PURPOSE: Research on obesity and the built environment has often featured logistic regression and the corresponding parameter, the odds ratio. Use of odds ratios for common outcomes such obesity may unnecessarily hinder the validity, interpretation, and communication of research findings. METHODS: We identified three key issues raised by the use of odds ratios, illustrating them with data on walkability and body mass index from a study of 13,102 New York City residents. RESULTS: First, dichotomization of continuous measures such as body mass index discards theoretically relevant information, reduces statistical power, and amplifies measurement error. Second, odds ratios are systematically higher (further from the null) than prevalence ratios; this inflation is trivial for rare outcomes, but substantial for common outcomes like obesity. Third, odds ratios can lead to incorrect conclusions during tests of interactions. The odds ratio in a particular subgroup might higher simply because the outcome is more common (and the odds ratio inflated) compared with other subgroups. CONCLUSION: Our recommendations are to take full advantage of continuous outcome data when feasible and to use prevalence ratios in place of odds ratios for common dichotomous outcomes. When odds ratios must be used, authors should document outcome prevalence across exposure groups.
PURPOSE: Research on obesity and the built environment has often featured logistic regression and the corresponding parameter, the odds ratio. Use of odds ratios for common outcomes such obesity may unnecessarily hinder the validity, interpretation, and communication of research findings. METHODS: We identified three key issues raised by the use of odds ratios, illustrating them with data on walkability and body mass index from a study of 13,102 New York City residents. RESULTS: First, dichotomization of continuous measures such as body mass index discards theoretically relevant information, reduces statistical power, and amplifies measurement error. Second, odds ratios are systematically higher (further from the null) than prevalence ratios; this inflation is trivial for rare outcomes, but substantial for common outcomes like obesity. Third, odds ratios can lead to incorrect conclusions during tests of interactions. The odds ratio in a particular subgroup might higher simply because the outcome is more common (and the odds ratio inflated) compared with other subgroups. CONCLUSION: Our recommendations are to take full advantage of continuous outcome data when feasible and to use prevalence ratios in place of odds ratios for common dichotomous outcomes. When odds ratios must be used, authors should document outcome prevalence across exposure groups.
Authors: Mia A Papas; Anthony J Alberg; Reid Ewing; Kathy J Helzlsouer; Tiffany L Gary; Ann C Klassen Journal: Epidemiol Rev Date: 2007-05-28 Impact factor: 6.222
Authors: Amy J Schulz; Srimathi Kannan; J Timothy Dvonch; Barbara A Israel; Alex Allen; Sherman A James; James S House; James Lepkowski Journal: Environ Health Perspect Date: 2005-12 Impact factor: 9.031
Authors: Tanya K Kaufman; Andrew Rundle; Kathryn M Neckerman; Daniel M Sheehan; Gina S Lovasi; Jana A Hirsch Journal: J Urban Health Date: 2019-08 Impact factor: 3.671
Authors: Andrew Rundle; Yun Wang; Sudha Sadasivan; Dhananjay A Chitale; Nilesh S Gupta; Deliang Tang; Benjamin A Rybicki Journal: Prostate Date: 2017-03-27 Impact factor: 4.104
Authors: Andrea E Cassidy-Bushrow; Suzanne Havstad; Niladri Basu; David R Ownby; Sung Kyun Park; Dennis R Ownby; Christine Cole Johnson; Ganesa Wegienka Journal: Biol Trace Elem Res Date: 2015-09-11 Impact factor: 3.738
Authors: Andrea E Cassidy-Bushrow; Ganesa Wegienka; Suzanne Havstad; Albert M Levin; Susan V Lynch; Dennis R Ownby; Andrew G Rundle; Kimberley J Woodcroft; Edward M Zoratti; Christine Cole Johnson Journal: Matern Child Health J Date: 2015-06
Authors: Gina S Lovasi; Ofira Schwartz-Soicher; James W Quinn; Diana K Berger; Kathryn M Neckerman; Risa Jaslow; Karen K Lee; Andrew Rundle Journal: Prev Med Date: 2013-05-31 Impact factor: 4.018