An error occurred in the article by Kaur et al, titled “Epidemiology of Acute Otitis Media in the Postpneumococcal Conjugate Vaccine Era” published in the September 2017 issue of Pediatrics (2017;140(3): e20170181; doi:10.1542/peds.2017-0181). In the authors’ database, the BREASTFEEDING variable was represented as an ordinal variable, assuming values in the set {0,50 100}. The intention was to represent BREASTFEEDING as a dichotomous factor in a logistic regression model (combining 50 and 100 into a single factor level), in which case the odds ratio can be correctly estimated by exp(b), where b is the coefficient associated with BREASTFEEDING. However, BREASTFEEDING was mistakenly left as a numerical variable with range [0100]. The main effect on the fitted model was to approximately rescale the BREASTFEEDING coefficient, and its SE, by a factor of close to 1/100 (the mapping is not exact, due to the conversion of three distinct values {0,50 100} to two values {0,1}). The authors recalculated the model, using the originally intended dichotomous factor representation of BREASTFEEDING. BREASTFEEDING remains significant, and the odds ratio is now correctly reported, with a change in the order of magnitude compatible with the rescaling effect just described. In addition, the remaining covariate effects are changed only minimally. A corrected Table 8 appears below.
TABLE 8.
Multivariable Analysis of Factors Associated With sOP Condition
| Variable | Regression Coefficient | Odds Ratio | 95% CI | P |
|---|---|---|---|---|
| Constant | −3.17 | — | — | — |
| Male sex | 0.99 | 2.69 | 1.90–3.82 | 0.005 |
| Breastfed | −0.92 | 0.40 | 0.28–0.57 | 0.011 |
| Family history of AOM | 0.80 | 2.23 | 1.62–3.06 | 0.013 |
| Daycare attendance | 1.19 | 3.29 | 2.39–4.53 | 0.0002 |
—, reference.