The Best Decision May Not Be What the Guidelines Say

John M Mandrola, MD


March 18, 2016

Good news: I have found something to like about the recent national coverage decision on percutaneous closure of the left atrial appendage.[1] The part I like is the requirement that an evidence-based decision tool be used in a formal shared decision-making interaction before the device is deployed.

The Institute of Medicine has called shared decision making one of  its six pillars of high-quality care.[2] But that's not the reason to like it. The reason to like shared decisions is it's the right thing to do—especially in cases where treatments are of uncertain benefit or have significant trade-offs between benefit and harms.

I think a lot about decision quality. Almost every decision in electrophysiology (EP) is preference sensitive. But EP isn't the only field dominated by preference sensitive decisions. A group of researchers from BMJ Clinical Evidence looked at 3000 treatments as reported in randomized controlled trials. They concluded that only 11% of treatments were clearly beneficial, 24% were likely to be beneficial, 7% were beneficial but had trade-offs between benefit and harms, and 50% were of unknown effectiveness.

That sort of uncertainty leaves a lot of room to align care with a person's preferences.

Another thing that bothers me about the state of decision quality is the power imbalance in the doctor-patient relationship. When a patient presents to a specialist with a problem, real or potential, that specialist has too much power to influence the decision.

A practice partner told me of an elderly lady he recently treated with intermittent atrial fibrillation. She was symptomatic and scared; her family, too, was scared. Fear brings vulnerability. But my partner is experienced and wise. He chose a conservative minimally disruptive approach. It wasn't easy, he said. It took a lengthy discussion. In the end, the patient and family were delighted. The point of his story was to set up this comment: "John, I could have taken this in a different direction. I could have ordered numerous tests, extended her hospital stay, and treated her more aggressively. It was my choice."

I realize that medicine will always be nuanced; no computer algorithm will ever replace a good doctor, and asymmetry cannot be vanquished from all medical decisions, but surely we can do better in communicating choices and reducing practice variation. This is, after all, the information age.

One way to increase the deliberation and consensus part of decisions is to use decision support. The best studied support tool is the decision aid. A decision aid has three goals. The first is to explicitly state the decision that needs to be made. This is no small step these days. Many people do not realize they have a choice. The second goal of a decision aid is to provide evidence-based information about the disease, its treatment options, benefits, harms, and uncertainties. The third goal is to help patients recognize the values-sensitive nature of their decision.

I recently stumbled upon a trove of evidence in support of decision aids. I've always favored decision aids, but I did not know the evidence was so strong.

Evidence for Decision Aids

Researchers from the Cochrane Group[3] performed a systematic review of decision aids for people facing health treatment or screening decisions. This review considered a total of 115 studies involving more than 34,400 patients in which decision aids were compared with standard care. The primary outcome measures were "decision-made" attributes and "decision-process" attributes. Here are some selected highlights:

  • In 42 studies of more than 10,000 patients, patients in the decision-aid group scored 13 times higher on knowledge scores.

  • In 19 studies of more than 5800 patients, patients in the decision-aid group were twofold more accurate in predicting risk.

  • In 22 studies of more than 4300 patients, patients in the decision-aid group scored significantly lower in measures of decision conflict—"feeling uniformed."

  • In 14 studies of more than 3200 patients, decision aids reduced practitioner-controlled decision making by 34%.

My favorite decision aid, one for primary-prevention implantable cardioverter defibrillator (ICD) choices, comes from the University of Colorado. It's called It's both informational and interactive and includes video vignettes from doctors and patients. I dare you to look at this decision aid and deny its utility in most patients considered for an ICD.

It's easy to understand why the Centers for Medicare & Medicaid Services (CMS) required a decision aid for left atrial appendage closure. At best, Watchman (Boston Scientific) is noninferior to warfarin, and all would agree that the device comes with trade-offs between benefit and risk. Clearly, then, Watchman is a preference-sensitive decision.

What's the Problem?

Why don't we use decision aids for many other decisions? Why isn't it normal to take data from figures 1 and 2 of major clinical trials, display them in a clear way, and then show them to patients? What are we afraid of? The data are the data. If a number-needed-to-treat is 20, what's wrong with saying that 19 of 20 patients get no benefit (but all the risk) of a treatment? You would want to know that if you were the patient, right?

When discussing the absolute risk reductions of anticoagulation in patients with atrial fibrillation, I often use a decision aid. Some people see the risk reduction of stroke as significant; others see it as small and not worth taking the drug. How is it that we know what is best for any given person?

I'm not naive. Many barriers slow the uptake of decision aids. Minor barriers include the likelihood that decision aids will slow decisions, change workflow, and are less useful in certain populations of patients—for example, the elderly with cognitive defects or non-English speakers.

Major barriers include the leveling of power in the doctor-patient relationship. Decision aids reduce the traditional paternalism, giving us less leeway in framing. Physicians in my community rarely use decision aids. I've even heard colleagues claim that decision aids are a way for government to ration care.

The largest barrier for acceptance of decision aids is illustrated in a randomized controlled trial from the United Kingdom.[4] Researchers compared a computerized decision aid against evidence-based paper guideline advice for anticoagulation in patients with atrial fibrillation. Patients shown the computer decision aid reported lower decisional conflict, but they also were less likely to start warfarin.

Therein lies the central problem for the new generation of doctors: when you improve decision quality, you may reduce uptake of a treatment proven effective in populations and recommended in guideline statements.

That need not bother us. And here is why: Shared decision making is a meeting of two experts. Doctors may be the expert in medical science, but patients are experts in what is best for them. The best decision, therefore, may not be what doctors and guideline writers decree it to be.

It's time the medical profession saw good decisions as a true quality measure. Good on the CMS for nudging us in the right direction.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.