Note from the editor: This article is the first in our new “Spotlight on statistics” series, which aims to clarify statistical practices used in research articles. The series will cite specific examples of how important terms are used in the literature, demonstrate a simple example of a statistical test using real data, and explain why understanding statistical techniques is part of the clinical nurse’s responsibility toward evidence-based practice.
Future articles will focus on odds ratios and correlations.
Increasingly, nurses are expected to participate in evidence-based practice (EBP) to help ensure use of the best available evidence to make clinical decisions. To do this, they must be able to read and evaluate the research literature. Specifically, they need to become familiar with basic statistical procedures and be able to evaluate the appropriateness of a statistical analysis and judge the author’s interpretation of results.
Every nurse has critical clinical questions that may lend themselves to statistical analysis. But like many nurses, you may view statistical analysis as difficult or at least tedious, and thus something to avoid. To better meet EBP expectations, you need to feel confident enough when reading a research article to understand the analysis and results sections.
In this article, we define relative risk and describe how this construct helps clinicians determine the relevance of an experimental intervention on an outcome or the relevance of an identified risk factor that increases an outcome’s likelihood. For instance, based on the evidence, nurse clinicians might decide to change their practice if an intervention reduces the risk of an undesirable outcome by 25%. Alternatively, they might better predict or understand a patient’s prognosis by identifying certain factors that predict the relative risk of the undesirable outcome occurring.
What is risk?
Some statistical analyses include such terms as risk factors, risk reduction, absolute risk, and relative risk. Risk factors are characteristics or experiences of patients that make them more likely to develop a disease (or not) or to get better (or worse) when exposed to the risk-increasing factor. Risk factor was coined in 1961 by Kannel and associates, who conducted the large observational study known as the Framingham study. In that study, 5,209 men and women were recruited to determine characteristics linked to cardiovascular disease. Certain conditions, including smoking, high blood pressure, and high cholesterol levels, were associated with (that is, correlated with increased risk of) heart disease development.
Related to risk is the opportunity for risk reduction after exposure to an intervention or a risk-reducing factor. Researchers commonly compare groups that receive potentially risk-reducing interventions with control groups not exposed to the intervention.
Absolute risk is the probability that the outcome or event will occur, whether or not the intervention group (or the control) was exposed to the risk-increasing factor. The absolute risk reduction (sometimes called the risk difference) compares the two risks; the absolute risk for the nonexposed group is subtracted from the absolute risk for the exposed group. This measure indicates the proportion of people who were helped by being exposed to the intervention or who were harmed by exposure to the risk-increasing factor. It may be displayed as a percentage (for example, 1%), a decimal (0.01), or a number (10 out of 1,000). When using a number, the researcher must indicate how many people were in the population (1,000 in this example).
Relative risk is the risk ratio—that is, the ratio of the proportion of individuals who had the undesirable outcome and were exposed to the risk-increasing factor, divided by the proportion of individuals who had the undesirable outcome but weren’t exposed to the risk-increasing factor. While relative risk is based on a specific sample, the estimate can be used to help practitioners decide the relevance of applying knowledge of the risk factor within their practice. Relative risk reduction estimates the effectiveness of the intervention and likelihood of the outcome occurring for those not exposed to the intervention.
For example, Nakagami and associates (2009) conducted a descriptive study of healing in patients with pressure ulcers based on wound temperature. Thirty-five subjects were chosen; of these, 23 had low wound temperatures while 12 had high wound temperatures. Of the 35 patients, 22 had pressure ulcers that healed normally. Among patients with high wound temperatures, the relative risk for not healing was 4.35. This can be interpreted as follows: Pressure ulcers with high temperatures are four times more likely to not heal than ulcers with lower temperatures.
This finding has tremendous implications for such nursing interventions as frequent patient repositioning, meticulous skin care, and use of egg-crate versus plastic mattresses. For readers who have access only to the article abstract from which these data were taken, knowing how to calculate risk would allow them to determine that three of the 22 healed ulcers were in the higher-temperature group while the other 19 were in the low-temperature group.
Calculating relative risk
We’ll use a descriptive example from our own research to explain how to calculate relative risk. We calculated the relative risk of not breastfeeding at 6 weeks postpartum for women who had previous breastfeeding experience, compared to women who lacked previous experience. Using the data shown in Example of relative risk calculation, the absolute risk (ARe) for breastfeeding cessation in women without breastfeeding experience (ARne, a risk-increasing factor) can be calculated as: ARe = a ÷ (a + b) = 14 ÷ 52 = 0.269. This means that 26.9% of those with no previous breastfeeding experience stopped breastfeeding by 6 weeks postpartum.
The absolute risk for women with breastfeeding experience who hadn’t been exposed to the risk-increasing factor was: ARne = c ÷ (c + d) = 2 ÷ 26 = 0.076. To put it another way, 7.6% of those with breastfeeding experience stopped breastfeeding by 6 weeks.
The absolute risk reduction (ARR) for this sample is calculated as: ARR = ARne – ARe = 0.269 – 0.076 = 0.193. In other words, women with no previous breastfeeding experience had a 19.3% higher likelihood of stopping breastfeeding by 6 weeks postpartum than those with previous breastfeeding experience.
In our sample, the relative risk (RR) is RR = ARe ÷ ARne = 0.269 ÷ 0.076 = 3.54. This means women with no breastfeeding experience were 3.54 times more likely to stop breastfeeding by 6 weeks. The relative risk reduction is: ARR ÷ ARne = 0.193 ÷ 0.076 = 2.54. The absolute risk for women without previous exposure to breastfeeding is reduced by 2.54 times the absolute risk for the nonexposed group (those who’d breastfed before).
The University of British Columbia’s Clinical Significance Calculator is one of several websites recommended as helpful in calculating risk indices based on known percentages. We found our results consistent with those on this website.
When you understand the concepts of absolute risk, risk reduction, and relative risk, you can better comprehend risks and changes in risks, and apply the relevant evidence to your practice. With this knowledge, you can critique research reports and understand the statistical underpinnings of more complex statistical analyses. Relative risk relates closely to the odds ratio—a frequently reported statistic that will be the focus of the next article in this series.
Aslam M, Burris HH, Zupancic JAF. Are we satisfied with the way we review articles? Adv Neonatal Care. 2009;9:40-42.
Kannel WB, Dawber TR, Kagan A, Stokes JI. Factors of risk in the development of coronary heart disease: six-year follow-up experience. The Framingham Study. Ann Intern Med. 1961;55(1):33–50.
Kollef MH, Afessa B, Anzueto A, Veremakis C, et al. Silver-coated endotracheal tubes and incidence of ventilator-associated pneumonia: the NASCENT randomized trial. JAMA. 2008;300(7):805-813. http://jama.ama-assn.org/cgi/content/abstract/300/7/805. Accessed November 23, 2010.
Nakagami G, Iizaka S, Kadono T, et al. Prediction of delayed wound healing in pressure ulcers by thermography. J Wound Ostomy Continence Nurs. 2009;36(suppl 3):S65. http://journals.lww.com/jwocnonline/Fulltext/2009/05001/Wound_Management_of_Complex_Wounds__3441_.196.aspx. Accessed November 23, 2010.
Newhouse R, Dearholt S, Poe S, Pugh LC, White K. Johns Hopkins Nursing Evidence-based Practice Model and Guidelines. Indianapolis, IN: Sigma Theta Tau International; 2007.
Polit DF. Statistics and Data Analysis for Nursing Research. 2nd ed. Boston, MA: Pearson; 2010.
Pugh LC, Serwint JR, Frick KD, et al. A randomized controlled community-based trial to improve breastfeeding among urban low-income mothers. Acad Pediatr. 2010;10(1):14-20.
Linda C. Pugh is a professor of nursing at York College of Pennsylvania in York. Renee A. Milligan is a professor of nursing at George Mason University School of Nursing in Fairfax, Virginia. Kevin D. Frick is a professor of health policy and management at Johns Hopkins Bloomberg School of Public Health in Baltimore, Maryland.