People with complex and/or severe health issues represent a small—though growing—proportion of the population yet use the majority of health resources. This concentration of utilization is seen across health care settings and payers. Addressing the health of these high-needs populations has been the topic of substantial research and policy,1 yet the ability to mitigate the poor health outcomes of these populations is mixed. In part, this group of high-needs patients is not uniform and likely requires different approaches to tailoring care. Segmenting the high-needs population is challenging, especially as most models involve embedded team members—either at transitions2 or in primary care3 or in the community4—and these individuals, no matter how highly trained, may have limited ability to address the breadth and depth of needs of the population segments. Research has shown, however, that outcomes may be influenced through providing a certain breadth of services.5 In the paper by Grant et al., care managers at Kaiser Permanente receiving referrals of high-needs patients were asked to retrospectively rate the appropriateness of the referral; the elicited factors that determined appropriateness were then used to try to predict patients who may be better candidates for the care managers in these programs. As our ability to implement predictive models grows, these refinements to both the identification of high-needs patients and the approach to care they may need are extremely important to avoid increased burn-out and keep the care patient-centered. With large health reform models using these risk stratification and segmentation approaches, the need for scalable models that use the input of those charged with tailoring care will only grow over time. Next steps may include proactive implementation of prediction for both the needs of the patients and their appropriateness for certain services or approaches.