Child maltreatment, and recurrent maltreatment in particular, occurs at an alarmingly high rate. Frequency of reports to Child Protective Services (CPS) is associated with negative psychological outcomes, and children whose reports are unsubstantiated experience similar risk of behavioral, emotional, and substance use disorders as those whose reports are substantiated. Prior research has demonstrated that children with no CPS reports and children with one CPS report showed no significant differences in rates of maltreatment perpetration or substance use in adulthood, suggesting that prevention efforts after one report may have strong merit in reducing negative outcomes in adulthood. However, patterns and risk factors of unsubstantiated reports have been only minimally explored thus far, despite having been found to predict subsequent maltreatment. The current study extends upon previous research by (a) examining both substantiated and unsubstantiated reports to identify longitudinal patterns of timing and recurrence and (b) assessing the extent to which service provision mediates long-term recurrence after each type of report. Analyses were conducted using subsamples of a longitudinal national dataset from 2011-2015 containing data from CPS reports for 3,655,951 children. Measures included child, caregiver, and CPS case characteristics obtained at the time of first report in 2011. Latent class analysis of referral patterns indicated four classes of recurrence patterns: (1) 2011 unsubstantiation followed by moderate recurrence, (2) 2011 unsubstantiation followed by low recurrence, (3) 2011 substantiation followed by moderate recurrence, and (4) 2011 substantiation followed by low recurrence. Multinomial logistic regression with most likely class membership as the outcome variable indicated that domestic violence, caregiver substance abuse, and poverty were better predictors of initial substantiation status than of long-term recurrence. Prior victimization was predictive of initial substantiation status as well as long-term recurrence. Asian American race predicted low rates of recurrence. Latent class analysis of service provision revealed only two classes: a class of children who received services and a class of children who did not. Service provision partially mediated associations between initial substantiation status and five-year maltreatment recurrence, as measured by number of subsequent reports, number of subsequent substantiated reports, and number of subsequent years in foster care. Limitations are considered and implications of using predictive modeling to drive service prioritization are discussed.