Background: Optimal hepatitis B virus (HBV) screening strategies for patients with cancer awaiting chemotherapy are unknown. We aimed to identify predictors of HBV infection in a large cohort of patients with cancer who were systematically screened for HBV before chemotherapy. Methods: In this prospective observational study, patients with cancer at MD Anderson Cancer Center awaiting first administration of chemotherapy in an outpatient chemotherapy unit were identified and approached for HBV screening. Enrolled patients completed a CDC risk survey and had blood tests to screen for HBV (HBsAg, anti-HBc) and hepatitis C virus (anti-HCV) infection. We developed a logistic regression model to identify clinical predictors of either chronic (HBsAg+/anti-HBc+) or resolved (HBsAg-/anti-HBc+) HBV infection. Model fit was assessed using the Hosmer-Lemeshow goodness-of-fit test. Results: From July 2013 to December 2014, 2124 patients (mean [SD] age, 58 [13] years) completed the risk survey and had HBV screening. Of these, 54% were women, 11% Hispanic, 77% non-Hispanic white, 8% black, and 4% Asian. Nearly 20% had a hematologic malignancy, 1% liver cancer, and 79% had a solid tumor other than liver cancer. Almost 12% of patients were born outside of the US. Seven patients (0.3%) had chronic and 128 (6%) had resolved HBV infection. Among the latter, 13 patients were also anti-HCV+. Predictors (odds ratio [OR]; 95% CI) of having either HBsAg+ or anti-HBc+ result included black race (4; 3-7), Asian race (4; 2-7), birthplace outside the US (5; 3-9), and having lived with someone with HBV (2; 1-4). Older age and lower education level were also predictive of positive HBV test results. Patients who reported ever injecting drugs (OR 10; 5-22) or being a man currently having sex with another man (OR 28; 6-135) had a high risk of HBsAg+ or anti-HBc+ result. The area under the curve for the model predicting positive HBV tests was 0.80. Conclusions: Known HBV risk factors are predictive of chronic or resolved HBV infection in patients with cancer. Future model validation using bootstrapping will allow varying levels of sensitivity in order to develop best HBV screening strategies.