JAMA Insights 

How to Communicate Medical Numbers

Brian J. Zikmund-Fisher, Alistair Thorpe, Angela Fagerlin

JAMA 2025;334;(16):1474-1475

doi:10.1001/jama.2025.13655

Health care professionals often use numbers to guide decisions, monitor patient health, and communicate information. Although nonnumeric conversations may be recommended for some situations (eg, many cancer screening decisions),¹ providing numerical information, especially when requested by patients, may be helpful in shared decision-making.

However, the effectiveness of sharing numerical information depends on patients’ ability to understand and interpret these numbers. Understanding numbers may be difficult for patients with limited numeracy skills. Data from the 2023 Program for the International Assessment of Adult Competencies (a survey conducted by the US National Center for Education Statistics) reported that only 34% of 4637 US adults were able to perform simple numerical tasks (eg, identify the largest value in an unordered list).² Numeracy also overlaps with other foundational information-processing skills: 97% of adults at the lowest numeracy level had the lowest literacy, and 81% had not completed high school.²

Presenting numbers using evidence-based strategies may support patient understanding and engagement in care. In contrast, poor number communication may lead to patient misunderstanding, confusion, and/or frustration. Five evidence-based recommendations for communicating numbers to patients are presented below and in the Figure.

Figure.  Guidance for Communicating With Patients About Medical Numbers

HbA_1c indicates hemoglobin A_1c.

First, numbers (even if approximate) should be used to describe risk probabilities. The meaning of verbal probability terms (eg, rare, common, or unlikely) varies among people and situations. A systematic review of 24 studies found that people interpreted “rare” to mean probabilities ranging from 0% to 80%, while interpretations of “common” ranged from 10% to 100%.³ Using numbers helps patients understand whether a rare risk is 4% or 0.004%. In a large systematic review, verbal probability terms led to perceptions of likelihood as larger (7 of 9 studies; N = 2972) and stronger (8 of 10 studies; N = 3002) effects on behavioral intentions than numbers did.⁴Furthermore, 8 studies (N = 5079) showed that most people prefer to receive health risk information in numerical or number plus verbal formats vs verbal formats alone.⁴ One caveat to this recommendation is that nonnumeric communication may be appropriate when it is more important that a patient understands a category (eg, that their risk is “high”) than the exact likelihood.

Second, consistent risk denominators should be used when discussing numerical data with patients instead of a 1-in-X format in which the denominator varies while the numerator is fixed at 1 (eg, a 1 in 384 risk of trisomy18). A systematic review found that in 16 of 22 randomized clinical trials, risks presented as 1-in-X instead of consistent denominator formats (eg, percentages or rates per 1000) led people to perceive probabilities to be larger, which affected behavioral intentions in 2 of 3 studies.⁴ Furthermore, 1-in-X formats make it more difficult for people to determine which risk is higher or lower.⁵ A study of a multiethnic population of 633 US women from obstetrics and gynecology outpatient clinics found that 73% accurately identified 8.9 per 1000 as larger than 2.6 per 1000, whereas 56% were accurate when these risks were presented as 1 in 112 vs 1 in 384.⁶

Third, differences in probability generally should be presented using absolute differences instead of relative reduction or increases. Patients often want to know how likely is this intervention to actually help me, and is it more likely to help me or hurt me? How these probability differences are presented can influence people’s perceptions of these trade-offs. A recent systematic review reported that using relative difference formats (eg, 33% relative risk reduction) increased behavioral intentions (eg, choice of risk-reducing treatments) in 14 of 17 studies and perceptions of effectiveness in 2 of 2 studies compared with presenting the before and after absolute rates or the absolute probability difference (eg, 3 percentage point reduction from 9% to 6%).⁷

Clinical guidelines often present recommendations based on relative risk instead of absolute risk. Among 55 recommendations from 32 cancer guidelines, only 31% (n = 17) presented absolute effects for both benefits and harms, whereas 14.5% (n = 8) did not quantify benefits and harms and 54.5% (n = 30) presented them in an asymmetric manner (ie, benefits and harms presented using different formats), including 4 that quantified benefit in terms of a relative risk reduction and harms in terms of absolute risk increase.⁸ Describing the advantages of treatment using relative terms and the disadvantages in absolute ones creates a false equivalence that potentially overestimates benefits and underestimates risks. Therefore, for patients to understand the magnitude of benefits and risks associated with specific treatments, patients need to be presented with absolute risk differences, even if these data are approximate. For example, “this treatment reduces your risk from roughly 15% to maybe 7% to 8%” is preferable to stating that it “cuts the risk in half” or “results in a 50% risk reduction.”

Fourth, visual displays of probability should show both the numerator and the denominator of this ratio (Figure). Both icon array formats that use 100 icons in a matrix and stacked bar formats accurately represent this part-to-whole relationship, whereas bar charts do not, especially if scaled to less than 100%. A systematic review found that part-to-whole graphical formats led to smaller perceptions of both single probabilities and probability differences than graphic formats that only showed the risk numerator.^4,5 For example, a study of 1931 US adults showed that people rated both the likelihood of getting a hypothetical disease and the degree of protection provided by a vaccine as larger when relevant data were presented with numerator-only icon arrays vs displays that showed both the numerator and denominator.⁹

Fifth, contextual information should be provided for numbers that may be unfamiliar to patients (eg, biomarker levels). For example, in a study of 1618 US adults, providing a visual line display of creatinine levels with a threshold for concern (3.0 mg/dL labeled “many doctors are not concerned until here”) reduced concern for a result of 2.2 mg/dL vs an identical display lacking this threshold, yet the presence of the threshold label did not significantly change reactions to a more extreme value of 3.4 mg/dL.¹⁰ Beyond normal or standard ranges, discussing target values, action thresholds, and/or clinically significant difference information enables patients to understand their data and their clinical relevance.

Conclusions

When communicating health numbers to patients, clinicians can improve patient understanding by presenting information using numbers rather than providing verbal probability terms such as rare, common, or unlikely; using consistent denominators; discussing absolute rather than relative risks; and providing context for unfamiliar types of data.