Closing the gap between research and practicet

JOHN R.GEDDESand PAULJ. HARRISON

Has Mrs A got schizophrenia? Would Mr B do

better with a tricyclic or a selective serotonin re
uptake inhibitor? Is Miss C likely to be recovered
in six months time? Is St John's wort an effective
antidepressant?

Clinical psychiatry involves making difficult

decisions about diagnosis, therapy and
prognosis. Sometimes we may be entirely
confident about our decisions, but often we
are uncomfortably aware that we are mak
ing a choice without being sure there is
convincing evidence to justify it. Maybe we
don't know or have forgotten what the
evidence is, or perhaps there isn't any.
All doctors have considerable informa
tion needs which are often unrecognised and
unmet (Smith, 1996). Clinicians turn to
colleagues, textbooks and reviews to keep
abreast of developments and to inform their
decision-making.
However these sources
suffer from a number of limitations and
biases
and may be out of date. For exam
pIe, conventional review articles may appear
authoritative but contain a considerable, and
often unknown, degree of subjective opinion,
such that deciding
which review to believe is
like deciding which toothpaste to use. It is a
question of taste rather than a question of
science(Oxman & Guyatt, 1988). On the
other hand, it is unrealistic to expect anyone
to locate, let alone keep up with, the two
million papers published in 20 000 biomedi
cal journals every year (Mulrow, 1994) or to
identify the tiny fraction which are both
scientifically sound and clinically important
to psychiatrists.
The effect of this unmet need for
accurate information is a knowledge
gap'
whereby many clinical decisions are made
with a greater degree of opinion and
uncertainty than is necessary. The gap is
filled by a number of other factors which
together influence the decisions we make:
such factors include the conceptual aetiolo
gical school to which we subscribe (e.g.
biological v. social psychiatry) and the
tSee commentaries, pp. 226227.

220

combination of experience and habits which

we accumulate during our career. In this
context, it is easy to understand the marked
differences between psychiatrists' decisions,
for example in the use of electroconvulsive
therapy (Pippard, 1992; Hermann et a!,
199S), continuation antipsychotics(Meise
et a!, 1994) and the treatment of depression
(Wells et a!, 1994). Variation in therapeutic
(and diagnostic) practice is justified, even
desirable, if all the variants are equally
effective or if there really is no evidence.
However, most psychiatrists would agree
this is unlikely always to be the case, and
that better application
of the existing
evidence would lead to greater uniformity
and to higher overall standards of psychia
tric care with improved outcomes.
Given the knowledge gap and reason to
doubt the effectiveness and objectivity of
traditional ways of keeping up with develop
ments in clinical practice, psychiatrists need a
new strategy to ensure that their clinical
decision-making is based upon the most
appropriate, accurate, and up-to-date infor
mation: evidence-based medicine (EBM).

evidence, EBM also aims to be a method

of self-directed, career-long learning.
Views about EBM tend to be polarised.
Critics of EBM caricature its proponents as
evangelists who fail to appreciate the corn
plexity of everyday medical practice and who
overlook the wisdom of experienced cmi
cians; the advocates of EBM view its oppo
nents as Luddites who have an overvalued
opinion of their clinical acumen. In this paper
we have tried to address both concerns. We
outline the principles and practice of EBM as
it applies

to psychiatry.

We

focus

upon

the

use of EBM to help a psychiatrist to make his

or her routine clinical decisions more efficient
and evidence-based. We also recognise and
discuss the various limitations of EBM. Our
stance, however, is unequivocally in favour of
EBM; the only sensible debate concerns how
it

can

be

applied

most

effectively

and

pragmatically.

HOW TO PRACTISE
EVIDENCE-BASED
MEDICINE
The practice of EBM focuses on five linked
activities, as outlined below.
Defining the clinical question to be

answered
Clear answers require clear questions. The
formulation of a precise clinical question
involves definition of the patient's problem,
identification
of the manoeuvre to be
performed, and specification of the outcome
of interest.
Example
Encouraged by the health authority, yourTrust is
trying to improve the local implementation of the
Care Programme Approach

WHAT IS EVIDENCE-BASED
MEDICINE?
EBM is the conscientious,explicit and
judicious use of the current best evidence
in making decisions about the care of
individual patients (Sackett et a!, 1996a).
EBM closes the gap between research and
practice by incorporating the advances in
clinical epidemiology and medical infor
matics into clinical activities (Evidence
Based Medicine Working Group, 1992).
EBM is intended to overcome the problems
outlined above: that clinicians may not
always make the best decisions; that they
may not realise this and, if they do, they may
not know how to improve the situation. By
providing clinicians with a set of skills
which allow them to base clinical decisions
on the best available, most up-to-date

(CPA). The official

guidance. which introduces the concept of tiered'

CPAfor different levelsofclinical need, is open to

several interpretations. Several ofyour colleagues
suggest that the CPA is worthless, that resources
for its implementation are not available, and that
the government should be encouraged to with

draw it. One cites a damning editorial (Anon

ymous, 995). Other colleagues feel that it should
be fully implemented for all patients and that the
purchasers should be billed accordingly. You sug

gest that althoughthe useofthe CPA is compul

sory, it will be easier to implement if you know
what effect it might be expected

to have. You

therefore decide to look for some evidence on its

effectiveness.Youposethe question:For
patients
with schizophrenia or other severe mental illness,
what are the effects ofCPA on the patient's clinical
state and service utilisation?

Searching for the evidence

Recent advances in information technology
have made it far easier to search the medical

literature. Access to electronic databases

such as the National Library of Medicine's
Medline is now widespread in medical
libraries, and the British Medical Associa
non offers free Medline access to members
with computers and modems. Increasingly,
medical journals offer World Wide Web
pages on the Internet, displaying contents,
abstracts and some full text articles.
Electronic databases are useful but they
have limitations. Not all journals are indexed,
the relevant articles can be difficult to locate,
and the searcher's access and time may be
limited. One solution to these problems is to
use systematic reviews prepared by others.
These are a major advance. Unlike a conven
tional reviewer, a systematic reviewer uses
explicit methods of searching for and criti
cally appraising the primary studies. If these
are comparable, the reviewer may then per
form a formal quantitative synthesis (meta@
analysis) of the results (Mulrow, 1987). As
well as summarising the effectiveness of
health care interventions, a systematic review
can reveal problems with the literature, for
example revealing duplicate publications
(Huston & Moher, 1996). Two increasingly
useful
sources
ofsystematic
reviews
ofhealth
care interventions,
including psychiatric
treatments, are the Cochrane Collaboration
(1996) and the NHS Centre for Reviews and
Dissemination (Sheldon, 1996).
While these initiatives fulfil an essential
role,thereis inevitably
a delaybetween
important new evidence appearing and it
being incorporated into a systematic review.
There is therefore a need for a method of
identifying, critically appraising and disse
minating new evidence as it becomes avail
able. One solution has been the introduction
oftwo journals
ofsecondarypublication

Table I A hierarchyof evidencefor therapy

the American Co!!ege of Physicians (ACP)

Journa! C!ub and Evidence-Based Medicine
(which both include psychiatric topics).
Their aim is to locate and summarise the
small number of articles, concerning ad
vances in diagnosis, therapy, prognosis,
aetiology and economic evaluation which
are clinically important, and to publish them
in the form of structured abstracts prepared
by trained clinical epidemiologists
with
commentaries provided by experienced din
icians. Now a new journal
Evidence-Based
Mental Health
is being planned to meet
the psychiatrist's needs more specifically.
Examplecontd:
You try to find a relevant systematic review of
randomisedcontrolled trials (RCTs).CPAis really
a variant of casemanagementand so you search
for reviews of case
management' in the Cochrane
Database ofSystematic Reviews on the Cochrane
Library. You find a systematic review ofcase man

agementfor people with severemental disorders

which appears to meet your needs (Marshall et a!,

996).

for the evaluation of treatments. However,

for many interventions, RCTs may not exist
and the practitioner has to use evidence
from the next level of the hierarchy. The
practical importance of the hierarchy is that,
when choosing between treatments, the
clinician selects the intervention which is
backed up by the best available evidence.
When evaluating a paper evaluating ther
apy, the main questions to ask are those
addressing the study's internal validity (are
the results trustworthy?) and external valid
ity

(can

they

be

applied

to

my

patients?)

(Table 2) (Sackett et a!, 1997). Critical

appraisal of an overview of RCTs asks
several further questions (Table 3).
Example
contd:
Youcriticallyappraisethe Cochrane review for its
validity and applicability to your problem.The

re

view includes nine RCTs ofcase management v.

standard care. The search strategy and inclusion

and exclusion criteria are clearly described and

there is an assessment ofthe quality ofthe indivi

dual studies.All useda form ofcase management

Appraising the evidence

Evidence needs to be critically appraised for
its reliability and whether it can be extra
polated to the particular clinical problem. A
fundamental
principle of EBM is that
evidence from clinical research studies is
more trustworthy than reasoning from basic
pathophysiology, psychology or clinical in
tuition (Evidence-Based Medicine Working
Group, 1992). Implicit in the process of
critical appraisal is the idea that there is
hierarchy in the quality of evidence derived
from different types of study, which is
inversely proportional to their susceptibility
tobias(that
is
thesystematic
deviation
of
the results from the truth). A commonly
used hierarchy of research designs for
evaluating therapy is shown in Table 1.
RCTs, or better still, systematic reviews
of RCTs, are the most reliable study design

which was comparable

to that included in the

CPA:standardcarewasdefined asthe normal Ic

vel of psychiatriccare provided in the areawhere
the trial was conducted.The studies examined a
rangeofoutcomes including:lossto follow-up, ad
mission to hospital, death, mental state at one

year, socialfunctioning at one year, quality of life

at oneyear andhealthcarecosts.Therewassome
variation in the results from study to study, butthe

reasonsfor this are investigatedandthe variation

is mainly quantitative rather than qualitative. You
decide that you can believe the results ofthe over
view. The studies in the overview

were per

formed in the UK and the USA and included

patientswith severe mental illness(however de
fined).This islikelyto includethe group of patients
which interest you.

Applying the evidence

Translating
research
findings
intoclinically
usable information is one of the most
challenging areas of EBM. It involves

Ia. Evidence from a meta-analysis of RCTs

lb. Evidencefrom at leastone RCT

la. Evidencefrom at leastonecontrolled study
without randomisation
llb. Evidence from at least one other type of
quasi-experimental
Ill.

study

Evidence from non-experimental

descriptivestudies,suchascomparative
studies,correlation studiesand
casecontrolstudies

IV. Evidencefrom expert committee reports,

Table 2

Questions

for evaluating an article about therapy

Are the results ofthis trial valid?

Was the assignment ofpatients to treatments truly randomised?
Were the groups similar at the start ofthe trial?

Were patientsandclinicians
keptblindto whichtreatmentwasbeingreceived?
Asidefrom the experimentaltreatment, were the groupstreated equally?
Were all the patients who entered the trial accounted for at its conclusion?
The evidence is valid and important. Should it be applied to my patient?
Isthe patient so different from those in the trial that the results can't help me?

or opinions and/or clinical experience

How great would the potential benefit of therapy actually be for my patient?

of respected authorities

Are my patient's values and preferences satisfied by the regimen and its consequences?

221

Table 3

care subjects. The Absolute Risk Reduction

(ARR) is 7.5%and the Number Needed toTreat

Guidelines for evaluating an overview of randomised controlled trials

(NNT) is 13(95% CI 840).Thismeans that 3 pa

Isit anoverview of RCTsofthe treatment you're interestedin?

Doesit includea methodssectionthat describes:
Whether andhow all the relevanttrials were identified andincluded?
How the validity ofthe individualtrials wasassessed?
Were the resultsconsistentfrom study to study?

tients have to be treated with case managementto

prevent one less patient from being lost to follow
up than would occur with standard care.The ratio

ofthe odds of being lost to follow-up in the case

management group to the odds of being lost to
follow-up

in the standard care group was 0.68

(95%CI0.530.88).
Youknowfromanauditof
Box I

your service,carried out before the introduction

ofthe CPA, that only 10%of patientswith severe

Indices for translating research results into clinical practice

mental illness were lost to follow-up

each year.

This is about 30% (0.100/0.327=0.306)

of the

CER in the review. Assumingthat case manage

ment has a constant effect in all patient groups,

Experimental

Control

treatment, X

treatment, Y

you can therefore adjust the NNT to apply to the

lossto follow-up rate of your serviceby dividingit

by this proportion (Cook & Sackett, 1995).This
produces a NNTofabout 42 (13/0.306). How
ever, among patients with a previous history of
lossofcontact, your drop-out rate was 80%. Ad

Positive
outcomeabNegative

justment in a similar way produces an NNT esti

mate ofabout five for this group of patients.

outcomecd
Admission to hospitalOneofthe key aims of com
munity care is to reduce the rate of hospital ad
mission. In the review, case management actually
increased the risk ofadmission; the CER was 9.4

Relativerisk (RR)

andthe EER30.7.TheARRwas
Il.3%and num
ber needed to harm (NNH) was 9 (6
15),This
meansthat for every nine patientscasemanaged
one more would be admitted to hospital than

Theratio ofthe proportionswith a positiveoutcomeontreatmentsX andY =@a/(a+cfl/[b/b+d))=EER/CER

would have occurred if standard care had been

Control EventRate(CER) = b/(b+d)

Experimental EventRate(EER) = a/(a+c)
AbsoluteRisk Reduction(ARR)

Thedifferenceinthe proportionswith a positiveoutcomeontreatmentsX andY (=CER

EER)
Oddsratio (OR)

used.Againthis figure could be adjustedto fit lo

Theratio ofthe oddsofa positiveoutcomeontreatmentsX andY (=(a/c)/(b/d)=ad/bc)

Number Needed to Treat(NNT)
how many patients need to be treated with treatment X to get one more posi

tive outcomethanwouldbeexpectedontreatmentY (=l/ARR)

cal readmission rates by making the same assump

tion ofconstant effect across subgroups.

You present these figures to your working

group. The group concludes that CPA is likely to

summarising the research findings in ways

which contain the maximum amount of
information and which are meaningful to
the clinician (or to patients or purchasers).
Many studies present their results in terms
of ratio measures such as the odds ratio and
the relative risk (see Box 1). However this
can lead to overestimates of the treatment
effect because they do not give any idea of
the control event rate (Fahey et a!, 1995;
Sackett et a!, 1996b). For example, if a
treatment
reduces
theriskofdeath,
itseffect
would be lessimpressive
if the baseline
mortality was 2% (reduced to 1% by the
treatment) than if the baseline risk was 80%
(reduced to 40% by the treatment),
althoughinboth instances
therelative
risk
of deathis0.5.Most commentatorsfavour
measures of effect such as the absolute risk
reduction or the number needed to treat,
which are more intuitive and more clinically
useful (Laupacis et a!, 1988). Unfortunately
at present, the reader often has to calculate
theseindices
from thereview.

222

Patients in RCTs are likely to be

different from the majority of patients
encountered in real-life clinical practice,
for example because patients consenting to
enter the trial are unrepresentative of all
subjects approached, or because comorbid
ity is an exclusion

criterion.

Hence

the need

for pragmatic RCTs on representative pa

tients (Simon et a!, 1995), and for extra
polation of RCT results to clinical situations
by combining them with appropriate event
rates derived from cohort studies, or by the
use of local data (see below: Cook &
Sackett, 1995; Glasziou & Irwig, 1995).
Example
contd:
The mainfindingsofthe review werethatcase man
agement assiststeams to maintain contact with pa

tients but alsoincreasesthe admissionrate.There

was no significanteffectonthe

otheroutcome vari

ables (such as reduction in clinical symptoms).

Lossto follow-up
151of 600 (Experimental
EventRate(EER)=25.2%)casemanagedsubjects
were lost to follow-up compared with 200 of 611
(Control Event Rate (CER)=32.7%) standard

be most useful for keeping contact with patients

who have previously dropped out of care. They
will concentrate on implementing full'
CPA for this
group initially, expanding to other groups as re

sourcesbecome available.The in-patient service

will need to be supported to cope with the pre

dicted increaseddemand. (The group alsonoted

that admissioncould be viewed asa positiveevent
rather than as a negative one
it is decided to
investigate reason for admissions locally)

The example makes it clear that interpreta

tion of evidence, and health care decisions,
are complex even when good quality cvi
dence is available and when EBM is
adopted. Evidence is helpful but not suffi
cient for medical decision-making: the key
aspect of EBM is that it ensures the best use
is made of the available evidence.
We have used an example of an evidence
based approach to a decision about therapy.
The same EBM process can be applied to
decisions about diagnosis (e.g. how useful are
screening instruments for detecting cases of
depression in primary care? How useful are
Schneider's first-rank symptoms in diagnos
ing schizophrenia?) (Geddes et a!, 1996a) or
prognosis (e.g. what is the risk of suicide in

someone who has taken an overdose? What

is the outcome following a first episode of
schizophrenia?). The Evidence-Based Mcdi
cine Working Group is publishing a series of
articles in the Journal of the American
Medical Association which provide a guide
to how to appraise and interpret articles
when using EBM to answer these different
types of question (Table 4).

Evaluation and development

evidence-based practice

of

The final stage of EBM involves the devel

opment and evaluation of one's own EBM
skills: the ability to ask answerable ques
tions, to search for evidence, and to appraise
itand to applyitappropriately.
Through a
recursive process, gaps in practice or teach
ing can be identified and rectified, and the
proportion of one's clinical decisions which
are soundly evidence-based can be audited.
There are several ways of developing the
numerical and critical appraisal skills neces
sary for EBM. EBM teaching workshops are
becoming widely available. Educational
prescriptions'
are useful for introducing
EBM into teaching and into one's own

clinical practice (Sackett et a!, 1997). A

prescription defines a problem for which
uncertainty exists and specifies a time and
date for the question to be answered.
Evidence is then sought, critically appraised,
and applied to the problem. Academic meet
ings and journal clubs can be similarly
structured (Gilbody, 1996). The content of
continuing medical education, especially
now that it is becoming obligatory, should
obviously be evidence-based too, and effec
tive strategies used to deliver it (Davis et a!,
1995).Finally,
practice
in generalcan be
come more evidence-based by the adoption
of evidence-based guidelines, often devised
nationally, but critically appraised and tai
lored for local implementation (NHS Centre
for Reviews and Dissemination, 1994; Eccles
eta!, 1996).

ADDRESSING CONCERNS
ABOUT EVIDENCE-BASED
MEDICINE
Apart from ad hominem attacks (Shahar,
1995), EBM has been criticised on a number
ofgrounds.Thesehaveoftenarisen
from a
misunderstanding of what EBM is and what

it can

do

(Sackett

et a!,

1997).

However,

there remain some important

concerns
about the theoretical and practical implica
tions of EBM.

Is EBM really different from what

goodclinicianshavealways
practised?
The idea that practice should be based on
good evidence is hardly new, and the aims of
EBM are the same as traditional medical
practice (Lewis, 1958). While exceptional
clinicians may always have kept fully up-to
date without using EBM, the existence of
variations in practice implies that this is not
generally true for all clinicians. At the very
least, by making the link between evidence
and practice efficient and explicit, EBM
allows outstanding clinicians to demonstrate
their prowess
and allows the rest of us to
emulate them.

Hypocrisy: the strategies of EBM

are flawed,

and there

Is no evidence

that EBM works

The methods of EBM are still being devel

oped and they need to be used carefully and
critically (Sackett et al, 1996a). For exam
plc, it is known that meta-analysis can be
Ikble 4 The Evidence-BasedMedicineWorking Group@s
Users
guideto the medicalliteratureseries
published
misleading (Eysenck, 1994; Egger & Smith,
in the JournalofcheAmericanMedicalAssociation(JAMA)
1995), and that interpretation of the same
evidence can come up with different con
clusions (Jackson & Sackett, 1996). Such
Oxman,A. D., Sackett,D. L. & Guyatt,G. H. (1993)1,Howtogetstarted.JAMA, 270,20932095.
factors point to the need for continuing
Guyatt, G. H., Sackett,D. L. & Cook, D,J.(1993)II.How to useanarticle abouttherapy or prevention.
refinement and testing of the EBM tools,
A, Are the resultsofthe studyvalid?JAMA,270,2598260
I.
including meta-analysis (Sim & Hlatky,
, & (1994)
II. How
to
use
an article
about
therapy
or
prevention.
B.What
were
the
results
1996), and emphasise that the clinician must
andwill they helpme in caringfor mypatients?JAMA,271,5963.
continue to use his or her critical faculties
Jaeschke, R. , Guyatt, G. H. & Sackett, D. L. (1994) III. How to use an article about a diagnostic test.
when appraising evidence, even when it is
A. Are the resultsofthe study valid?JAMA,271,38939
I.
presented on an evidence-based plate in the

& (1994)
III. How
to
use
an article
about
a diagnostic
test.
B. What
are
the
results
and
will
form of a systematic review. EBM does not
they help me in caring for my patients?JAMA, 271, 703707,
claim to be an infallible, discrete system
Levine,M.,Walter, S.,Lee,H., etal(l994)$V. How to useanarticle about harm.JAMA,271,16151619. which renders other approaches invalid;
rather, it complements and strengthens
Laupacis,A,,Wells, G., Richardson,W.S.,etal(1994)V. How to useanarticle about prognosis,JAMA,
traditional medical skills.
272,234237.
A related claim of hypocrisy is that there
Oxman, A. D., Cook, D.J. & Guyatt, G. H. (1994)Vl. How to use an overview,JAMA, 212, I 367l371.
is no proof that EBM is effective. Ideally,
Richardson,W,S.& Detsky,A. S.(1995)VII.How to useaclinicaldecisionanalysis.A. Are the resultsof
this would require an RCT showing that the
the study valid?JAMA, 273, 1292l295.
practice of EBM improves patient outcome.
Hayward,R,S. Wilson,M.C. Tunis,S.R.,eta! (1995)VIII.How to useclinicalpracticeguidelines.A. Are
Though there is no evidence for the effec
the recommendationsvalid?JAMA,274,570574.
tiveness of EBM as a system-level interven
Wilson, M. C., Hayward, R. S.,Tunis, S. R., et a! (1995) VIII. How to use clinical practice guidelines.
tion (and one could argue that such a study
B,Whatarethe recommendationsandwill they helpyou in caringfor your patients?JAMA,274,
is not feasible), there are empirical findings
16301632.
suggesting that the elements of EBM do
Guyatt, G. H,, Sackett, D. L. , Sinclair,J. C., et al(1995) IX. A method for grading health care
achieve their aims. First, the effectiveness of
critical
appraisal
(essentially,
theadoption
recommendations.JAMA,
274, 1800
1804.
of an EBM approach to medicine) has been
Naylor,C. D. & Guyatt,G. H. (1996)X. How to useanarticle reporting variationsinthe outcomesof
shown; medical students trained in this way
healthservices.JAMA,275,554558,
were more likely
than theirpeersto use

223

critical appraisal skills and keep up-to-date

with therapeutic advances several years after
qualifying (Bennett et a!, 1987; Shin et a!,
1993).

Second,

recent

observations

suggest

that the application of EBM strategies

improves outcome
for example the fall in
breast cancer death rates since the publica
tion of the meta-analysis by the Early Breast
Cancer Trialist Collaborative Group (Beral
eta!, 1995).
We would also argue that EBM has high
face validity. EBM represents an explicit,
coherent and comprehensive approach, en
abling clinicians to ensure that their practice
is based on the best available evidence.
Although formally unproven, it is difficult
to challenge the assumption that this will
lead to the best clinical outcomes.

Evidence is lacking, and what there

is is Inapplicable to real-life practice
The proportion of medical interventions
which are based on evidence has been
underestimated (Smith, 1991; Ellis et a!,
1995).

In psychiatry,

the

tendency

may

be

more marked because of the widely-held

beliefs about the nature of mental illness
(Kerr et a!, 1995; Priest et a!, 1996). This
can lead to pessimism about the feasibility
of applying EBM to psychiatry because it
may be thought that the speciality com
prises islands of evidence in a sea of
opinion, and that the interventions for
which evidence does exist are self-evidently
effective anyway. In fact, the proportion of
treatment decisions in psychiatry which are
based on RCT evidence is similar to that
observed in general medicine (Ellis et a!,
1995; Geddes et a!, 1996b; Summers &
Kehoe, 1996). However, it could be argued
that these data only apply to medical'
type
interventions and that most psychiatry lies
in grey
zones' of clinical practice which lie
outside the scope of EBM (Tanenbaum,
1993; Naylor, 1995). Indeed, these grey
zones may be considered to be the essence
of psychiatry since they include the nature
of the doctorpatient relationship, therapist
qualities, and even the effectiveness of
compulsory treatment. While we are not
suggesting that everything in psychiatry can
currently be based soundly on evidence
which we can find if only we know where
to look (see Greenhalgh, 1996), we would
argue, in line with Sullivan & MacNaugh
ton (1996), that the interview and other
key elements of psychiatry are amenable to
an evidence-based approach. For example,
one could carry out an RCT to establish

224

JOHN R.GEDDES,MRCPsych,PAULJ.HARRISON,MRCPsych,University Department of Psychiatry,Warneford

Hospital, Oxford

Correspondence: Dr J.Geddes, University Department of Psychiatry,Warneford Hospital, Oxford

0X3 7JX. Fax: 01865 793101; e-mail:john.geddes@psychiatry.oxford.ac.uk
(First received 18November 1996,final revision 20 November 1996,acceptedI May 1997)

whether scrupulously balanced prognostic

statements or unfailingly optimistic (but not
untrue) ones lead to better clinical out
comes.

EBM will be used as an excuse to cut

services and limit clinical freedom
There is an understandable anxiety that
purchasers, under the guise of evidence
based purchasing, will attempt to disinvest
in established services for which there is no
high quality (RCT) evidence of effectiveness
(Grahame-Smith, 1995). It is easy to accept
the aim of EBM to increase awareness and
implementation of genuine treatment ad
vances. Equally, hallowed procedures must
be abandoned with the same enthusiasm if
evidence of their relative ineffectiveness or
actual harm emerges (Eddy, 1994) as, for
example, in the case of high-dose antipsy
chotics (Hirsch & Barnes, 1994).
As health care rationing becomes cx
plicit, battles are likely to be fought over
old'
treatments which were introduced
before the current requirements for cvi
dence of effectiveness (Eddy, 1993). It is
over these interventions that disagreements
are most likely to occur between providers
and purchasers, since the two groups have
different priorities (Eddy, 1992): clinicians
should stress that absence of RCT evidence
of effectiveness is not the same as evidence
of ineffectiveness, and should take the view
that the burden of proof lies with those
who wish to disinvest in an established
treatment. Informed debate about service
provision will be facilitated by the use of
EBM. Moreover, if clinicians are able to
argue effectively for the introduction of
evidence-based
interventions,
EBM may
lead to increased costs rather than to
savings (Sackett et a!, 1996a).
Finally, there has been concern about
the loss of clinical freedom because of EBM.
However, it is only when the term is
misappropriated to mean the right to carry
on practising demonstrably inferior medi
cine that EBM
rightly
restricts clinical
freedom.

CONCLUSION
Psychiatrists must keep abreast of therapeu
tic advances, cope with rapidly changing
mental health policies, and face increasing
public expectations
and demands from
purchasers. We would argue that the best
response to these pressures is to use EBM,
since EBM optimises our clinical decisions
and allows us to justify them. We would
also suggest that it makes practice easier and
more efficient. As with the implications of
recent scientific advances, the changes in
practice which EBM requires are challen
ging. However, the EBM approach to
clinical decision-making is like a scalpel
compared with the meat-axe of topdown
policy-making (Eddy, 1990). Psychiatrists
will be better off adopting EBM and
applying it judiciously, rather than waiting
for it to be foisted upon us.

REFERENCES
Anonymous (1995)Care-management:a disastrousmistake.
Lancet.345, 399401.
Bennett, K.j., Sack.tt, D. L, Haynes, R. B., at a! (1981)
A controlledtrial ofteaching critical appraisalofthe clinical
literature to medicalstudents.Journalofthe AmericanMedical
Association,257,24512454.
B.ral,V., Harmon, C., Resyss,G., it al (I9@S)Suddenfall
in breastcancerdeath rates in Englandand Wales.Lancet,
345, 16421643.
Cochran. Collaboration (1996) TheC.ochrane
Library.
Oxford: Update Software.
Cook, R.J. & Sack.tt, D. L. (199S)The number needed
to treat: a clinicallyusefulmeasureoftreatment effect. British
MedicalJournal,310,452454.
Davis, D. A.,Thomson, M. A., Oxman, A. D., it al (1995)
Changingphysicianperformance.A systematicreview of the
effect ofcontinuing medicaleducationstrategies.Journalof the
AmericanMedicalAssociation,274,700705.
Eccies,M., Clapp, Z., Grimshaw, J., at al (l@6) North
of Englandevidencebasedguidelinesdevelopmentproject:
methodsofguideline development.BritishMedicalJournal.
312,760-762.
Eddy,D. 1.1.(1990) Clinicaldecisionmaking:from theory
to practice.Whatdo we do about costs?Journalof the
AmericanMedicalAssociation,264, I161,I 65, 1169
I170.
(19fl)

Clinical

decision

making:

from

theory

to

practice.

Cost-effectivenessanalysis.Willit be accepted?Journalof the

AmericanMedicalAssociation,268, 132136.

 (19fl)

Clinical

decision

making:

from

theory

to

practice.

Three battlesto watch in the l990s.Journalofthe American

MedicalAssociation,270, 520526.
(l@4)

Clinical

decision

making:

from

theory

to

practice.

Principlesfor makingdifficult decisionsin difficult times.Journol

ofthe AmericanMedicalAssociation,271,l792l798.
EUsr@M. & Smith, G. D. (1995) Misleadingmeta-analysis.
BritishMedicalJournal,310,752754.
Ellis, J., Mulligan, I., Rowe, J., .t aS(IflS) Inpatientgeneral
medicineisevidencebased.Lancet.346, 4074l0.
Evidence-Based M.dklne Working Group (1992)
Evidence-basedmedicine.A new approachto teachingthe
practice of medicine.Journalofthe AmericanMedical
Association,261 24202425.
Eysenck, Hi. (I@84)Meta-analysisand its problems.British
MedicalJournal,3O@789792.

Jackson,R.T. a Saclcstt, D. L. (1996) Guidelinesfor

managingraisedblood pressure.BritishMedicalJournal.
3l@6465.

Sack.tt, D. L, Roesnb.rg,W H., Gray, j. A., it ci (1996.)

Evidencebasedmedicine:what it is and what it isn't.British
MedicOIJOUrnOJ,
312,7172.

Kerr, M., Blizard, R. & Mann, A. (i9@S)General

practitioners and psychiatrists:comparisonof attitudes
to depressionusingthe depressionattitude questionnaire.
BritishJournalof GeneralPractice,45, 8992.

Desks,

Lau@s, A., Sack.tt, D. L & Roberts, R. S. (198$) An

assessmentofclinicallyusefulmeasuresofthe consequencesof
treatment. New EnglandJournalof Medicine,318,1728
l733.
Lewis, A. (1958) Betweenguessworkand uncertainty in
psychiatry.Lancet,i, l7l
l75.
MaiihaN, M., Gray, A., Lockwood, A., it .1(1996)
Casemanagementfor peoplewith severement@disorders
In Schizophrenia
Module(edsC. Adams,J.Anderson &

J.DcJesusMan).Oxford:UpdateSoftware.
FaMy,T.,Grlfflths, S. & P.tSrS,T.J. (l@@S)
Evidencebased
purchasing:understandingresultsofclinical trials and
systematicreviews.BritishMedicalJournal.311,l056I059.
G.dd.s, J. R., Christofi, G. 1 SaCk.tt, D. L (1996.)
Commentary on @first-rank
symptoms'.BritishJournalof
Psychiatry,169,544545.
G.dd.s, j. R., Game, D., Jenkins, N. E., .t .1 (19%b) What
proportion of primary psychiatricinterventionsare basedon

randomised
evidence?
QualityinHealthCare,5,2l52l7.
GUbod@5. (I@@6)Evidence-basedmedicine:an improved
format for journal clubs.PsychiatricBulletin,20,673675.
Giaazlou, P.P.& irwig, L M. (I9@5)An evidencebased
approachto individualisingtreatment. BritishMedicalJournal,
311.l356l359.
Graham. Smith, D. (1995) Evidencebasedmedicine:
Socraticdissent.BritishMedicalJournal.310,11261127.
Gr..nhalgh,T. (I@) Ismy practice evidence-based?Bntish
MedicalJournal,3l@957958.
H.rmanii, R. C., Dorwart, R. A., Hoover, C.W., it .1
(199S)Variationin ECTuseinthe UnitedStates.American
Journalof Psychiatry,152,869875.
Hirsch,
S.R.& Barn.s,T.
R.(1994)
Clinical
useofhigh-dose
neuroleptics.BritishJournalof Psychiatr@i
164,94-96.
Huston, R & P4oh.r, D. (l9@6)Redundancy,disaggregation.
and the integrity of medicalresearch.Lancet.347, 10241026.

patients:istherea consensus?
Schizophrenia
Bulletin,
20,
215225.
Muirow, C. D. (1987)The medicalreview article: state of the
science.Annalsof InternalMedicine,106,485488.
Rationale

for

systematic

reviews.

British

&

Altman,

D. G.

(1996.)

Down

with

odds

, Rkhardson,

S.,

Roesnb.rg,W,

it

.1

(1997)

Evidence

BasedMedicine:How to Practiseand TeachEBM.London:

Churchill-Livingstone.
Shahar,E. (1995) Evidence-basedmedicine.Lancet,346,

l772.
Shddon,T. A. (1996) Researchintelligencefor policy and
practice: the role ofthe National Health ServiceCentre for
Reviewsand Dissemination(EBt1 note).Evidence-Based
Medicine,I, 167
l68.
Shin,j. H., Haynes,R. B. & Johnston,N. E. (1993)Effect

of problem-based.
self-directed
undergraduate
educationon

Muse, U., Kurz, M. a FI.4schhacIm.,@WW.

(1994)
Antipsychoticmaintenancetreatment of schizophrenia

(1994)

J.J.

ratios! (EBM note).Evidence-Based

Medicine,I. 164166.

Medical

Journal.30@597599.
Naylor, C. D. (1995)Grey zonesofclinical practice: some
limits to evidence-basedmedicine.Lancet,345, 840842.
NHS Centre for R.vi.ws and Dissemination (1994)
ImplementingClinical PracticeGuidelines.EffectiveHealth
Care,8.
Oxman, A. D. & Guyatt,G. H. (198$)Guidelinesfor reading
literature reviews.CanadianMedicalAssociationJournal,131

697703.
Plppard, J. (1992)Audit of electroconvuslivetreatment in
two nationalhealthserviceregions.BritishJournalof
Psychiatry.160,621637.
Priest, R. G.,Vlzs, C., Roberts, A., it ii (1996)Laypeople's
attitudesto treatment of depression:resultsof opinion poll for
Defeat DepressionCampaignjust before its launch.British
MedicalJournal,3l@858859.

life-long learning.CanadianMedicalAssociationJournal,14$,

969976.
Sins,L & Hiatk@ H. A. (1996) Growing painsof meta
analyses.BritishMedicalJournal.3l@702703.
Simon, G.,Wagn.i@E. &
@nkorff,II. (1995)Cost
effectivenesscomparisonsusing real world' randomized
trials: the caseof new antidepressantdrugs.Journalof Qinical
Epiderniolog,t
4$ 363373.
Smith, R. (1991)Where isthe wisdom . . . ?BritishMedical
Journal.303,798799.
(1996)

What

clinical

information

do

doctors

need?

British

MedicalJournai,313 10621068.
Sullivan,F. @4,
& MacNaughton,R.J. (1996) Evidence
in consultations:interpreted and ind@iduaIked.Lancet,341
941943.
Summers, A. a Kaho., R. F. (1996) Ispsychiatrictreatment

evidence-based?
Lancet,347.409410.
1b.nsnbaum,S.j. (1993)What physiciansknow New England
JournalofMedicine, 329, 2681271.
@Is,
K. B., Katon,W, Rogers, B., it .1(1994) Useof minor
tranquilizersand antidepressantmedicationsby depressed
outpatients: resultsfrom the medicaloutcomesstudy.

American
Journal
of Psychiatry,
ISI,694700.

225

Closing the gap between research and practice.

J R Geddes and P J Harrison
BJP 1997, 171:220-225.
Access the most recent version at DOI: 10.1192/bjp.171.3.220

References
Reprints/
permissions
You can respond
to this article at
Downloaded
from

This article cites 0 articles, 0 of which you can access for free at:
http://bjp.rcpsych.org/content/171/3/220.citation#BIBL
To obtain reprints or permission to reproduce material from this paper, please write
to permissions@rcpsych.ac.uk
/letters/submit/bjprcpsych;171/3/220
http://bjp.rcpsych.org/ on March 24, 2016
Published by The Royal College of Psychiatrists

To subscribe to The British Journal of Psychiatry go to:

http://bjp.rcpsych.org/site/subscriptions/