Objective To assess transient-evoked otoacoustic emissions (TEOAE) data from 15 years of a newborn hearing screening program and evaluate how well various criteria separate ears with and without hearing loss. Design Retrospective review of TEOAE data using logistic regression, receiver operating characteristic curves, and cumulative percentage graphs. Study sample: Children with hearing loss who passed TEOAE screening as a newborn were compared to children who failed TEOAE screening and normal hearing children who either passed or failed. Exclusions were applied for acquired hearing loss or auditory neuropathy. Results Ears with hearing loss that passed screening had significantly lower TEOAE response levels compared to ears with normal hearing. Noise levels, test times, and number of sweeps were also lower. Most of these ears had mild hearing loss. Logistic regression results showed that high-frequency TEOAE response level is the best predictor of hearing loss. A multivariate “logit” score calculated from the regression was the best indicator for separating ears with hearing loss from ears with normal hearing. Conclusions TEOAE response levels or an algorithm which incorporates logit scores should be considered as a minimum passing criterion to increase the sensitivity of the TEOAE screening.