Shifting Towards Teaching How to Communicate Statistics in Medical Education

Abstract

Despite the widespread inclusion of statistics in medical school curricula as per the Liaison Committee on Medical Education requirements, the statistical competency among medical students and clinicians remains low. A 2007 study of 277 medical residents revealed only 41.1% scored correctly on a statistical knowledge survey, with minimal understanding of key concepts such as confidence intervals and adjusted odds ratios. A more recent 2023 study of 898 clinicians showed similar deficiencies in understanding efficacy, p-values, and discrimination metrics despite high confidence. This perspective argues for a paradigm shift from teaching statistical applications to focusing on statistical communication. We believe current statistics instruction lacks emphasis on communicating statistical results to patients. Teaching statistical concepts as tools for patient communication, rather than extensions of mathematics, can enhance understanding and ensure patients make informed decisions. Reframing statistical education to focus on communication could potentially address traditionally perceived learning barriers, improve understanding, and foster confidence. In this article, we outline several example reframings of teaching classical statistical concepts emphasizing interpretation and communication. Future strategies such as aligning statistics education closer to residency, revising exam content, updating accreditation requirements, and developing standardized communication primers can help ensure future clinicians are well-equipped to practice evidence-based medicine and effectively communicate statistical information in our increasingly data-driven world.

Accredited medical schools are required to develop curricula that “provides opportunities for medical students to acquire skills of critical judgment … in solving problems of health and disease.” ¹ Especially in our era of “evidence-based medicine,” this broad requirement takes the form of statistics instruction (144 out of 145 medical schools offer biostatistics according to the Liaison Committee on Medical Education [LCME] Annual Questionnaire, Part II survey 2016–17) seems like a step in the right direction. ² However, despite the inclusion of statistics in curricula at accredited medical schools, the overall statistical competency among medical students and clinicians remains low. In a 2007 multiprogram study of 277 medical residents, the overall percentage scored on the statistical knowledge survey was 41.1% (SD = 15.2%). ³ Only 11.9% could correctly interpret a confidence interval and 37.4% understood how to interpret an adjusted odds ratio from a multivariate regression analysis. ³ A recent 2023 study involving 898 clinicians found an overall poor proficiency in understanding efficacy, p-values, and discrimination metrics such as sensitivity and specificity despite high confidence. ⁴ These troubling numbers demonstrate a lack of competence in meeting part of LCME's requirements and raise the question of why physicians are still not adequately prepared to understand and interpret statistical results.

In this perspective piece, we argue that there needs to be a paradigm shift toward teaching how to communicate statistics instead of the current status quo of applying statistics. Statistics instruction, as with preclinical instruction in general, is done with the intent of preparing medical students for the USMLE STEP 1 exam, where “use and interpretation of statistical principles and measures of association” is tested. ⁵ While we do not have granular data on how each medical program structures its statistical component, we nonetheless believe there is a lack of explicit emphasis on communicating statistics in national medical student education policies. Statistical communication emphasizes a framework of understanding the implications of numerical evidence (such as positive predictive value, an odds ratio, or a p-value) and knowing how to describe it so that a patient can make an informed decision together with the physician. ⁶ In this way, key statistical concepts can and should still be taught, but additionally framed as primers for how to be explained to patients instead of presented only as abstract mathematical formulas or concepts that medical trainees have historically struggled with understanding. For instance, an absolute risk can be described more simply as the chances of an event happening. If the risk of stroke is 10%, then we would expect 10 out of 100 people to have a stroke.

The emphasis on statistical communication will also synergize with a proposal to shift the timing of statistics education to as close to residency as possible. The vast majority of statistics education takes place during the preclinical years. ¹ During the preclinical years, statistics instruction often fit into an already packed schedule of core medical topics as required by the LCME. The authors of the study on medical residents found that having prior biostatistics training in the preclinical years of medical school was not associated with higher statistical knowledge scores. ³ This may not be surprising given the premature timing, application-based orientation of statistics, and lack of context and reinforcement. We can imagine that teaching medical students how to communicate concepts such as risk and statistical significance once they’ve already experienced clinical rotations and are on the precipice of becoming licensed clinicians will be the most impactful. It is important here to distinguish between the role of communicating clinical significance, which is as much an art as science to holistically make patient-centered decisions that may extend beyond numbers; however, that discussion is presently out of scope for this perspective.

The focus on building statistical communication skills may also help guide decisions on which concepts to teach. Statistics is a complex and highly technical subject, so clearly there should not be an expectation to achieve competency within a short time frame and without repeated exposure. We believe that the emphasis on the application of statistics by entities such as USMLE is misplaced, as not all physicians need or want to be statisticians or quantitative researchers. For clinicians who do want research to form substantial roles in their careers, ample opportunities where they can more rigorously learn statistics and clinical research design exist, such as Master's programs, PhD programs, or clinical research fellowships. However, all clinicians should be able to understand the basic details of statistical concepts so that they can explain them in layperson's terms without necessarily knowing all the mathematical theory.

Reframing to statistical communication may circumvent barriers to learning, such as a perceived lack of confidence or confusing representations. Seventy-five percent of medical residents felt low confidence in their knowledge of statistics, which was reflected in their questionnaire scores. ³ This may be due to the perception among students and physicians that statistics is an inherently mathematical and thus unobtainable subject. ⁶ Thus, fostering transparent interpretations of statistical concepts that can be taught in a short and efficacious time should be the goal instead. We outline several such reframings here (Table 1); some studies show that presenting statistical concepts in absolute terms, for instance, probabilities as frequencies, helps physicians more accurately understand presented risk estimates. ⁶

Table 1.

Incorporating statistical communication in medical education.

TOPIC	CURRENT PRACTICE OF DOING	REVISED PRACTICE OF COMMUNICATING
Teaching relative risk and odds ratios	While medical professionals need to understand relative risks as this is the most common way associations are presented in biomedical literature, the level of education often stops with the definition of relative risk as the ratio of probabilities of events in the treatment group compared to the control group. There is little emphasis on interpreting the numerators and denominators. “Mammography screening reduces the risk of dying from breast cancer by 20%.”	We should teach students to be able to fluently convert relative risks to absolute risks. Absolute risk removes the mental math required to navigate percentages and quantifies risk in terms of counts. Example: “Mammography screening reduces the risk of dying from breast cancer by 1 in 1,000, from 5 in 1000 to 4 in 1000.” In addition, teaching concepts such as number-needed-to-treat with an emphasis on communication can help students and physicians more easily make the transition to explaining these concepts to patients. Example: “We need to screen 1000 people to prevent one person from dying from breast cancer.”
Teaching assessment of accuracy	Metrics such as sensitivity, specificity, positive predictive value, and negative predictive value are commonly taught so that medical professionals can read the literature and understand the performance of diagnostic tests. These metrics are based on understanding conditional probabilities. While this understanding is important, the level of education often stops with the calculations of these metrics (eg, Bayes’ Theorem) As an example of how sensitivity is portrayed: “If a woman has breast cancer, the probability that she tests positive is 90%.”	We should teach students to think in natural frequencies. This removes a lot of mental math and confusion and makes the process of understanding this concept more transparent. Example: “Of 10 women with breast cancer, 9 test positive.” Furthermore, given that sensitivity and specificity answer the question “What is the probability that a patient with the disease has a positive test result” (sensitivity), or “What is the probability that a patient without disease will have a negative test result” (specificity); whereas in practice, what clinicians and patients really are interested in is “What is the probability that a patient with a positive test result has the disease,” or “What is the probability that a patient with a negative test result does not have the disease,” emphasis should be placed on teaching how to communicate metrics such as positive predictive value and negative predictive value.
Teaching statistical inference	Many medical school curricula cover the concept of statistical significance, often teaching some form of “if the p-value is less than 0.05, it's statistically significant.”	We should teach students how to correctly interpret and communicate a p-value. Example: “It is the probability of what we observe or more extreme if there were no true effect of treatment and we assume that our study was free of all forms of bias.” We should also highlight common misconceptions such as p-values representing the extent that which chance produced the observed result, p-values <.05 meaning that results are scientifically important, or small p-values indicating a large effect size. Simultaneously, covering confidence intervals as another way of interpreting the results of a study is warranted. If the confidence interval is very wide, we are very uncertain about the effect estimate. Generally, well-equipped medical professionals should be astute questioners of test results. For example, if given a risk estimate, they should verify what the risk refers to, the time frame of the risk, how big the risk is in absolute terms, whether the risk applies to a patient, and how confident they can be in the risk.

While the inclusion of statistics in medical training in the past several decades has largely been a positive step in equipping our future generation of physicians in an increasingly data-driven world, the dangers of mislearning important concepts are too great in medical statistical education's current form. We believe that carefully reframing statistical education to communicate statistics should motivate incoming generations of physicians to appropriately practice evidence-based medicine. Some strategies include reevaluating the content of exams (eg, reconsidering how the scope of statistics is officially tested by USMLE to focus more on the explanation of statistical results instead of its calculation), including more explicit language in accreditation requirements so that medical students can achieve statistical competence, and disseminating standardized statistical communication education primers (eg, brief, high-yield documents teaching how to say the correct statistical interpretations). Some initiatives have been established abroad, such as the Winton Centre for Risk and Evidence Communication, which provides tools and leaflets that help promote statistical communication education. However, these interventions have not yet entered medical education in the United States. In our data-driven society, the time to successfully equip the next generation of clinicians with statistical communication tools is now.