Introduction
Measures of association are crucial for epidemiologic studies, as they help to compare different groups of participants. There are four different types of odds ratio (OR), relative risk (RR), analysis of variation (ANOVA), and correlations. Different types of measures of association are used for studies of different designs. The present paper defines the measures of association, provides examples of studies that use these measures along with chi-square tests and t-tests, and describes the differences between OR and RR. The present paper concludes that while OR is not always intuitive, it may be used when the calculation of RR is impossible due to the study design.
Definitions
RR is one of the most intuitively comprehensible measures of association. According to the Centers for Disease Control and Prevention (CDC, 2012), RR compares the risk of a health event in one group to another. RR is calculated by dividing the risk of the event in one group by the risk of this event in another group. The result demonstrates how much more prevalent the condition is in one group as compared to the other. OR is similar to the RR; however, it is not always intuitively understandable. Szumilas (2010) defines OR as a ratio that is “the odds that an outcome will occur given a particular exposure, compared to the odds of the outcome occurring in the absence of that exposure” (p. 227). The measure is similar to RR; however, it is calculated by dividing of odds of the event in one group compared to the other.
Another two vital measures of association are ANOVA and correlation. ANOVA is a method that helps to compare the mean values of two or more independent groups (Miller, 1997). For instance, the method can help to estimate if alcohol consumption was different among age or ethnic groups. There are two types of ANOVA: one-way ANOVA and two-way ANOVA. One-way ANOVA compares two groups based on one variable, while two-way ANOVA compares several groups based on several independent variables. Correlation is a measure of the strength of a linear association between two variables. It is a standardized coefficient that varies between -1 and 1, where -1 is a perfect negative correlation, 1 is a perfect positive correlation, and 0 is a lack of correlation. The method is used to understand how much one variable affects the changes in the other variable. In summary, all measures of association help to make a comparison of different kinds depending on the research’s design and purpose.
Examples of Studies
RR is an intuitive measure of association that can be used for studies with large samples. For instance, a study that surveys ten thousand students to ask if they and their parents smoke can be conducted using RR. The researchers can divide the number of smokers by the total number of students in the group with non-smoking parents, then divide the number of smokers by the total number of students with smoking parents and then find an association between these risks. The result will demonstrate how much students with smoking parents are more likely to smoke than those with non-smoking parents. While RR is the most intuitive measure of association, it cannot always be used due to cost- and time efficiency. OR is commonly used in case-control studies (Szumilas, 2010). For instance, a researcher can use OR to estimate if using a two-handed technique is more effective for preventing tennis elbow than using a one-handed technique. Since it is economically unreasonable to conduct a large study for the condition, a researcher can recruit a hundred people with the condition and a hundred people without the condition and calculate OR. Since tennis elbow is a rare condition OR will be close to RR.
ANOVA and t-tests are used in similar research designs. For instance, a researcher needs to understand the difference in the average amount of alcohol patients consume in a group that underwent treatment and a control group. In this case, a t-test will be used, as the test helps to estimate the difference in means of the two groups. ANOVA will be used if the researcher decides to add a third group that underwent another treatment. For example, the first group may have received only a pharmaceutical treatment, the second group received both medications and therapy, and the third group received no intervention. Since there are three groups, one-way ANOVA will be used.
A correlation coefficient can be used to understand how much one numerical variable affects the other. For instance, a researcher may calculate how much job satisfaction is associated with workplace learning among healthcare professionals, as was demonstrated in the study by Iliopoulos et al. (2018). Chi-square analysis can be sued in studies with a pretest-posttest design. For instance, a researcher collects the frequency of occurrence of STIs after surgery before and after an intervention. Chi-square is used to understand if the proportion of patients with STIs after surgery is the same.
Difference between RR and OR
RR is a more intuitive method to use for measuring association. RR is calculated by calculating the risks in two groups. These risks are calculated by dividing the number of adverse cases by the number of measurements in one group, and then the same is done for the second group. Such action demonstrates a percent chance of an adverse event happening in any of the groups, which is essentially a risk of occurrence. Dividing one risk in one group by risk in another demonstrates how much more one group is exposed to the condition compared to another group.
OR is calculated by dividing the odds of an adverse event in one group by the odds of the adverse events in another group. Odds are calculated by dividing the total number of adverse cases by the total number of non-adverse cases. As mentioned above, OR can be very close to RR, if the adverse cases are uncommon. However, if OR is used for common conditions like smoking, the results may be misleading. According to Viera (2008), an outcome is considered uncommon if its prevalence is below 10%.
Conclusion
There are numerous measures of association, including OR, RR, ANOVA, t-test, correlation coefficient, and chi-square statistic. The present paper demonstrated different types of research appropriate for different types of analyses. The discussion also showed that RR and OR are used in similar circumstances. While RR is preferred, OR can still be used to decrease the cost and time of research. OR may be used only with the adverse condition is uncommon.
References
Centers for Disease Control and Prevention. (2012). Section 5: Measures of Association. Web.
Iliopoulos, E., Morrissey, N., Baryeh, K., & Polyzois, I. (2018). Correlation between workplace learning and job satisfaction of NHS healthcare professionals. British Journal of Healthcare Management, 24(5), 226-233.
Miller, R. G. (1997). Beyond ANOVA: Basics of applied statistics. CRC press.
Sedgwick, P. (2012). Pearson’s correlation coefficient. BMJ, 345-346.
Szumilas, M. (2010). Explaining odds ratios. Journal of the Canadian academy of child and adolescent psychiatry, 19(3), 227-229.
Viera, A. J. (2008). Odds ratios and risk ratios: what’s the difference and why does it matter? The southern medical journal, 101(7), 730-734.