Preventing Tomorrow’s High-Cost Claims: The Rising-Risk Patient Opportunity in Medicaid

ABSTRACT

This commentary notes the superiority of targeting rising-risk patients rather than high-cost claimants for Medicaid cost containment based on analysis of 13.1 million beneficiaries across 15 states. In 2019, spending for rising-risk patients (13.6% of sample) increased by 98.5% whereas spending for high-cost claimants (0.64%) decreased by 41.6%. Significantly, 54% of high-cost claimants in the first half of 2019 fell below the cost threshold in the second half of the year, and 50% of new high-cost claimants were previously identified as rising risk. Our findings reveal the limitations of focusing solely on high-cost claimants, whose costs naturally decrease due to regression to the mean. We argue that Medicaid programs should shift from reactive, cost-management interventions to proactive, prevention-oriented outreach, particularly as new predictive algorithms become more sensitive and specific. Early identification of and intervention for rising-risk patients is a more effective way to prevent the progression of chronic conditions and manage associated costs than attempting to reduce extreme utilization, which tends to decrease naturally over time.

Am J Manag Care. 2025;31(11):In Press

Takeaway Points

Medicaid programs should implement proactive, prevention-oriented approaches for the more than 80 million individuals they cover.

• We analyzed 13.1 million Medicaid beneficiaries across 15 states, identifying 1.78 million rising-risk patients (13.6%) and 83,386 high-cost claimants (0.64%).
• Half of new high-cost claimants were previously identified as rising-risk patients, and these former rising-risk patients accounted for 60.1% of total high-cost claimant spending.
• Early intervention trials in rising-risk patients have shown modest but meaningful reductions in avoidable hospital admissions and costs compared with a control group, supporting that the results were not simply due to regression to the mean.
• Our study builds on this evidence by demonstrating the population- and cost-level significance of rising-risk patients, highlighting their scale and impact across Medicaid, and informing how new predictive algorithms can guide proactive, system-level care management investments.

Medicaid, which provides health coverage for more than 80 million low-income individuals in the US,1 faces a critical inflection point due to ongoing challenges in improving health outcomes and managing escalating costs. Traditionally, Medicaid care management programs focused on telephonic outreach to patients who reach high-cost claimant status—ie, those whose total care costs have exceeded a certain threshold—to steer them toward more cost-effective care. Yet results of recent randomized trials2-4 cast doubt on the effectiveness of this standard approach, which critics have compared to addressing cancer only after it has metastasized.

In this commentary, we argue that Medicaid programs must take a more proactive approach, one that widens the aperture of interventions to prioritize rising-risk populations who have an increasing risk of emergency department (ED) visits or hospitalizations for ambulatory care–sensitive conditions that timely primary care could potentially prevent. We review evidence indicating that this approach could prevent adverse health outcomes for patients and their attendant costs for millions of Medicaid beneficiaries.

Differences Between High-Cost Claimants and Rising-Risk Patients

To understand the differences between rising-risk and high-cost claimants, consider 2 patients with type 2 diabetes. First consider Linda, aged 62 years, a high-cost claimant whose poorly controlled diabetes (with a hemoglobin A1c level of 11%) has led to multiple recent hospitalizations for hyperglycemic crises. Linda has been assigned case managers, has frequent specialist appointments, and has escalated to second- and third-line medications. Her health plan and providers are utilizing multiple services to manage her care, from nutritionists to mail-order pharmacy services to newer food-as-medicine programs. Yet findings from recent randomized controlled trials (RCTs) cast doubt on the impact of such programs for high-cost claimants like Linda. The most notable of these trials evaluated work by the Camden Coalition of Healthcare Providers to “wrap” intensive care management interventions around high-cost claimants to meet their complex medical and social needs and found no significant reduction in hospital readmissions for the treatment group compared with the control (ie, usual care) group.2 Similar RCTs that focused on care management services run by health plans to conduct outreach for high-cost claimants reported null effects on key outcomes such as ED visits, hospitalizations, and total cost of care.3,4

Although findings from earlier observational studies suggested that care management programs could be helpful, the null results of recent RCTs call into question these initial assessments. Why? The most salient reason is the fundamental statistical phenomenon known as regression to the mean. In the context of health care utilization and costs, high-cost claimants—who are often identified during periods of exceptionally high health care use—naturally tend to have lower utilization and costs in subsequent periods, even without intervention, by virtue of having started as extreme outliers. Thus, observational studies that compare pre- and postintervention periods often mistakenly attribute this natural pattern to the effectiveness of care management interventions.5 More rigorous RCTs address this discrepancy and provide a means to distinguish between the true causal effects of medical interventions and natural statistical tendencies that are observed in the control group and in the group that receives care management.

In contrast to Linda, consider the challenges facing another patient at an earlier stage of disease. Susan is a 58-year-old rising-risk patient recently diagnosed with diabetes (with a hemoglobin A1c level of 8.2%). She struggles with the challenges of taking insulin for the first time and not knowing how to properly check her blood sugar, has no transportation to outpatient appointments, and is facing financial difficulties due to worsening fatigue and its impact on her ability to work. Without intervention, she is on a trajectory toward potentially becoming a high-cost claimant.

Although results of RCTs focused on high-cost claimants showing null or minimal improvements2-4 seem discouraging, recent research on early intervention among rising-risk patients has shown more promising outcomes. A large, randomized study conducted in Contra Costa County, California, by Brown et al demonstrated the potential of early intervention for patients enrolled in Medicaid.6 Study participants like Susan received early insulin and glucometer education from pharmacists, transportation assistance from care coordinators, and help with employment and appointment adherence from a community health worker. Unlike the Camden Coalition study, which enrolled patients after they were already high cost, the Contra Costa study pursued early identification and engagement of patients at risk of health deterioration, producing meaningful reductions in avoidable hospital admissions and associated costs.6 Importantly, although the studies by Finkelstein et al and Rowe et al yielded null results,2,4 the Contra Costa study by Brown et al demonstrated positive impacts,6 as did a later real-world effectiveness and implementation study among a much larger cohort (Baum et al).7 The Brown and Baum studies used concurrent control groups to estimate intervention effectiveness with methodologies to address regression to the mean, and both Brown et al and Baum et al underscored the potential benefits of effective interventions.6,7 Most notably, Brown and colleagues observed beneficial effects despite only 40% of the intervention group engaging with the program.6

Early intervention trials among rising-risk patients underscore the value, for patients like Susan and their providers and health plans, of preventing the escalation of health care crises and associated costs, rather than focusing exclusively on efforts to reduce already extreme utilization.

Population-Level Differences Between High-Cost Claimants and Rising-Risk Patients

How often are patients like Linda and Susan found in practice? To answer this question, we evaluated Medicaid claims data from 13.1 million Medicaid beneficiaries spanning 15 states whose high-quality, individual-level health care utilization and social information were deposited into a national database.8 Our analysis compared health care utilization and costs for rising-risk and high-cost claimants.

We identified 2 cohorts of high-cost claimants—those incurring at least $25,000 in medical costs in the first half (January-June 2019) and those incurring at least $25,000 in medical costs in the second half (July-December 2019) of the year—to analyze cost persistence and transitions into or out of high-cost status over time. We also identified rising-risk patients using a peer-reviewed, published machine learning algorithm developed by Patel et al incorporating 2700 variables from claims and social determinants of health data9 to predict rising-risk patients at heightened risk for ambulatory care–sensitive ED visits or hospitalizations during the second half of the year (July-December 2019). Our analysis identified 1.78 million rising-risk patients (13.6% of the total sample) and 83,386 high-cost claimants (0.64%).

The Figure illustrates the stark contrast in spending trajectories across these groups. From January through June 2019, the 1.78-million rising-risk patients accounted for $494.8 million in spending per month, with this spending surging to $982.0 million per month in the latter half of the year—a 98.5% increase. In contrast, spending in 2019 for the 83,386 high-cost claimants dropped by 41.6%, from $394.9 million to $230.8 million monthly—illustrating regression toward the mean.

Of the 83,386 high-cost claimants identified in the first half of 2019, 54.0% no longer met the cost threshold in the second half of the year, illustrating the expected statistical phenomenon of regression to the mean. Moreover, in the second half of 2019, of the 73,827 patients who met criteria for being high-cost claimants, 36,877 (50.0%) were previously identified as rising risk. Nearly two-thirds (60.1%) of the total high-cost claimant spending came from these former rising-risk patients.

Our findings illustrate 2 key insights: (1) that most high-cost claimants regressed toward the mean and were no longer high-cost claimants in the second half of the year and (2) that half of new high-cost claimants and most of the spending associated with high-cost claimants come from individuals who could be proactively identified before reaching high-cost status. We found that these patterns persisted even after applying an enrollment criterion requiring at least 3 months of eligibility in both the first and second halves of the year, indicating that patient churn into and out of Medicaid did not explain the results. Together, these findings extend the existing literature by quantifying the scale and cost implications of rising-risk patients at the population level—moving beyond individual-level RCT results to demonstrate systemic impact.

We argue that Medicaid programs should shift from purely cross-sectional cost-management interventions to more proactive, prevention-oriented outreach, particularly in an era where predictive algorithms have become substantially more sensitive and specific.9 Although prior RCTs have shown limited effects for care management among high-cost patients and more promising results for rising-risk interventions, our findings uniquely demonstrate that rising-risk patients not only comprise a large share of future high-cost claimants but also account for the majority of high-cost spending. This positions rising-risk targeting as a scalable and effective strategy for Medicaid programs. Early identification and intervention among rising-risk patients offers a clear path to prevent worsening of chronic conditions and manage associated costs, rather than attempting to reduce already extreme utilization that is likely to decrease naturally over time.

Data Availability

The data sets utilized in this study are not publicly accessible. They can be obtained from CMS. Accessing these data entails a comprehensive procedure, involving completion of an institutional review board process and the procurement of a seat on the data portal. However, we did include county-level utilization and spending metrics used in our analysis. In accordance with CMS guidelines, we excluded counties with fewer than 12 visits. Researchers can find these data and the code necessary to replicate and extend our study findings on GitHub.10

Author Affiliations: Clinical Product Development, Waymark (SYP, SB), San Francisco, CA; School of Social Policy and Practice, University of Pennsylvania (SYP), Philadelphia, PA; Sloan School of Management, Massachusetts Institute of Technology (SYP), Cambridge, MA; Crown Family School of Social Work, Policy, and Practice, University of Chicago (HAP), Chicago, IL; University of California, San Francisco (SB), San Francisco, CA.

Source of Funding: None.

Author Disclosures: Dr Patel receives salary support from Waymark. Dr Pollack receives salary support from the University of Chicago and unrelated grants from the National Institutes of Health and Arnold Ventures. Dr Basu is an employee, board member, and stock owner of Waymark, which provides Medicaid patients with services; has received grants from the CDC and the National Institutes of Health related to proactive population health; has received an honorarium from the University of California, San Francisco, which provides Medicaid services; and has received patents for identifying patients for proactive care services. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (SYP, HAP, SB); acquisition of data (SYP, SB); analysis and interpretation of data (SYP, SB); drafting of the manuscript (SYP, HAP, SB); critical revision of the manuscript for important intellectual content (SYP, HAP, SB); statistical analysis (SYP, HAP, SB); administrative, technical, or logistic support (SYP, SB); and supervision (SB).

Address Correspondence to: Sadiq Y. Patel, PhD, MSW, Waymark, 2021 Fillmore St, Ste 1059, San Francisco, CA 94115. Email: sadiq.patel@waymarkcare.com.

REFERENCES

1. June 2025. Medicaid & CHIP enrollment data highlights. Medicaid.gov. Accessed October 2, 2025.
https://www.medicaid.gov/medicaid/program-information/medicaid-and-chip-enrollment-data/report-highlights

2. Finkelstein A, Zhou A, Taubman S, Doyle J. Health care hotspotting — a randomized, controlled trial. N Engl J Med. 2020;382(2):152-162. doi:10.1056/NEJMsa1906848

3. Kim SE, Michalopoulos C, Kwong RM, Warren A, Manno MS. Telephone care management’s effectiveness in coordinating care for Medicaid beneficiaries in managed care: a randomized controlled study. Health Serv Res. 2013;48(5):1730-1749. doi:10.1111/1475-6773.12060

4. Rowe JS, Gulla J, Vienneau M, et al. Intensive care management of a complex Medicaid population: a randomized evaluation. Am J Manag Care. 2022;28(9):430-435. doi:10.37765/ajmc.2022.89219

5. Morton V, Torgerson DJ. Effect of regression to the mean on decision making in health care. BMJ. 2003;326(7398):1083-1084. doi:10.1136/bmj.326.7398.1083

6. Brown DM, Hernandez EA, Levin S, et al. Effect of social needs case management on hospital use among adult Medicaid beneficiaries: a randomized study. Ann Intern Med. 2022;175(8):1109-1117. doi:10.7326/M22-0074

7. Baum A, Batniji R, Ratcliffe H, DeGosztonyi M, Basu S. Supporting rising-risk Medicaid patients through early intervention. NEJM Catal Innov Care Deliv. 2024;5(11). doi:10.1056/cat.24.0060

8. Transformed Medicaid Statistical Information System (T-MSIS) Analytic Files (TAF). Medicaid.gov. Accessed October 10, 2024. https://www.medicaid.gov/medicaid/data-systems/macbis/medicaid-chip-research-files/transformed-medicaid-statistical-information-system-t-msis-analytic-files-taf

9. Patel SY, Baum A, Basu S. Prediction of non emergent acute care utilization and cost among patients receiving Medicaid. Sci Rep. 2024;14(1):824. doi:10.1038/s41598-023-51114-z

10. Patel SY. ACS acute care Medicaid. GitHub. Accessed October 10, 2024. https://github.com/sadiqypatel/acs-acute-care-medicaid