A conflict of evidence and personal experience: trust the evidence
Being a “good doctor” involves an engagement with patients as individuals with their own needs, hopes and expectations. We advise and care for them to the best of our ability, based on clinical experience. At the same time we are under pressure to follow guidelines and implement protocols, based on evidence. We need to do both and usually they are not at odds - but if they are in conflict, which is dominant?
Analysis of Canadian data published in the BMJ (2006;332:141-3) revealed that doctors tend to be so influenced by experience that it overrides evidence. Sometimes I read a paper and cannot resist referring to it over and over again in the following weeks and months, so great is its impact on me. This was one such. It tells us that a recent bad experience has a powerful effect on practice, but the extent to which it biases subsequent treatment decisions away from best practice is salutary.
The clinical issues dealt with in this paper include:
- atrial fibrillation (AF)
- stroke as a consequence
- stroke prophylaxis with warfarin
- anticoagulant related bleeds
All headline stuff in cardiovascular medicine.
The BMJ paper caught my attention for a very personal reason. About ten years ago a screenwriter, working on a story involving a medical family, asked for my advice. Her story centred around a high achieving heart surgeon whose wife, a rural general practitioner, was to be portrayed as a more gentle person. The writer wanted a medically authentic incident to make the contrast. We had the lady doctor visit a favourite elderly patient. I suggested that she make the clinical diagnosis of AF and say something like “I know the clever chaps would send you to the anticoagulant clinic but … (sigh) … perhaps we’ll give you some digoxin and settle for that”. The contrast was made, the doctor eschewed intervention in favour of compassion and it worked well in the drama. With hindsight it was not such a good example. AF is responsible for 15% of strokes. Sparing the old man anticoagulation was not the better course of action in his overall interests although, to be fair to myself and to our fictional doctor, the definitive evidence on which practice is based had not then been published.
Whenever we use drugs to reduce the coagulability of the blood, there is a balance between the risk of the thrombus formation we are trying to prevent and the risk of bleeding that we induce with our therapy. Warfarin was first developed as a rat poison and is in part an acronym for the Wisconsin Alumni Research Foundation. It is the commonest oral antocoagulant in use and is also known by its trade name Coumadin.
In the case of lone AF, anticoagulation reduces the number of strokes and despite the risk of bleeding, there is a net benefit (Ann Intern Med 1999;131:492-501). Anticoagulation is not only clinically beneficial but highly cost effective – there is a reduction of need for health care. But doctors fail to anticoagulate a third to two thirds of patients with AF in spite of good evidence for its benefit. You might think that if a doctor had not prescribed anticoagulation for AF and the patient subsequently had a stroke, the doctor would tighten up on practice, but not so.
The authors used a large clinical database – the Canadian Institutes of Health Information (CIHI) records. From this they identified 116,200 patients with lone AF.
- There were 3,921 patients anticoagulated for AF and subsequently admitted with gastrointestinal or intracerebral haemorrhage.
- There were 8,720 patients not anticoagulated for AF and subsequently admitted with a stroke.
They then turned to another database held by the Ontario Health Insurance Plan and from the prescribing information they identified the physician most involved in the care of each patient. Now comes the clever bit. They are no longer interested in these index clinical events but in the physicians’ prescribing habits. It is a bit like the comedy routine for making hot whisky toddy which ends with throwing out the hot water and drinking the whiskey. Our attention is shifted to the physicians and their drug prescriptions for patients under their care in the 90 days before and after the time point determined by the index patient under their care who suffered the serious clinical event of bleeding or stroke. Did they change in treating subsequent patients?
Yes, if their patient had a bleed, but not in line with evidence. The 530 physicians who prescribed anticoagulation for AF were less likely to anticoagulate after a bleed - a reduction from 49% to 42%.
No, if their patient had a stroke. The 704 physicians who had not prescribed warfarin for a patient in AF were not influenced by the fact that their patient had a stroke – before and after warfarin use was similar at 37% and 36% and too low on the basis of the evidence.
You can make mistakes both ways round – errors of commission and errors of omission. In this instance it appears that if the doctors’ treatment resulted in a complication, some changed their practice, contrary to the research evidence, to reduce the chance of that happening again. It is as though the doctor was inappropriately taking the blame, wrongly seeing this as an error of commission. Certainly the INR should have been checked to confirm that anticoagulation was well controlled but a risk of bleeding is included in the calculation for net benefit of anticoagulation in AF.
On the other hand a stroke may be seen as nature taking its course. It did not impinge upon these doctors (based upon the pooled behaviour of 704 of them) as a danger that they could and should have averted. The stroke did not prompt more active management of AF in line with the evidence.
Recall our kindly fictional GP. Some people cannot understand how human beings can be ruled by numbers, while others see no logical way to make decisions without numerical evidence. In this instance, a personal policy founded on the shibboleth “in my clinical experience” is bound to do more harm and less good than following evidence based guidelines. Clinical events such as stroke due to AF and anticoagulant related bleeding are too sparse for us to derive any rule from our own experience, however experienced the individual doctor and indeed, however analytical. The same applies in the prophylaxis of deep vein thrombosis and pulmonary embolism. We cannot compute the competing risks for ourselves. For decision such as these, clinical trials, meta-analysis and databases in the thousands are much more likely to inform best practice than “in my experience”. That is what we should be teaching the next generation of doctors (Student BMJ 2006;14:162-3).
For my contemporaries who chose a life of surgery it must be galling. They used to refer to the less glamorous Public Health physicians as the “drains doctors” but now, with their massive databases and their sophisticated analytical methods, they make the rules. We need to find a balance between “clinical judgement” and the imperative to follow evidence based guidelines and protocols. If they are at odds, it is probably best to follow the rules. As I tell my juniors, there is plenty in their day to day work for which we have no secure evidence. That is where they can and must still exercise judgement.