OBJECTIVE: To examine whether typical pain behavior, as reported by caregivers, could be used prospectively to predict future pain behavior and to derive a subset of core items from the Non-Communicating Children’s Pain Checklist. STUDY DESIGN: Caregivers (n = 33) of children with cognitive impairments completed the Non-Communicating Children’s Pain Checklist retrospectively and immediately after subsequent episodes of pain and distress in their homes. Odds ratios were computed for checklist items, and multiple regressions were used to predict numerical pain and distress ratings with items that had significant odds ratios. A logistic regression was used to test whether the items found to predict pain could correctly classify the presence or absence of pain in a new cohort of 63 children with similar cognitive impairments. RESULTS: Seven of the checklist items had significant odds ratios: Cranky, Seeking Comfort, Change in Eyes, Less Active, Gesture to Part That Hurts, Tears, and Gasping. This subset of items significantly predicted numerical pain ratings by caregivers (multiple R =.70), but not distress ratings (multiple R =.31). In a second group of 63 children with cognitive impairments, this subset of items displayed 85% sensitivity and 89% specificity for pain. CONCLUSION: A subset of items from the Non-Communicating Children’s Pain Checklist could predict pain in children with cognitive impairments. Caregivers’ retrospective reports may be useful for clinicians making judgments about pain in these children.