Male prisoners with (n = 132) and without (n = 132) histories of engaging in self-injurious behavior (SIB) were matched on conviction prefix and custody level. Conditional logistic regression revealed that a combination of risk factors from domains defined by developmental, offense history, mental health, and institutional functioning factors correctly classified 93% of the prisoners in the sample (ROC AUC = 0.89, S.E. = 0.005, p < 0.0001). Model specificity was 92.6% and sensitivity was 95.3%. False positive and false negative rates were 2.3% and 3.4%, respectively. Implications for the assessment of prisoners at risk for SIB as well as suggested future directions for SIB prisoner research are discussed.