The Resourceful Patient

2.5 The doctor as adviser - communicating information without framing

'Frame: To fake the result of a race etc.' (1910)

Shorter Oxford English Dictionary

People are framed by circumstantial evidence but data can be framed as well, and the framing of data can influence decision-making, particularly when it is so easy, using modern statistical software, to change the way in which data is presented. Furthermore, it is extremely easy, using the Web, for pharmaceutical companies to present data to clinicians and patients in ways most likely to influence the decision to prescribe; of central importance in successful framing is the use of relative rather than absolute risk reduction data.

2.5.1 Distinguishing between absolute and relative risk reduction

People who particularly wish to influence others, as opposed to letting them make up their own minds, often choose to express the data as what is called a relative risk reduction.

For example, in the late eighties, following the publication of a meticulously organised randomised controlled trial of the effects of lipid-lowering drugs on mortality from coronary heart disease, an advertisement claimed that the drug 'reduces coronary heart disease risk by 34%.' At about the same time, another study carried out in Helsinki was published, and the two articles together led to what was called 'an unprecedented wave of enthusiasm for treating patients who have hypercholesterolaemia', raised levels of cholesterol in the bloodstream.

However, the results of the research were able to be presented in two ways, as shown in Table 3.

Table 3: Ways of presenting research for hypercholesterolaemia
	1	2	3	4
	% having heart attacks in the group which did not receive lipid- lowering drugs(the control group)	% having heart attacks in the group which did receive lipid-lowering drugs(the intervention group)	Absolute risk reduction (ARR) (2-3)	Relative risk reduction ARR x 100% _____ in the control group
Lipid Research Centres Trial	9.8%	8.1%	1.7%	19%
Helsinki Trial	4.1%	2.7%	1.4%	34%

Brett, A. S. 'Treating hypercholesterolemia' (5)

Column 1 shows the rate of heart attacks in the control group which did not receive the lipid lowering drugs, and column 2 the rate in the group which did receive the lipid lowering drugs.

Column 3 is the absolute reduction in risk which is the difference in the percentages of patients having heart attacks in the two groups.

The relative risk reduction (Column 4) takes the absolute risk reduction and converts it into a percentage of the percentage of patients having heart attacks in the group not receiving the lipid lowering drugs, called the control group by research workers.

The presentation of research results as either absolute or relative risk reduction has dramatic effects on decision-making by patients, clinicians and policy makers.

2.5.2 The impact of relative risk reduction on patient decision-making

One hundred Torontonian outpatients were asked whether or not they would take lipid lowering drugs. They were offered 'information' in two different ways - as either relative, or absolute, risk reduction. The difference in the responses of the two groups is striking, as shown in Table 4, and other studies have shown the same effect. (6,7)

Table 4: Effect of expressing risk in absolute and relative terms
Methods of expressing the reduction in risk	Percentage of patients choosing treatment
'34% Reduction in heart attacks'(Relative risk reduction)	88%
'1.4% reduction in heart attacks'(Absolute risk reduction)	42%

Hux, J.E. and Naylor, C.D. (8)

In one study, (9) 56.8% of patients chose medication when the benefit was expressed in relative terms, more than three times the 14.7% who chose the medication when the benefit was expressed in absolute terms.

2.5.3 Physicians and policy-makers can be misled by framing

2.5.3.1 Misleading physicians

Physicians too are influenced by framing and three studies (7,10,11) all demonstrated that physicians were more influenced by relative than by absolute risk reduction.

2.5.3.2 Misleading the experts

Few clinicians have received extensive training in epidemiology; their concern is for individuals rather than populations. However, even decision-makers whose job is to appraise decisions from the perspective of the population rather than the individual, are influenced by framing. A research team presented 182 health authority members with the results from randomised trials on breast cancer screening and a systematic review of trials of cardiac rehabilitation. They presented the results in four different ways, as shown in Table 5.

Table 5: Testing the framing effect - presentation of the same information about two programmes in four different ways
Information presentation	Mammography	Cardiac rehabilitation
Relative risk reduction	34%	20%
Absolute risk reduction	0.06%	3%
Percentage of event-free patients	99.82% vs 99.8%	84% vs 87%
Number needed to treat (NNT)	1592	31

Fahey et al, Evidence-based purchasing (12)

140 of the 182 decision-makers returned questionnaires and demonstrated clearly that their willingness to fund both programmes was influenced significantly by the way in which the results were presented; again they opted for relative risk reduction. Especially worryingly, perhaps, for the trained epidemiologists, in the group being studied, only 3 of the 140 'all non-executive members claiming no training in epidemiology' spotted that the four sets of data expressed the same results in different ways.

2.5.4 Giving information in words, numbers, images and hieroglyphs

Recognising that words can often be interpreted in different ways, particularly adverbs of probability such as 'probably' and 'possibly', numbers are increasingly used to influence decisions. It has long been recognised that basic numeracy skills vary in the population and those who are less numerate, not surprisingly, have more difficulty in appreciating options. (13) However, numeracy alone is not enough; clinicians, patients and all decision-makers in health and healthcare need to understand the impact of framing, particularly when clinicians are committed to shared decision-making.

2.5.5 Promoting the understanding of absolute risk reduction

Fortunately the concepts of absolute and relative risk reduction are easy to understand and the importance of presenting information to clinicians, patients and policy makers is now accepted and is being promoted.

The absolute benefits of treatments can be described in terms of the Number Needed to Treat (NNT) and examples of this are shown in Table 6.

Table 6: NNTs for various interventions (extracts from Bandolier 1995; 17:7, and McQuay and Moore)
Intervention	Outcome	NNT
CABG in left main stenosis	Prevent 1 death at 2 years	6
Carotid endarterectomy in high-grade symptomatic stenosis	Prevent 1 stroke or death in 2 years	9
Simple antihypertensive therapy for severe hypertension	Prevent 1 stroke, MI or death in 1 year	15
Simple antihypertensive therapy for mild hypertension	Prevent 1 stroke, MI or death in 1 year	700
Treating hypertension in the over-60s	Prevent 1 coronary event	18
Aspirin in severe unstable angina	Prevent MI or death in 1 year	25
Aspirin in healthy US physician	Prevent MI or death in 1 year	500
Graduated compression stockings for venous thromboembolism	Episodes of venous thrombo-embolism	9
Triple therapy for peptic ulcer	Eradication of H. pylori	1.1
Triple therapy for peptic ulcer	Ulcers remaining cured at 1 year	1.8
Permethrin for headlice	Cure	1.1
Antibiotics for dogbite	Infection	16

In the United Kingdom, increased professional awareness of the NNT is principally due to the excellent and highly readable journal Bandolier, and a visit to its website allows the reader to find many examples of the number needed to treat and its converse, the NNH - the number needed to harm.

The concept of absolute benefit is also being given prominence, for example when clinicians who feel that the public are being misled by newspaper headlines take the trouble to write to the press. An example of this was the excellent letter published in the Glasgow Herald which had headlined the relative risk reduction claims of a group of research workers:

Your report (August 16) on the trial published in this week's New England Journal of Medicine correctly states that in combination treatment with aspirin and clopidogrel in a selected group of high-risk patients, the relative reduction in risk of death, heart attack, and stroke was 20%. It is important to note that these events happened to 11.4% of patients receiving aspirin alone compared to 9.3% of patients receiving the two drugs together; an absolute reduction of 2.1%.

You quote one of the researchers saying that this treatment is simple and safe. Simple perhaps but examination of the trial shows that major bleeding occurred in 2.7% of those receiving aspirin alone and 3.7% of those receiving aspirin plus clopidogrel. This was an absolute increase in risk of 1% and a relative increase in risk of 38%.

Major bleeding was defined as substantially disabling, intraocular bleeding leading to loss of vision or bleeding necessitating the transfusion of at least two units of blood.

Please note the editorial piece on the website -
(http://content.nejm.org/this_week/345/7/index.shtml). 'Although clopidogril plus aspirin had clinical benefit beyond that of aspirin alone, the benefit was small and was partially offset by an increased risk of bleeding, including bleeding necessitating transfusion."

Researchers are recommending doctors to start prescribing this immediately and not to wait for the combination to become licensed.

But doctors are always legally responsible for the effects of drugs that they prescribe and if they prescribe a drug outwith its licence, their professional responsibility alters and increases. The Medical and Dental Defence Union of Scotland states: 'Although the Medicines Act allows doctors to recommend the use of drugs outside the parameters of their licences, it does not mean that the doctor thus advised has to prescribe any drug recommended. If they feel that the use of any recommended drug is beyond their experience or knowledge they are under no obligation to prescribe it.

Developments in the drug treatment of coronary events are welcomed, but the reporting of their benefits must be balanced with the clear safety issues.

Audrey Thompson, Medicines Management Adviser
Dr Andrew Power, Head of Medicines Management,
Gartnavel Royal Hospital, Glasgow.

Doctors sometimes consciously modify the information they give to patients with the best of intentions, and a study of the language used in the consultation demonstrated that doctors used phrases such as 'what they call' or adjectives such as 'little' to minimise the possible impact of terms such as fibro-adenoma and cervical erosion. The use of such terms is kindly meant but can sustain the imbalance of power between clinician and patient, maintaining the paternalistic power of the clinician and the childlike status of the patient, who responds by using phrases such as 'I was wondering' instead of asking direct questions.

2.5.6 The doctor as human being

The doctor here is portrayed as a logical calculating machine, relating to the patient as the watchmaker relates to the watch. The consultation, however, is different: it is an interaction between human beings, and the patient is influenced by the human elements of this interaction, partly because many patients bring more than one problem to the consultation - the sign or symptom that has led them to seek help as well as their anxiety about it.

Top