Background: Generally, a SLNB is offered to the patient whose likelihood of positivity is ≥ 5%. Using population-based cohorts, we sought to [1] develop and validate a predictive model to estimate the individualized risk (IR) of SLNB positivity (SLNBp) and its confidence interval (CI), [2] evaluate an IR-based CDR to identify patients with sufficiently low IRs to avoid SLNB, and [3] compare the performance of the CDR with that of the 2019 NCCN guideline. Methods: The learning cohort (n = 18,214; SLNBp rate = 10.7%; 2010-2014) and validation cohort (n = 3,924; SLNBp rate = 11%; 2015) included SEER patients 18-99 years old who had a SLNB as part of definitive surgery. A multivariable logistic regression model for SLNBp, including 4 AJCC related-factors (thickness, ulceration, level and mitotic rate) and age, was used to estimate the IR of SLNBp and its one-sided 95% CI. The CDR was defined using the IR and CI and then used to classify patients into 3 categories: SLNB not indicated (IR and upper limit of the one-sided CI < 5%), SLNB indicated (IR and lower limit of the one-sided CI ≥ 5%), and borderline. Results: In the learning cohort all 5 factors were significant in the multivariate model, which had a c-statistic of 0.742 (95%CI, 0.731-0.753). In the validation cohort the model c-statistic was 0.728. Based on the CDR, 21.6%, 56.9%, and 21.5% of patients would not be offered SLNB, would be offered SLNB, or would need further shared decision making. Compared to the guideline, the CDR classified more patients as “SLNB not indicated” (21.6% vs. 7.1%) and fewer as “SLNB indicated” (56.9% vs. 68.5%). Of the 16,137 SLNB negative patients, 3815 (23.6%) would not be offered SLNB based on the CDR compared to 1258 (7.8%) who would not be offered SLNB based on the guideline. The false negative rates associated with “not indicated” by the CDR and guideline were 4.6% and 1.2%, respectively. Conclusions: Use of this CDR rather than the NCCN guideline will spare more low-risk patients the expense and harms of SLNB and increase only marginally their likelihood of a false negative test (still < 5%).