Study shows that medical residents lack necessary biostatistics knowledge

Points to problems with knowing research

Medical residents often lack the knowledge and skills they need to interpret medical research studies, which can jeopardize their ability to integrate new findings into their clinical practice, a recent study says.

But it also points to potential problems when these same doctors become researchers.

"We rely on medical literature to keep us up-to-date, and if residents can't understand the methods and results, it could potentially lead to incorrect interpretation and errors and application of what they read," says Donna M. Windish, MD, MPH, an assistant professor of medicine at Yale University School of Medicine in New Haven, CT, and at Waterbury Hospital in Waterbury, CT.

There are notable examples of how incorrect interpretations of studies have led to public controversy and changes in clinical practice, such as media reports that exaggerated the coronary risk associated with the diabetes drug rosiglitazone, Windish notes.

Through proper biostatistics and research training, physicians can make their own determinations about the validity of a study's findings, and not rely on headlines in the popular media.

Windish decided to research biostatistics knowledge among residents after noting some educational deficits in her own and in her students' learning experience.

"The study was initially generated through my own teaching at Yale, Johns Hopkins, and when I was a resident myself," Windish says.

"We were expected as physicians to read the most updated literature and interpret the results without having a strong knowledge of the methodology," Windish explains. "I personally found it challenging, and when I became a professor I found the medical students had the same fears and lack of understanding."

Windish conquered her trepidations about biostatistics by learning more as she earned a master's public health degree, but she says that most medical students encounter too few biostatistics courses during their medical training.

"It's very variable," Windish says. "If it's taught at all, it's taught during the first year or two of training, the pre-clinical years, and that training can last a couple of hours to several courses."

The study, published in the Journal of the American Medical Association, surveyed 277 internal medicine residents of 11 residency programs in Connecticut, and found that the overall average percentage correct on statistical knowledge and interpretation of results was 41.4%.1

"We limited it to internal medicine trainees, but their characteristics represented trainees from across the country," Windish says.

This compared with an average of 71.5% among fellows and general medicine faculty with research training.1

"We looked at 11 programs that allowed us to study their residents, survey them, and give them some training in statistics when we were done," Windish says.

"I heard the same anxieties about statistics from everyone," Windish notes. "If they were taught anything, it was so long ago and not placed in clinical context, so it wasn't meaningful to them."

Researchers created the survey instrument after reviewing the literature for three months, including JAMA, and summarized the statistics found in 239 articles, based on which ones were used most frequently, Windish explains.

They wrote questions that best represented the studies' findings.

"Some questions were borne out of biostatistics courses," she says.

For example, Johns Hopkins University gave the investigators permission to use their quizzes and tests, Windish adds.

Altogether the multiple-choice test contains 20 questions, and it's designed to be completed within 20 to 25 minutes, Windish says.

The full test is published as an appendix in JAMA, available to other researchers or academics who are interested in it.

The results show that some work needs to be done in improving new physicians' biostatistical knowledge, she adds.

"Residency programs should include biostatistics training to reinforce what was taught prior to residency, if it was taught at all," Windish says. "And they need to make it clinically relevant to their trainees."

Another reason why medical residents should have a firm foundation in biostatistics education is because they might one day want to become researchers.

"Part of the residency training that's mandated by the Accreditation Council for Graduate Education is that all residents have to do some type of scholarly work, and most people try to do research," Windish says. "We see a struggle among our own residents who want to do research, but have a fear they won't be able to analyze the results or even design the study."

Also, there is a national trend of increasing numbers of research studies moving to community settings, where physicians have very little, if any, experience in research. This is another reason why more statistical training is needed.

"Even with our current program, we've been working on developing not only more research methodology training through statistics' courses, but also to give them more formal education and some higher research methods in statistics," Windish says.

This will be a challenge for most educational institutions, she notes.

"It's challenging because even some of our faculty don't have research training," Windish says. "So how can they teach someone else when they don't have the background?"

Windish started a statistics course last year for trainees, and she used the same instrument employed in the study to see whether the training improved the students' knowledge and skills.

"It was only a four-hour course, and those who took the course one time scored 59% on the instrument," Windish says. "Those who took the course a second time, a year later, scored 70%."

These preliminary data support the idea that medical residents don't need large interventions or years of statistics, but would benefit from better information about biostatistics, she adds.

While researchers sometimes do need to consult with biostatisticians to help with study design and statistics, it still is important for medical researchers to demonstrate some competency in research methodology.

"The only other thing we need to consider is whether or not faculty are trained enough in our residency training program to be effective teachers of research and biostatistics," Windish says. "We're trying to work on our own residency training and build a course."

If these efforts succeed, then the information will be distributed to other institutions and training programs for help in developing their own training courses, she adds.


  1. Windish DM, et al. Medicine residents' understanding of the biostatistics and results in the medical literature. JAMA 2007;298:1010-1022.