):91–97.[CrossRef][Medline]
13. Haslinghuis-Bajan LM, Mijnhout GS, Hoekstra OS, Hooft L, van Tulder MW, Hoekstra OS. Literature search strategy for gamma camera emission tomography using coincidence imaging. Eur J Nucl Med 2001;28(3):396–397.[Medline]
14. Mijnhout GS, Riphagen II, Hoekstra OS. Update of the FDG PET search strategy. Nucl Med Commun 2004;25(12):1187–1189.[CrossRef][Medline]
15. Centre for Evidence-based Medicine, Oxford, Web site, 1995–2008. http://www.cebm.net/index.aspx?0-1038. Accessed March 8, 2007.
16. PubMed. PubMed clinical queries. PubMed Web site. http://www.ncbi.nlm.nih.gov/entrez/query/static/clinical.html. Updated September 27, 2007. Accessed March 8, 2007.
17. Scottish Intercollegiate Guidelines Network. Search filters. Scottish Intercollegiate Guidelines Network Web site, 2001–2008. http://www.sign.ac.uk/methodology/filters.html. Updated January 4, 2008. Accessed March 8, 2007.
18. University of Sheffield, School of Health and Related Research. Searching for diagnostic studies. University of Sheffield Web site, 2008. http://www.shef.ac.uk/scharr/ir/adept/diagnosis/searching.htm. Updated April 11, 2007. Accessed March 8, 2007.
19. University of Rochester, Edward G. Miner Library. Evidence-based healthcare. University of Rochester Web site, 2006. http://mdl.miner.rochester.edu/mdl/newmdl.cgi?categories=EBMRES. Updated January 25, 2008. Accessed March 8, 2007.
20. Mackenzie R, Dixon AK. Measuring the effects of imaging: an evaluative framework. Clin Radiol 1995;50(8):513–518.[CrossRef][Medline]
21. ISI Web of Knowledge Service for UK Education. The Thomson Corporation Web site, 2008. http://admin=apps.isiknowledge.com/JCR/JCR?SID=1BK8f21bdgg1MmKFfCm. Updated October 18, 2007. Accessed March 8, 2007.
22. Irwig L, Tosteson AN, Gatsonis C, et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann Intern Med 1994;120(8):667–676.[Abstract/Free Full Text]
23. Reid MC, Lachs MS, Feinstein AR. Use of methodological standards in diagnostic test research: getting better but still not good. JAMA 1995;274(8):645–651.[Abstract/Free Full Text]
24. Jaeschke R, Guyatt G, Sackett DL; for the Evidence-based Medicine Working Group. Users' guides to the medical literature. III. How to use an article about a diagnostic test. A. Are the results of the study valid? JAMA 1994;271(5):389–391.[Abstract/Free Full Text]
25. Jaeschke R, Guyatt GH, Sackett DL; for the Evidence-based Medicine Working Group. Users' guides to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? JAMA 1994;271(9):703–707.[Abstract/Free Full Text]
26. How good is that test? Bandolier. http://www.jr2.ox.ac.uk/bandolier/band26/b26-2.html. Accessed March 8, 2007.
27. Diagnostic testing emerging from the gloom? Bandolier. http://www.jr2.ox.ac.uk/bandolier/band70/b70-5.html. Accessed March 8, 2007.
28. Bossuyt PM, Reitsma JB, Bruns DE, et al. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. Radiology 2003;226(1):24–28.[Abstract/Free Full Text]
29. Shen MY, Calabretta N, Cavanaugh S, Datwani N, Lew C, Dadhania M. Analysis of current nuclear cardiology literature in MEDLINE database: a study of gated SPECT imaging using PubMed. J Nucl Cardiol 2003;10(6):650–655.[CrossRef][Medline]
30. Astin MP. Development of a search strategy for identifying diagnostic imaging studies of test accuracy, and a feasibility study for a register of diagnostic imaging studies. Aberdeen, Scotland: Health Services Research Unit, University of Aberdeen Press; 2003.
31. Fraser C, Mowatt G, Siddique R, Burr J. Searching for diagnostic test accuracy studies: an application to screening for open angle glaucoma. Presented at 14th Cochrane Colloquium, Dublin, Ireland, October, 2006.

This Article Abstract Figures Only Full Text (PDF) Appendix E1 Supplemental Tables All Versions of this Article: 2472070101v1 247/2/365    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Astin, M. P. Articles by Grimshaw, J. M. Search for Related Content PubMed PubMed Citation Articles by Astin, M. P. Articles by Grimshaw, J. M.