Prevalence odds ratio versus prevalence ratio: choice comes with consequences

Abstract

Odds ratio (OR), risk ratio (RR), and prevalence ratio (PR) are some of the measures of association which are often reported in research studies quantifying the relationship between an independent variable and the outcome of interest. There has been much debate on the issue of which measure is appropriate to report depending on the study design. However, the literature on selecting a particular category of the outcome to be modeled and/or change in reference group for categorical independent variables and the effect on statistical significance, although known, is scantly discussed nor published with examples. In this article, we provide an example of a cross-sectional study wherein PR was chosen over (Prevalence) OR and demonstrate the analytic implications of the choice of category to be modeled and choice of reference level for independent variables.

Clustering of four major lifestyle risk factors among Korean adults with metabolic syndrome

“…Generally, prevalence ratio (PR) is regarded more appropriate measure than POR, but the reason we used POR instead of PR was that only POR has the property of reciprocity (changing reference category of a dichotomized variable will yield ‘reciprocal’ estimates) which is needed for our analyses [39]. In fact, previous researches with similar study design used POR, too [8–10,17].…”

Global state of deworming coverage and inequity in low-income and middle-income countries: a spatiotemporal study of household health surveys

“…The analysis used a multivariable logistic regression model using a dependent variable of mother-reported deworming receipt to compute the correlates of deworming. 21 The analysis included a pre-specified set of correlates based on prior literature as described above. 6 We included country fixed effects, which controlled for all time-invariant differences between countries (e.g.…”

“…We estimated the marginal effects for correlates of deworming receipt among children for ease of interpretation, and used robust standard errors clustered by country and survey. 21 We computed the ecological correlates of equity in deworming using all available survey data averaged at a regional level. We used a ordinary least squares regression model using a dependent variable of region-level equity index, and independent variables of region-level averages of the pre-specified set of correlates as described above.…”

Fibromyalgia Among Patients With Chronic Migraine and Chronic Tension‐Type Headache: A Multicenter Prospective Cross‐Sectional Study: A Comment

“…Finally, in Cho et al's table 2, “Widespread Pain Index of Participants With CM and CTTH,” they used chi‐square to compare rates of prevalence, but not all of the results complied with the assumptions of this test, specifically the variable “frequency pain in the right jaw.” For that reason, it would be highly recommended for the authors to use the Fisher test instead . Likewise, for multivariate analysis, the odds ratio was calculated; however, due to the high prevalence of the outcome (66.9% for chronic migraine), it would have been more suitable to use the prevalence ratio (PR) with Poisson regression, in order to avoid the strength of the association being overestimated …”

“…4 Likewise, for multivariate analysis, the odds ratio was calculated; however, due to the high prevalence of the outcome (66.9% for chronic migraine), it would have been more suitable to use the prevalence ratio (PR) with Poisson regression, in order to avoid the strength of the association being overestimated. 5…”