Abstract
Efficient literature search is essential to the practice of Evidence-Based Medicine. PubMed provides freeaccess to one of the largest searchable biomedical databases. Efficient literature search using PubMed requiresa good understanding of the available search strategies and tools. In this article we present a step-by-stepapproach for performing literature search using PubMed. Several PubMed tools including ‘Single CitationMatcher’, ‘Clinical Queries’, ‘Clipboard’, ‘Field Tags’, and ‘Cubby’ are highlighted using case basedscenarios. ©

INTRODUCTION

Evidence-based medicine (EBM) is ‘judiciousapplication of the best research evidence integratedwith individual physicians’ clinical expertise andpatient values’.1 Several physician surveys suggest thatmost patient encounters generate one or more clinicalquestions that remain unanswered and need a quickand focused literature search.2-4 Thus, in anobservational study at a university setting, an averageof five unanswered questions were raised for everypatient encounter.2 With biomedical informationdoubling every five years,5 the standard textbooks andreviews usually lag behind in their ability to provide themost current information on relevant clinical questions.6

The practice of EBM can be summarized into fourbasic steps: 1) develop a focused clinical question;2)search for the best evidence; 3) critically appraise theevidence; 4) apply the evidence and evaluate theoutcome. In this review, we will present the techniquesinvolved in formulating a well-built clinical questionand searching the literature using the PubMed database.

*Assistant Professor, **Associate Professor, Department of
Medicine, Mayo Clinic College of Medicine, Mayo Clinic,
Rochester, MN, USA.

Several electronic databases are available forsearching citations. Most of these are subscription basedand may not be freely available to the individual user.The source of biomedical information on the web can bedivided into two major categories: 1) original publishedarticles in the journals; and 2) synthesized informationavailable as systematic reviews, synopses or decision support systems.7 However, a lack of experience withsearch techniques often leads to missing many relevantcitations that may have an impact on patient care.8

MedLine is the U.S. National Library of Medicine’spremier bibliographic database. MedLine covers thefields of medicine, nursing, dentistry, veterinarymedicine, health care system, pre-clinical sciences, anda few other areas of the life sciences. MedLine containscitations from over 4,600 journals, published in 70countries. Over 14 million records are indexed inMedLine with most of the records from English languagesources or having English abstracts.

The MedLine database can be searched with severalsearch engines. The two most popular are Entrez andOvid. Entrez is the life sciences search engine that ispart of the National Center for BiotechnologyInformation’s retrieval system. Entrez searches severaldatabases including PubMed. PubMed includesMedLine, Old MedLine (1951 to 1965), articles from somenon-MedLine journals and issues of journals publishedprior to their selection for MedLine indexing, and out-of-scope articles from ‘selectively indexed’ MedLinejournals. Thus PubMed database is more comprehensivethan MedLine. Furthermore, PubMed is more currentand allows access to citations even prior to their indexingwith MedLine.

The initial essential requirement for performing aneffective search using PubMed is the ability to generatea well-built clinical question. The individual keycomponents of this question can then be used as searchterms in PubMed. Three basic steps are recommendedfor formulating the search question: 1) define thecomponents of the search question; 2) identify the studydesign that will answer the search question; 3) identify the study type that will answer the search question.

Each search question has four components remembered best by the mnemonic-PICO.9

1.Patient (P) – What is the patient group of intereste.g. acute cholecystitis.

2.Intervention (I) – What is the intervention of intereste.g. expectant management.

3.Comparison (C) – What is the comparisonintervention of interest e.g. surgery.

4.Outcome (O) – What is the primary outcome e.g.survival.

A search is usually started with two of the fourcomponents, usually patient and intervention.Depending on the question and retrieval, the search canbe further focused by adding rest of the components ofthe question.

Identify the study design that will answer the search question - The study design for most therapeuticinterventions is likely to be a randomized double blindclinical trial (RCT). For clinical questions where severalRCTs are reported in the literature, a systematic reviewor meta-analysis might be the most appropriate articleto review. The optimal study design to assess prognosisand harm is likely to be a prospective cohort study.

Identify the study type that will answer the search question - Most clinical questions can be answered byone of the four recognized study types i.e. diagnosis,prognosis, therapy and harm. Several databasesincluding PubMed provide appropriate search filters toretrieve citations with the particular study design.

The rest of this article will cover the important featuresof PubMed. Clinical scenarios will be used to elaboratethe search options. Readers are encouraged to conductthe searches as they go along in the article. All theactions that readers have to take are italicized, with theterms that they have to type placed in double quotes.Individual PubMed options are placed in single quotes.

Log on to www.pubmed.gov

The PubMed homepage has three key areas that are frequently used (Fig. 1):

1.‘Query box’ – The ‘Query box’ is used to enter thesearch terms as a free text entry area.

2.‘Features bar’ – Immediately beneath the ‘Querybar’, the ‘Features bar’ provides access to ‘Limits’,‘Preview/Index’, ‘History’, ‘Clipboard’, and‘Details’.

3.‘Side bar’ – The blue ‘Side bar’ has some of theadvanced PubMed features including the ‘MeSHDatabase’, ‘Single Citation Matcher’, ‘ClinicalQueries’ and ‘Cubby’.

In addition PubMed homepage also has a black ‘Menu bar’ at the top that allows the user to link to other Entrezdatabases.

MeSH Database

It is important to have a basic understanding of theMeSH database prior to exploring actual searchstrategies further. Medical subject headings (MeSH)database is the hierarchical library that is used to indexcitations. MeSH terms are arranged in a tree like structurewith the more general terms representing the stem andnarrower and more focused terms representing the distalbranches. Each citation that is indexed with MedLine isassigned up to 15 MeSH terms based on the content ofthe article. An important feature of PubMed thatincreases search yield is ‘automatic term explosion’.This means that when a search term is entered in PubMedthat matches with a MeSH term, PubMed automaticallyincludes all the specific narrower terms that areassociated with the search term.

Select MeSH Database from ‘Side bar’.

Type “Lung cancer” in the ‘Query box’ and select‘Go’.

Select ‘Lung neoplasms’.

Look at the ‘Subheadings’, ‘Entry Terms’ and thenscroll down to look at the MeSH tree associated withlung cancer. Typing any of these ‘Entry Terms’ in the‘Query box’ leads PubMed to match that term to lungcancer. Selecting ‘Subheadings’ allows you to restrictsearch to the particular subheadings within the MeSH term.

Basic PubMed search features - ‘Boolean logic’,‘Limits’, ‘Display options’

Boolean logic represents relationships betweensearch terms. The three Boolean operators are ‘AND’,‘OR’ and ‘NOT’. The operator ‘AND’ is used to retrievecitations that contain all the search terms, while theoperator ‘OR’ is used to retrieve citations that contain atleast one of the search terms. The operator ‘NOT’ is usedto restrict citations containing a particular search term.The use of ‘NOT’ is however confusing and should beavoided. While joining two search terms using Boolean logic, the Boolean terms should be typed in capitals (e.g.“AND” instead of “and”).

The ‘Limits’ option allows the user to narrow thesearch thereby retrieving fewer citations. The mostcommonly used ‘Limits’ are Publication Types, Ages,Publication Dates and Subsets.

CASE 1

A 50-year old male with new onset type II diabetesconsults you for lipid management. Patient has an LDLCholesterol of 96 mg/dl with HDL Cholesterol of 48 mg/dl. He recently heard that diabetic patients with evenlow LDL cholesterol might derive benefit from the use ofstatin medications. Your question is: Do patients withdiabetes and low LDL cholesterol have better outcomesif they are treated with a statin medication compared tono treatment?

Log on to www.pubmed.gov

Type in the search term, “diabetes AND low LDL ANDtreatment”. Look at the search results. You retrieve over onethousand citations.

Select ‘Limits’ and make the following selectionswithin ‘Limits’:

Publication types - Randomized controlled trials

Ages – All Adult: 19+ years

Publication Date – From 2002 To 2004

Subsets – Core clinical journals (Core Clinical Journalsis a list of 120 clinical journals that are likely to beavailable at most medical school libraries)

The final screen appearance is shown in Fig. 2.

Click Go and look at the output. You have much fewerarticles that are more relevant.

The article by Colhoun H et al published in Lancet in2004 addresses the specific question that you aresearching.10

Now look at the way the citations are displayed.PubMed defaults to show citations in the summaryformat, in a set of 20, with the citations arranged in a chronologic order with the most recent citation shownfirst. The way citations are displayed, number of citationsand how they are sorted can all be changed.

In the option ‘Display’ scroll down and select ‘Brief’.

In the option ‘Sort’ scroll down and select ‘Author’.

Click on ‘Display’.

Citations are now displayed in the ‘Brief’ format andarranged in an alphabetical order based on the author’sname.

Now select the ‘Back’ button to reach the previous screen.

Select the article by Colhoun H et al by checking the box to the left of the citation.

Scroll down in ‘Display’ and select ‘Abstract’.

Click on ‘Display’.

The ‘Abstract’ format is now seen. You can alsoperform the same function by simply clicking on theauthors’ names that is hyper-linked to the ‘Abstract’format.

Select ‘Clinical Queries’ from the ‘ Sidebar’.

Check the box to the left of ‘Systematic Reviews’.

In the search box enter the term, “Acupuncture AND back pain”

Look at the number of citations and see if your search retrieves relevant articles.

Now select the ‘Back’ button to reach the previous screen.

Check the box to the left of ‘Clinical Queries using Research Methodology Filter’.

In the ‘Category’, select ‘Therapy’ and in ‘Emphasis’, select ‘Specific search (narrow)’.

Your search term “Acupuncture AND back pain” should already be present in the ‘Query box’.

Select ‘Go’.

Look at the number of citations and see if your searchretrieves relevant articles. Most of the articles retrievedwould be clinical trials.

Now select the ‘Back’ button again to reach theprevious screen.

Change the Emphasis to ‘Sensitive search (broad)’.

Select ‘Go’.

Look at the number of citations and see if your searchretrieves relevant articles. You will retrieve a greaternumber of articles compared to the specific search andsome of the articles will likely have lesser relevance.

‘Clipboard’ helps you temporarily store citations forup to 8 hours. This option is particularly helpful whenyou are doing several different searches and would liketo select and save citations and work with them together.With the citations stored in the ‘Clipboard’, all thefunctions can be performed that are otherwise possiblewith a standard PubMed search.

Go back to the previous search just completed.

Select the first three citations by checking the box totheir left.

Go to the ‘Send to’ button. Scroll down to ‘Clipboard’and select ‘Clipboard’

Select ‘Send to’ button.

The message will appear on the left side of thewindow stating, ‘3 items were added to the Clipboard’(Fig. 4).

Select ‘Clipboard’ again from the ‘Features bar’. Look atthe 3 citations.

You can change ‘Display’ or ‘Sort’ these items like any PubMed search output. Note that if you want tosend all the citations obtained from a search to the‘Clipboard’, do not select any of the citations and selectthe option ‘Send to’ ‘Clipboard’. All the citations up to atotal of 500 will be sent to the clipboard.

While in the ‘Clipboard’, scroll down in the ‘Send to’button and select ‘Clip Remove’.

Select ‘Send to’.

Select ‘Clipboard’ again.

‘Clipboard’ should now be empty.

Work with other options in the ‘Send to’ button includingText, File, E-mail and Order. You can order full text of thearticles using service provided by ‘Loansome Doc’. Youwill need to first set up an account and the service is forfee.

Most of the search options that are used to do afocused search can be performed using the ‘Field Tags’.Field tags instruct PubMed to perform a search in thespecified fields. These tags are enclosed in squarebrackets and typed at the end of the search term. Some ofthe most useful tags are: author;6 publication date;10institutional affiliation of the author;6 journal title;6 andtitle of the citation.6

CASE 4

You are a general internist and would like to spendone year of research time at the Nicotine DependenceCenter at the Mayo Clinic. You are advised to developresearch ideas and contact Dr. Richard Hurt. You wouldlike to find out the articles that Dr. Hurt has beenpublishing in the last 10 years.

Enter the search term “Mayo[ad] AND Clinic[ad]” inthe ‘Query box’.

Select ‘Go’.

Look at the output. The results show all thepublications from the Mayo Clinic that are indexed inthe PubMed.

Add to the search term “hurt r[au]”. The final searchterm will be, “Mayo[ad] AND Clinic[ad] AND hurtr[au]”.

Look at the output. The results show all thepublications from the Mayo Clinic authored by Dr. RHurt.

Add to the search term “1994:2004[dp]”. The finalsearch term will be, “Mayo[ad] AND Clinic[ad] ANDhurt r[au] AND 1994:2004[dp]”.

Look at the output. The results show all thepublications from the Mayo Clinic authored by Dr. RHurt between 1994 and 2004.

Add to the search term “smoking”. The final searchterm will be, “Mayo[ad] AND Clinic[ad] AND hurt r[au] AND 1994:2004[dp] AND smoking”.

Look at the output. The results show all thepublications on smoking from the Mayo Clinic that areauthored by Dr. R Hurt and published between 1994and 2004.

A list of all the ‘Field tags’ is available on the PubMedHelp/FAQ page.

Select ‘Help’ in the ‘Side bar’.
Scroll down to ‘References.’
Select ‘Search Field Descriptions and Tags.’

‘Cubby’ helps the user to store search strategies forfuture reference. The most useful feature of the ‘Cubby’is to review the recent articles published on the topicsince the last search was completed. ‘Cubby’ feature isavailable for free but does require setting up a user nameand password.

CASE 5

You are an expert on the drug therapy of pulmonarytuberculosis. You would like to remain current with theliterature on this topic.

Select the ‘Cubby’ feature from the ‘Side bar’.
Set up your own ‘Cubby’ account.

Now go back to PubMed homepage and enter thesearch term “pulmonary tuberculosis AND drugtreatment” in the ‘Query box’.

Log on to your ‘Cubby’ account. Your search term ispresent in the ‘Cubby Search Name’ box.

Select ‘Store in Cubby’.

The search is stored in your ‘Cubby’ account. Thedate and time the search was performed is also stored.

Re-access your ‘Cubby’ account a few days from yourfirst search.

Select the search by checking the box to the left of it.

The output would comprise of citations added toPubMed since your last search.

CONCLUSIONS

Physicians practicing Evidence Based Medicine willfind it helpful to develop a well thought out literaturesearch strategy. PubMed is one of the mostcomprehensive biomedical database, is available for freeand offers several tools for performing a focusedliterature search. Familiarity with these search tools willhelp users perform efficient citation retrieval. PubMedprovides an online tutorial that might be helpful for bothnew and experienced users.

REFERENCES

1.Sackett DL SS, Richardson WS, Rosenberg W, Haynes RB.Evidence based medicine: How to Practice and Teach EBM. SecondEdition ed. London: Churchill Livingstone; 2000.

2.Osheroff JA, Forsythe DE, Buchanan BG, Bankowitz RA,Blumenfeld BH, Miller RA. Physicians’ information needs:analysis of questions posed during clinical teaching. AnnIntern Med 1991;114:576-81.

3.Green ML, Ciampi MA, Ellis PJ. Residents’ medicalinformation needs in clinic: are they being met? Am J Med2000;109:218-23.

4.Covell DG, Uman GC, Manning PR. Information needs inoffice practice: are they being met? Ann Intern Med 1985;103:596-9.

5.Mattox DE. Welcome to ARCHIVES CME. Arch OtolaryngolHead Neck Surg 2000;126:914.

6.Antman EM, Lau J, Kupelnick B, Mosteller F, Chalmers TC.A comparison of results of meta-analyses of randomizedcontrol trials and recommendations of clinical experts.Treatments for myocardial infarction. JAMA 1992;268:240-8.

7.Haynes RB. Of studies, syntheses, synopses, and systems:the "4S" evolution of services for finding currentbest evidence. ACP J Club 2001;134:A11-13.

8.Haynes RB, McKibbon KA, Walker CJ, Ryan N, Fitzgerald D,Ramsden MF. Online access to MEDLINE in clinical settings.A study of use and usefulness. Ann Intern Med 1990;112:78-84.

9.Oxman AD, Sackett DL, Guyatt GH. Users’ guides to themedical literature. I. How to get started. The Evidence-BasedMedicine Working Group. JAMA 1993;270:2093-5.

10.Colhoun HM, Betteridge DJ, Durrington PN, et al. Primaryprevention of cardiovascular disease with atorvastatin intype 2 diabetes in the Collaborative Atorvastatin DiabetesStudy (CARDS): multicentre randomised placebo-controlledtrial. Lancet 2004;364:685-96.

11.Hurt RD, Sachs DP, Glover ED, et al. A comparison ofsustained-release bupropion and placebo for smokingcessation. N Engl J Med 1997;337:1195-1202.

LATEST TREATMENT OF BLOCKED CORONARY THROUGH ECP IS THEEMERGING BIG TREND IN WEST. BE PIONEER IN EXTERNAL COUNTERPULSATION --------- HEAL SICK AND SECURE LIFE.

INNOVATIVE PHYSICIANS CAN GET RICH IN FAME, OTHERS CAN DELAY THE OPPORTUNITY

FDA (USA) approved for CAD, CCF --- ECP TECHNIQUE AND DEVICE DETAILS CONTACT NOW 098154-00433

ANAND CARDIOTECH (P) LTD. 73, CLUB ROAD, LUDHIANA 141 001 Ph. : 0161-2449474, 95 Fax : 5085084 Email : anandcardiotech@gmail.com Website : www.ecp-india.net