Objective Many newborns are investigated and empirically treated for suspected early-onset sepsis (EOS). This study aimed to describe neonatologists' self-identified risk thresholds for investigating and treating EOS and assess the consistency of these thresholds with clinical decisions. Study Design Voluntary online survey, available in two randomized versions, sent to neonatologists from 20 centers of the Brazilian Network on Neonatal Research. The surveys included questions about thresholds for investigating and treating EOS and presented four clinical scenarios with varying calculated risks. In survey version A, only the scenarios were presented, and participants were asked if they would order a blood test or start antibiotics. Survey version B presented the same scenarios and the risk of sepsis. Clinical decisions were compared between survey versions using chi-square tests and agreement between thresholds and clinical decisions were investigated using Kappa coefficients. Results In total, 293 surveys were completed (145 survey version A and 148 survey version B). The median risk thresholds for blood test and antibiotic treatment were 1:100 and 1:25, respectively. In the high-risk scenario, there was no difference in the proportion choosing antibiotic therapy between the groups. In the moderate-risk scenarios, both tests and antibiotics were chosen more frequently when the calculated risks were included (survey version B). In the low-risk scenario, there was no difference between survey versions. There was poor agreement between the self-described thresholds and clinical decisions. Conclusion Neonatologists overestimate the risk of EOS and underestimate their risk thresholds. Knowledge of calculated risk may increase laboratory investigation and antibiotic use in infants at moderate risk for EOS. Key Points