Newswise — Lawrence C. Kleinman, MD, MPH, Vice Chair for Research and Education and Associate Professor of Health Policy and Pediatrics at Mount Sinai School of Medicine, has authored a paper about estimating risk measures and the flaws that are entailed. The paper, which includes Edward C. Norton, PhD, a health economist at the University of Michigan, as co-author, will be published in Health Services Research.

The paper titled, What's the Risk? A Simple Approach for Estimating Risk Measures from Nonlinear Models Including Logistic Regression, describes an improved approach to data analysis that will reduce the potential for miscommunication about how large an effect is, when researchers present their data to one another and to the lay public. The paper validates this approach and uses an example from the literature to demonstrate shortcomings of current methods.

Frequently, researchers want to describe the impact of one thing on another, independent of other factors. For example, they may want to measure the impact of second hand smoke on admissions to the hospital, independent of other factors, such as age, race, and insurance type and whether or not they live in a city. The most common way that researchers adjust for these factors is called logistic regression. Almost always, the results of these logistic regressions are presented as adjusted odds ratio.

"The word adjusted means that the other factors were taken into account," said Dr. Kleinman. "An odds ratio is the odds of being hospitalized if you were exposed, in this example, to second hand smoke, divided by the odds of being hospitalized if you were not exposed."

The problem is that people often misunderstand odds and think in terms of risk (or percent chance). They often end up interpreting odds ratios as if they were risk ratios. Unfortunately odds ratios are often much larger than the adjusted risk ratios.

When a paper reports a forty percent reduction in hospitalizations by avoiding second hand smoke, they are almost always reporting a reduction in odds, not risk. When you report with odds, the size of the effect will always be larger than if you report a risk ratio. If only one percent of people exposed to second hand smoke are hospitalized, the odds ratio and risk ratio will be pretty similar. But if 10 or 20 percent are hospitalized, the odds and risk ratio will be very different.

"It's as if you knew that someone had put their thumb on the scale, but you had no idea how hard they were pressing," Dr. Kleinman noted. "This method will enhance the capacity for researchers to communicate their findings with one another and with the lay public. It can help the press to avoid errors or misinterpretations that heretofore have been common, and at times serious. It can also help patients and their doctors to make more informed decisions."

About The Mount Sinai Medical CenterThe Mount Sinai Medical Center encompasses The Mount Sinai Hospital and Mount Sinai School of Medicine. The Mount Sinai Hospital is one of the nation's oldest, largest and most-respected voluntary hospitals. Founded in 1852, Mount Sinai today is a 1,171-bed tertiary-care teaching facility that is internationally acclaimed for excellence in clinical care. Last year, nearly 50,000 people were treated at Mount Sinai as inpatients, and there were nearly 450,000 outpatient visits to the Medical Center.

Mount Sinai School of Medicine is internationally recognized as a leader in groundbreaking clinical and basic-science research, as well as having an innovative approach to medical education. With a faculty of more than 3,400 in 38 clinical and basic science departments and centers, Mount Sinai ranks among the top 20 medical schools in receipt of National Institute of Health (NIH) grants.

  • share-facebook-Researchers Show the Flaws in Estimating Risk Measures
  • share-twitter-Researchers Show the Flaws in Estimating Risk Measures


Chat now!