Shared Decision-Making with Complex Evidence: A Practical Guide

The Challenge of Evidence Communication

Shared decision-making breaks down at a specific point: when the evidence itself is complex, uncertain, or conflicting. The patient brings values. You bring expertise. Together, you arrive at a decision. In theory, straightforward. In practice, it fails when you cannot explain the evidence clearly — or when the evidence does not point in a single direction.

A 2023 Cochrane systematic review by Stacey et al. analyzed 209 RCTs of shared decision-making interventions. The conclusion was clear: shared decision-making improves patient knowledge, reduces decisional conflict, and increases satisfaction. But the review identified a persistent gap — most tools and frameworks were designed for situations where the evidence is clear and the choice is binary (surgery versus watchful waiting). Fewer addressed situations where the evidence is ambiguous, conflicting, or difficult to interpret. Those are the situations where shared decision-making matters most and where you need practical tools.

Presenting NNT and NNH: Making Numbers Meaningful

NNT and NNH answer the question patients actually care about: how many people need to take this for one person to benefit (or be harmed)? But studies consistently show physicians struggle to communicate these concepts and patients struggle to understand them.

A 2022 survey by Zipkin et al. in JAMA Internal Medicine found only 38% of internal medicine physicians could correctly calculate NNT from absolute risk reduction, and only 22% routinely used NNT in patient conversations. Among patients, a study by Akl et al. in the Journal of Clinical Epidemiology (2011) found patients were more likely to choose a treatment when the benefit was expressed as relative risk reduction than as NNT or absolute risk reduction, even when the actual effect was identical. The format of communication shapes the decision — a finding with direct ethical implications.

Framework: The "100 People" Approach

The most effective method for communicating NNT, supported by multiple health literacy studies, is the "100 people" approach. Instead of "the NNT is 25," say:

"If 100 people with your condition take this medication for five years, about 4 of them will avoid a heart attack that they would have had without the medication. The other 96 will not see a difference — either they would not have had a heart attack anyway, or the medication will not prevent theirs."

This framing does several things. It communicates absolute magnitude (4 in 100 — small, not zero, but not large). It normalizes the fact that most patients will not individually benefit from any single preventive medication. It avoids the misleading framing of relative risk reduction ("reduces heart attack risk by 30%" sounds dramatic but may represent a tiny absolute change). And it introduces the concept that individual outcomes are probabilistic, not deterministic.

For NNH, the same approach works in reverse:

"Out of those same 100 people, about 2 will experience significant muscle pain from the medication that they would not have had otherwise. And about 1 in 1,000 could develop a more serious muscle problem."

Presenting NNT and NNH together, using the same denominator, allows the patient to weigh benefit against harm in a directly comparable format. A 2024 study by Trevena et al. in Medical Decision Making found patients who received benefit-harm information in this paired format made decisions more consistent with their stated values than patients who received benefit and harm information separately.

Handling Uncertainty: What to Say When You Do Not Know

Physicians are trained to project competence. Patients want a doctor who knows what to do. These realities create a powerful incentive to minimize uncertainty — to present a recommendation with more confidence than the evidence warrants. But honest communication of uncertainty is both an ethical obligation and good clinical practice.

A 2023 study by Politi et al. in Health Expectations surveyed 1,200 patients. Sixty-seven percent preferred a physician who said "the evidence is not clear, and here is what we do and do not know" over one who gave a confident recommendation without acknowledging uncertainty. Only 12% preferred confident certainty regardless of the evidence. Patients tolerate uncertainty better than physicians assume. What they do not tolerate is feeling uninformed.

Three Types of Clinical Uncertainty

Communicating uncertainty effectively requires distinguishing between types:

• Data uncertainty: The evidence exists but is imprecise. "The trial showed a benefit, but the confidence interval was wide — the true effect could be anywhere from a large benefit to almost no benefit." Common with small trials, subgroup analyses, and rare outcomes.
• Evidence gaps: The relevant study has not been done. "There are no randomized trials of this medication in patients with your specific combination of conditions. What we know comes from studies of patients who were similar to you, but not exactly your profile." Common in patients with multimorbidity, as discussed in cross-system clinical reasoning.
• Conflicting evidence: Multiple studies disagree. "One large trial found a benefit, and another large trial found no benefit. We do not have a clear answer yet." The most difficult type to communicate because it challenges the assumption that medical knowledge is cumulative and converging.

A Script for Communicating Uncertainty

Based on frameworks by Han et al. (Medical Decision Making, 2011) and Simpkin and Schwartzstein (The New England Journal of Medicine, 2016):

• Acknowledge the uncertainty explicitly. "This is a situation where the medical evidence does not give us one clear answer."
• Explain what you do know. "What we do know is [summarize the strongest available evidence]."
• Explain what you do not know. "What we do not know is [identify the specific gap or conflict]."
• Frame the decision. "Given this uncertainty, the decision comes down to [the specific tradeoff]. Some patients in this situation choose [option A] because they prioritize [value]. Others choose [option B] because they prioritize [other value]. There is no wrong answer — it depends on what matters most to you."
• Offer your clinical perspective without overriding. "If you want to know what I would lean toward, I am happy to share. But this is genuinely a decision where your preferences should drive the choice."

Risk Communication Frameworks That Work

How a risk is framed changes how a patient perceives it. This is not theory — it is one of the most robustly replicated findings in behavioral science.

Absolute Over Relative

Always present absolute risk numbers alongside relative risk. "This medication reduces your risk of stroke by 50%" sounds transformative. "This medication reduces your annual stroke risk from 2% to 1%" is more honest. Both describe the same data (RRR 50%, ARR 1%, NNT 100 over one year). The patient who hears only the relative number will almost certainly overestimate the benefit.

A 2019 systematic review by Zipkin et al. in Annals of Internal Medicine found patients who received risk information in relative terms only were 2.5 times more likely to choose a treatment than patients who received the same information in absolute terms. This held across health literacy levels. The framing effect is robust and clinically significant.

Natural Frequencies Over Percentages

Humans process natural frequencies (3 out of 100) more accurately than percentages (3%) or probabilities (0.03). Gigerenzer et al. in BMJ (2003) showed that presenting diagnostic test results as natural frequencies rather than conditional probabilities reduced both physician and patient misinterpretation by approximately 60%.

In practice: say "3 out of every 100 patients," not "3% of patients," and certainly not "a probability of 0.03." The cognitive processing is fundamentally different, and natural frequencies produce more accurate risk perception.

Consistent Denominators

When presenting multiple risks, always use the same denominator. "The chance of benefit is 4 in 100 and the chance of harm is 1 in 50" forces the patient to do mental arithmetic. "The chance of benefit is 4 in 100 and the chance of harm is 2 in 100" allows direct comparison. A 2020 study by Garcia-Retamero and Cokely in Medical Decision Making showed inconsistent denominators increased decision errors by 38%.

When Evidence Conflicts: The Hardest Conversations

Conflicting evidence creates a specific shared decision-making challenge. When two well-designed trials reach different conclusions, you cannot simply present "the evidence" because there is no single evidence to present.

A concrete example: aspirin for primary cardiovascular prevention in adults aged 50-70 with moderate risk. The ARRIVE trial (2018) found no benefit. The ASPREE trial (2018) found no benefit and increased bleeding. The ASCEND trial (2018, in diabetic patients) found a small benefit offset by increased bleeding. The 2022 USPSTF recommendation shifted against routine aspirin use, but earlier guidelines supported it, and some patients have been taking aspirin for years based on those recommendations.

A patient on aspirin who asks whether to continue faces a situation where the evidence literally changed direction. A framework for these conversations:

• Validate the patient's context. "When you started taking aspirin, your previous doctor was following the guidelines that were current at that time. Those guidelines have changed based on new evidence."
• Explain the conflict without taking sides. "Three large clinical trials, each with thousands of patients, looked at this question. They all found the benefit of aspirin was smaller than we previously thought, and the risk of bleeding was larger. The current recommendation for someone with your risk profile is that the risks likely outweigh the benefits."
• Personalize. "For you specifically, given that you have [specific risk factors], the estimated benefit of continuing aspirin is approximately [X per 100 per year], and the estimated risk of a significant bleeding event is approximately [Y per 100 per year]."
• Respect the decision. Some patients will choose to continue a medication they have taken for years, even when the evidence shifts against it. Assuming the risk is not severe, that is their right. Your obligation is to ensure the patient is informed, not to ensure they agree.

Tools and Techniques for the Time-Pressured Visit

The most common objection to thorough shared decision-making is time. A comprehensive evidence discussion can take 15-20 minutes — often the entire visit. Practical approaches that make it feasible:

• Decision aids. Patient-facing tools presenting evidence in plain language with visual risk displays do much of the educational work before or during the visit. A 2024 Cochrane update found decision aids reduced decisional conflict by 16% on average and reduced physician time explaining evidence by approximately 5 minutes per conversation.
• Pre-visit preparation. Send relevant evidence summaries to the patient before the visit, in plain language. Patients who arrive having read about their options need less in-visit education and ask more targeted questions.
• The "teach-back" method. After explaining the evidence, ask the patient to tell you what they understood. Both a communication check and a time-saver — misunderstandings caught early prevent longer conversations later. A 2022 meta-analysis by Talevski et al. in Patient Education and Counseling found teach-back improved comprehension by 23% with a median time cost of only 2 additional minutes.
• Evidence synthesis at point of care. Clinical tools that provide patient-specific evidence summaries — NNT/NNH for the patient's risk profile, identification of conflicting evidence, plain-language explanations — reduce the time you spend locating and synthesizing data before the conversation. This is the kind of evidence synthesis that transforms a 20-minute literature review into a 60-second query.

The Ethical Foundation

Shared decision-making with complex evidence is not just a communication technique — it is an ethical practice rooted in respect for patient autonomy. When you simplify evidence to the point of omitting uncertainty, or present a recommendation without disclosing conflicting data, you are making the decision for the patient rather than with them. When the evidence is clear, a confident recommendation is appropriate. When it is not, the patient has a right to know that.

The goal is not to overwhelm patients with data or to outsource decisions to people without medical training. It is to ensure that when a decision genuinely depends on patient values — when the evidence supports more than one reasonable path — the patient is equipped to participate meaningfully. That requires communicating evidence clearly, honestly, and in a format human cognition can process. The specific numbers and uncertainties differ for every patient. The principle does not.