Education and Debate - 19 July 1997

Education and debate

How to read a paper

The Medline database

Trisha Greenhalgh

In 1928, in his introduction to Sceptical Essays, Bertrand

Russell wrote: “The extent to which beliefs are based

on evidence is very much less than believers suppose.”

Medical beliefs, and the clinical practices that are based

on them, are a case in point. Debate continues as to

whether scientific evidence alone is sufficient to guide

medical decision making, but few doctors would

dispute that finding and understanding relevant

research based evidence is increasingly necessary in

clinical practice. This article is the first in a series that

introduces the non-expert to searching the medical

literature and assessing the value of medical articles.

The Medline database

Over 10 million medical articles exist on library shelves.

About a third are indexed in the huge Medline

database, compiled by the National Library of Medicine

of the United States. The Medline database is exactly

the same, whichever company is selling it, but the

commands differ according to the software. Vendors

of Medline online and on CD ROM include Ovid

Technologies (ovid) and Silver Platter Information

(WinSPIRS).

Articles can be traced in two ways: by any word

listed on the database, including words in the title,

abstract, authors’ names, and the institution where the

research was done; and by a restricted thesaurus of

medical titles, known as medical subject heading

(MeSH) terms.

To illustrate how Medline works, I have worked

through some common problems in searching. The

scenarios have been drawn up using ovid software.

Problem 1: You are trying to find a known paper

Solution: Search the database by field suffix (title, author,

journal, institution, etc) or by textwords.

First, get into the part of the database which covers

the approximate year of the paper’s publication. If you

are already in the main Medline menu, select

“database” (Alt-B). If you know the approximate title of

the paper and perhaps the journal where it was

published, you can use the title and journal search keys

or (this is quicker) the .ti and .jn field suffixes. The box

shows some other useful field suffixes.

Thus, to find a paper called something like

“Confidentiality and patients’ casenotes,” which you

remember seeing in the British Journal of General Prac-

tice a couple of years ago,

type the following sequence:

1 confidentiality.ti

2 british journal of general practice.jn

3 1 and 2

You could do all this in one step:

1 confidentiality.ti and british journal of general

practice.jn

This step illustrates the use of the boolean operator

“and”; it will give you articles common to both sets.

Using “or” will simply add the two sets together.

Note that since 1988 the British Medical Journal is

abbreviated BMJ in ovid software, and Journal of the

American Medical Association is JAMA. Other useful field

suffixes to try when searching for a known article are

author (using the syntax haines-ap.au), institution (for

example, manchester.in), or title (for example,

evidence-based medicine.ti).

Summary points

Not all medical articles are indexed on Medline,

and many that are have been misclassified

Searching by textword can supplement a search

by MeSH headings

To increase the sensitivity of a search, use the

“explode” command and avoid using subheadings

Scan titles on screen rather than relying on the

software to find the most valid or relevant ones

Useful search field suffixes (ovid)

Syntax

.ab

.au

.jn

.me

.ti

.tw

.ui

.yr

Meaning

Word in abstract

Author

Journal

Single word, wherever it may

appear as a MeSH term

Word in title

Word in title or abstract

Unique identifier

Year of publication

Example

epilepsy.ab

smith-r.au

lancet.jn

ulcer.me

epilepy.ti

epilepsy.tw

91574637.ui

87.yr

This is the first

of 11 articles

introducing

non-experts to

finding medical

articles and

assessing their

value

Unit for

Evidence-Based

Practice and Policy,

Department of

Primary Care and

Population

Sciences, University

College London

Medical School/

Royal Free Hospital

School of Medicine,

Whittington

Hospital, London

N19 5NF

Trisha Greenhalgh,

senior lecturer

p.greenhalgh@ucl.

ac.uk

BMJ 1997;315:180–3

180 BMJ VOLUME 315 19 JULY 1997

Problem 2: You want to answer a specific question

Solution: Construct a focused (specific) search by combining

two or more broad (sensitive) searches.

I was recently asked by the mother of a young girl

with anorexia nervosa whose periods had ceased to

prescribe oral contraceptives for her so as to stop her

bones thinning. This seemed a reasonable request,

though there were ethical problems to consider. But is

there any evidence that taking oral contraceptives in

these circumstances really prevents long term bone

loss? I decided to explore the subject using Medline. To

answer this question, you need to search very broadly

under “anorexia nervosa,” “osteoporosis,” and “oral

contraceptives.” The search described below involves

articles from 1992; when replicating it, make sure the

database you are searching goes back that far. Type:

1 anorexia nervosa

You have not typed a field suffix (such as .tw), so the

ovid system will automatically try to “map” your

request to one of its standard medical subject headings

(abbreviated MeSH and colloquially known as “mesh

terms”). (Note that not all Medline software packages

will automatically map your suggestion to MeSH

terms. With Silver Platter search software, for example,

you need to enter your heading and click the “suggest”

button.) For this example, the screen offers you either

“eating disorders” or “anorexia nervosa” and asks you

to pick the closest one. Choose “anorexia nervosa”

(space bar to highlight the text, then press “return”).

The screen then asks you whether you want to

“restrict to focus.” Do you only want articles which are

actually about anorexia nervosa, or do you want any

article that mentions anorexia nervosa in passing? Let’s

say we do want to restrict to focus. Next, the screen

offers us a choice of subheadings, but we’ll ignore these

for a moment. Select “Include all subheadings.” We

could have got this far using a single line command:

2 *anorexia nervosa/

The * shows that the term is a major focus of the arti-

cle, and the / represents a MeSH term. You should

have about 750 articles in this set.

Similarly, to get articles on osteoporosis (which is

also a MeSH term), use the following single line

command:

3 osteoporosis/

You should get about 2200 articles. Note that in ovid,

if you know that the subject you want is an official

MeSH term, you can shortcut the mapping process by

typing a slash (/) after the word. Note also that we have

not used an asterisk here, because osteoporosis may

not be the focus of the article we are looking for.

Finally, put in the term “oral contraceptives” (with-

out an asterisk and without a slash) to see what the

MeSH term here is. You will be offered “contraceptives,

oral,” and if you had known this you could have used

the following command:

4 contraceptives, oral/

This set should contain around 1200 articles. You can

combine these three sets, either by using their set

numbers 1 and 2 and 3 or by typing the single line

command:

5 *anorexia nervosa/ and osteoporosis/ and

contraceptives, oral/

With this you will have searched over 4000 articles and

struck a single bull’s eye.

(If you don’t find it, check the

syntax of your search carefully, then try running the

same search through the previous five year database

using the Alt-B command.)

Problem 3: You want to get general information

quickly about a well defined topic

Solution: Use subheadings and/or the “limit set” options.

Subheadings are the fine tuning of the Medline

indexing system; they classify articles on a particular

MeSH topic into aetiology, prevention, therapy, and so

on. The most useful ones are listed in the box. I try not

to use subheadings unless I have unearthed an

unmanageable set of articles, since an estimated 50%

of articles in Medline are inadequately or incorrectly

classified by subheading. It actually doesn’t take long to

browse through 50 or so articles on the screen. It is

better to do this than to rely on the “limit set”

command (see box) to give you the best of the bunch.

The option “AIM journals” denotes all journals

listed in the Abridged Index Medicus

—

that is, the

“mainstream” medical journals. Alternatively, if you

want articles relating to nursing, rather than medical

care, you could limit the set to “Nursing journals.” This

is often a better way of limiting a large set than asking

for local holdings. If you are not interested in seeing

anything in a foreign language (even though the

abstract may be in English), select this option, again

bearing in mind that it is a non-systematic (indeed, a

very biased) way of excluding articles from your set.

Note that instead of using the “limit set” function

key you can use direct single line commands such as:

9 limit 4 to local holdings

10 limit 5 to human

Useful subheadings (ovid)

Syntax

/ae

/co

/ct

/di

/dt

/ed

/ep

/hi

/nu

/og

/pc

/px

/th

/tu

Meaning

Adverse effects

Complications

Contraindications (of drug)

Diagnosis

Drug therapy

Education

Epidemiology

History

Nursing

Organisation/administration

Prevention and control

Psychology

Therapy

Therapeutic use (of drug)

Example

thalidomide/ae

measles/co

propranolol/ct

glioma/di

depression/dt

asthma/ed

poliomyelitis/ep

mastectomy/hi

cerebral palsy/nu

health service/og

influenza/pc

diabetes/px

hypertension/th

aspirin/tu

PETER BROWN

Education and debate

181BMJ VOLUME 315 19 JULY 1997

Problem 4: Your search gives irrelevant articles

Solution: Refine your search as you go along in the light of

interim results.

Often, a search uncovers dozens of articles which

are irrelevant to your question. The boolean operator

“not” can help here. I recently undertook a search to

identify articles on surrogate endpoints in clinical

pharmacology research. My search revealed hundreds

of articles I didn’t want

—

all on surrogate motherhood.

The syntax to exclude the unwanted articles is:

1 (surrogate not mother$).tw

Deciding to use the “not” operator is a good example

of how you can (and should) refine your search as you

go along

—

much easier than producing the perfect

search off the top of your head. I used the truncation

symbol $ to find all words from a single stem, such as

mother, mothers, motherhood, and so on.

Another way of getting rid of irrelevant articles is to

narrow your textword search to adjacent words using

the “adj” operator. For example, the term “home help”

includes two very common words linked in a specific

context. Link them as follows:

1 home adj help.tw

Problem 5: The search gives no articles, or too few

Solution: Firstly, don’t overuse subheadings or the “limit set”

options. Secondly, search under textwords as well as MeSH

terms.Thirdly, learn about the “explode” command,and use it

routinely.

Many important articles are missed not because we

constructed a flawed search strategy but because we

relied too heavily on a flawed indexing system. For this

reason, you should adopt a “belt and braces” approach

and search under textwords as well as by MeSH terms.

After all, it is difficult to write an article on the

psychology of diabetes without mentioning the words

“diabetes,” “diabetic,” “psychology,” or “psychological,”

so the truncation stems “diabet$.tw.” and “psychol$.tw.”

would supplement a search under the MeSH

term “diabetes mellitus” and the subheading “/px”

(psychology).

Another important strategy for preventing incom-

plete searches is to use the powerful “explode”

command. The MeSH terms are like the branches of a

tree with, for example, “asthma” subdividing into

“asthma in children,” “occupational asthma,” and so on.

Medline indexers are instructed to index items by

using the most specific MeSH terms they can. If you

just ask for articles on “asthma” you will miss all the

articles indexed under “asthma in children” unless you

“explode” the term using the following syntax:

1 exp asthma/

Problem 6: You don’t know where to start searching

Solution: Use the “permuted index” option.

Let’s take the term “stress.” It comes up often, but

searching for particular types of stress would be labori-

ous and searching “stress” as a textword would be too

unfocused. We need to know where in the MeSH index

the various types of stress lie, and when we see that, we

can choose the sort of stress we want to look at. For

this, we use the command ptx (“permuted index”):

1 ptx stress

The screen shows many options, including post-

traumatic stress disorders, stress fracture, oxidative

stress, stress incontinence, and so on.

The command “ptx” is useful when the term might

be found in several subject areas. If your subject is a dis-

crete MeSH term, use the tree command. For example:

2 tree epilepsy

will show where epilepsy is placed in the MeSH

index

—

as a branch of “brain diseases,” which itself

branches into generalised epilepsy, partial epilepsy,

post-traumatic epilepsy, and so on.

Problem 7: Limiting a set loses important articles

but does not exclude those of low methodological

quality

Solution: Apply an EBQF (evidence based quality filter).

If your closely focused search still gives you several

hundred articles, and if applying subheadings or limit

set functions seems to lose valuable (and valid) papers,

you should insert a quality string designed to limit your

set to therapeutic interventions, aetiology, diagnostic

procedures, or epidemiology. Alternatively, you could

apply search strings to identify the publication type,

such as randomised controlled trial, systematic review,

or meta-analysis.

These EBQFs (evidence based quality filters), which

are listed in Appendix 1, are complex search strategies

developed by some of the world’s most experienced

medical information experts. You can copy them into

your personal computer and save them as strategies to

be added to your subject searches. Other search

strategies that will identify cohort studies, case-control

studies, and so on will soon be available from the UK

Cochrane Centre, Summertown Pavillion, Middle Way,

Oxford OX2 7LG ([email protected].uk).

Problem 8: Medline hasn’t helped

Solution: Explore other medical and paramedical databases.

Entry of articles onto the Medline database is open

to human error, both from authors and editors who

select key words for indexing, and from the librarians

who group articles under subheadings and type in the

abstracts. In addition, some sections of indexed

journals are not available on Medline (for example, the

News section of the BMJ). According to one estimate,

40% of material which should be listed on Medline can,

in reality, only be accessed by looking through all the

journals again, by hand. Furthermore, a number of

important medical and paramedical journals are not

covered by Medline at all. It is said that Medline lacks

comprehensive references in the fields of psychology,

medical sociology, and non-clinical pharmacology.

If you wish to broaden your search to other

electronic databases, ask your local librarian where you

could access the following:

AIDSLINE

—

Covers AIDS and HIV back to 1980.

x Allied and Alternative Medicine

—

Covers complementary

and alternative medicine.

x American Medical Association Journals

—

Provides the full

text of JAMA plus 10 specialty journals produced by the

American Medical Association; from 1982.

Useful “limit set” options

AIM journals

Nursing journals

Dental journals

Cancer journals

Review articles

Editorials

Abstracts

Local holdings

English language

Male

Human

Publication year

Education and debate

182 BMJ VOLUME 315 19 JULY 1997

x ASSIA

—

An applied social sciences database covering

psychology, sociology, politics, and economics since 1987.

All documents have abstracts.

x Cancer-CD

—

A compilation by Silver Platter of cancerlit

and Embase cancer related records from 1984. The CD

ROM version is updated quarterly.

x CINAHL

—

The nursing and allied health database

covering all aspects of nursing, health education, occupational

therapy, social services in health care, and other related disci-

plines from 1983. The CD ROM version is updated monthly.

x Cochrane Library

—

The Cochrane Controlled Trials Regis-

ter (cctr), Cochrane Database of Systematic Reviews (cdsr),

Database of Abstracts of Reviews of Effectiveness (dare), and

Cochrane Review Methodology Database (crmd) are

updated quarterly; authors of systematic reviews on cdsr

undertake to update their own contributions periodically.

x Current Contents Search

—

Indexes journal issues on or be-

fore their publication date. It is useful when checking for the

very latest output on a subject. Updated weekly; from 1990.

x Current Research in Britain

—

The British national research

database of trials in progress.

x DHData (formerly DHSS-Data)

—

The database of the UK’s

Department of Health indexes articles covering health service

and hospital administration; from 1983.

x Embase

—

Focuses on drugs and pharmacology but also

includes other biomedical specialties. It is more up to date

than Medline and has better European coverage. The CD

ROM version is updated monthly.

x HELMIS

—

The Health Management Information Service

at the Nuffield Institute of Health, Leeds, UK, indexes

articles on health service management.

x Psychlit

—

Produced by the American Psychological

Association as the computer searchable version of

Psychological Abstracts; covers psychology, psychiatry, and

related subjects; journals are included from 1974 and books

from 1987 (English language only).

x Science Citation Index

—

Indexes references cited in articles

as well as the usual author, title, abstract, and citation of arti-

cles themselves. Useful for finding follow up work done on a

key article and for tracking down addresses of authors.

x SHARE

—

Based at the King’s Fund library in London;

published and ongoing research into the health of, and

health services for, black and minority ethnic groups.

x Toxline

—

Information on toxicological effects of chemi-

cals and drugs on living systems; from 1981.

x UNICORN

—

The main database of the King’s Fund,

London. Covers health, health management, health

economics, and social sciences. Particularly strong on

primary health care and the health of Londoners.

Thanks to Mr Reinhard Wentz, Ms Jane Rowlands, Ms Carol

Lefebvre, and Ms Valerie Wildridge for advice on this chapter. I

am grateful to Carol Lefebvre of the UK Cochrane Centre for

permission to reproduce the EBQFs in Appendix 1.

1 Caman D, Britten N. Confidentiality and medical records: the patient’s

perspective. Br J Gen Prac 1995;45:485-8.

2 Seeman E, Szmukler GI, Formica C, Tsalamandris C, Mestrovic R. Osteo-

porosis in anorexia nervosa: the influence of peak bone density, bone

loss, oral contraceptive use, and exercise. J Bone Mineral Res 1992;7:1467-

74.

3 Moher D, Fortin P, Jadad AR, Juni P, Klassen T, Le Lorier J, et al.

Completeness of reporting of trials published in languages other than

English: implications for conduct and reporting of systematic reviews.

Lancet 1996;347:363-6.

4 Bero L, Rennie D. The Cochrane Collaboration: preparing, maintaining,

and disseminating systematic reviews of the effects of health care. JAMA

1995;274:1935-8.

Appendix 1: Evidence based quality filters for everyday use

(a) Therapeutic interventions (What works?)

1 exp clinical trials 2 exp research design

3 randomized controlled trial.pt.

4 clinical trial.pt.

5 (single or double or treble or triple).tw.

6 (mask$ or blind$).tw.

75and6

8 placebos/ or placebo.tw.

91or2or3or4or7or8

(b) Aetiology (What causes it? What are the risk factors?)

1 exp causality

2 exp cohort studies

3 exp risk

41or2or3

1 exp “sensitivity and specificity”

2 exp diagnostic errors

3 exp mass screening

41or2or3

(d) Epidemiology

1 sn.xs

(This would find all articles indexed under any MeSH term with any of

“statistics,” “epidemiology,” “ethnology,” or “mortality” as subheadings.)

Appendix 2: Maximally sensitive search strings (to be used mainly for research)

(a) Maximally sensitive qualifying string for randomised

controlled trials

1 randomized controlled trial.pt.

2 controlled clinical trial.pt.

3 randomized controlled trials.sh.

4 random allocation.sh.

5 double−blind method.sh.

6 single−blind method.sh.

7 or/1-6

8 animal.sh. not human.sh.

9 7 not 8

10 clinical trial.pt.

11 exp clinical trials

12 (clin$ adj25 trial$).ti,ab.

13 ((single or double or treble or triple) adj25 (blind$ or

mas$)).ti,ab.

14 placebos.sh.

15 placebo$.ti,ab.

16 random$.ti,ab.

17 research design.sh.

18 or/10-17

19 18 not 8

20 19 not 9

21 comparative study.sh.

22 exp evaluation studies/

23 follow up studies.sh.

24 prospective studies.sh.

25 (control$ or prospectiv$ or volunteer$).ti,ab.

26 or/21-25

27 26 not 8

28 26 not (9 or 20)

299or20or28

In these examples, upper case denotes controlled vocabulary and lower

case denotes free text terms. Search statements 8, 9, 19, and 27 could be

omitted if your search takes too long a time to run.

(b) Maximally sensitive qualifying string for identifying systematic reviews

1 review, academic.pt.

2 review, tutorial.pt.

3 meta−analysis.pt.

4 meta−analysis.sh.

5 systematic$ adj25 review$

6 systematic$ adj25 overview$

7 meta-analy$ or metaanaly$ or (meta analy$)

8 or/1-7

9 animal.sh. not human.sh.

10 8 not 9

Search statements 9 and 10 could be omitted if your search seems to be taking

alongtimetorun.

The articles in this series are excerpts from How to read a

paper: the basics of evidence based medicine. The book can be

ordered from the BMJ Bookshop: tel 0171 383 6185/6245;

fax 0171 383 6662. Price £13.95 UK members, £14.95

non-members.

Education and debate

183BMJ VOLUME 315 19 JULY 1997