Background: Newborn Screening for hearing impairment (NHS) is a crucial public health issue worldwide. Often, a two-stage screening with two different testing approaches is used. We aimed to investigate the optimal screening algorithm, based on data from the literature published in the last 30 years. A particular focus of the study was to synthesize the existing evidence on two-stage newborn hearing screening regarding the refer rate (RFR), the percentage of children that did not pass the second test or were lost after the first test. Methods: We searched MEDLINE for studies on two-stage NHS using transitory evoked otoacoustic emissions (TEOAE) or automated auditory brainstem response (AABR). All studies on newborns who received their first test as an inpatient and a second test up to one month later were eligible. Random effects meta-analysis and Bayesian modeling were performed to estimate RFR, effects of the second test phase on the RFR, and sensitivity and specificity of TEOAE and AABR, respectively. Risk of bias was assessed using QUADAS-II. The unfunded study was registered in PROSPERO (CRD42023403091). Results: Fifty-seven study protocols, including over 704,000 newborns, met the inclusion criteria. Certainty in the evidence was rated as moderate. The RFR was higher when the test method was changed than without a change of method (AABR-AABR: RFR = 1.4% (95% confidence interval (CI): 0.9, 2.2%), TEOAE-TEOAE: RFR = 2.5% (CI: 2.0, 3.1%), TEOAE-AABR: RFR = 4.9% (CI: 3.1, 7.5%), AABR-TEOAE: 5.9% (CI: 5.0, 6.9%). Across all protocols, both methods demonstrated high sensitivity (AABR: 98.9% (95% credibility interval (CrI): 96.0, 100.0%), TEOAE: 96.9% (CrI: 92.5, 99.8%)) and high specificity (AABR: 92.7% (CrI: 92.5, 92.9%), TEOAE: 91.3% (CrI: 91.2, 91.4%)). Conclusions: Strategies that did not involve changes to the screening method had lower RFR. Although both methods demonstrated high sensitivity, AABR appears to have slightly higher specificity compared to TEOAE.