The Medicare Care Choices Model was associated with reductions in disparities in the use of hospice care for Medicare beneficiaries with terminal illness.

OBJECTIVE: To assess the effects of the Medicare Care Choices Model (MCCM) on disparities in hospice use and quality of end-of-life care for Medicare beneficiaries from underserved groups-those from racial and ethnic minority groups, dually eligible for Medicare and Medicaid, or living in rural areas. DATA SOURCES AND STUDY SETTING: Medicare enrollment and claims data from 2013 to 2021 for terminally ill Medicare fee-for-service beneficiaries nationwide. STUDY DESIGN: Through MCCM, terminally ill enrolled Medicare beneficiaries received supportive and palliative care services from hospice providers concurrently with curative treatments. Using a matched comparison group, we estimated subgroup-specific effects on hospice use, days at home, and aggressive treatment and multiple emergency department visits in the last 30 days of life. DATA COLLECTION/EXTRACTION METHODS: The sample included decedent Medicare beneficiaries enrolled in MCCM and a matched comparison group from the same geographic areas who met model eligibility criteria at time of enrollment: having a diagnosis of cancer, congestive heart failure, chronic obstructive pulmonary disease, or HIV/AIDS; living in the community; not enrolled in the Medicare hospice benefit in the previous 30 days; and having at least one hospital stay and three office visits in the previous 12 months. PRINCIPAL FINDINGS: Eligible beneficiaries from underserved groups were underrepresented in MCCM. MCCM increased enrollees' hospice use and the number of days at home and reduced aggressive treatment among all subgroups analyzed. MCCM also reduced disparities in hospice use by race and ethnicity and dual eligibility by 4.1 (90% credible interval [CI]: 1.3-6.1) and 2.4 (90% CI: 0.6-4.4) percentage points, respectively. It also reduced disparities in having multiple emergency department visits for rural enrollees by 1.3 (90% CI: 0.1-2.7) percentage points. CONCLUSIONS: MCCM increased hospice use and quality of end-of-life care for model enrollees from underserved groups and reduced disparities in hospice use and having multiple emergency department visits.

Supportive Services at End of Life can Help Reduce Acute Care Services: Observations From the Medicare Care Choices Model.

OBJECTIVES: This study identifies the mechanisms through which supportive and palliative care services at the end-of-life helped prevent unnecessary use of acute care services. BACKGROUND: From 2016 to 2021, the Medicare Care Choices Model (MCCM) tested whether offering Medicare beneficiaries the option to receive supportive and palliative care services through hospice providers, concurrently with treatments for their terminal conditions, improved patients' quality of life and care and reduced Medicare expenditures. Previous MCCM evaluation results showed that the model achieved its goals, but did not examine in depth the causal mechanisms leading to these results. METHODS: Mixed-methods evaluation based on descriptive analysis of MCCM encounter data and qualitative analysis of interviews with staff from high-performing MCCM hospices. RESULTS: MCCM hospices provided 217 156 encounters to 7263 enrollees over 6 years. Enrollees received on average 30 encounters with hospice staff while enrolled in the model, representing about 10 encounters per month enrolled. Most encounters were delivered by clinically trained staff in the patient's home. Hospice staff identified five services critical for keeping patients from seeking acute care services: early and frequent needs assessments, direct observation of patients in their homes, immediate responses to patients' medical complaints, round-the-clock telephone access to nursing staff, and communication and coordination of care with primary care physicians and specialists. CONCLUSIONS: Palliative care approaches that are high-touch, employ clinically trained staff who visit patients in their homes, routinely evaluate how to manage patient symptoms, and are available when needs arise can improve outcomes and decrease costs at the end of life.

Medicare Care Choices Model Improved End-Of-Life Care, Lowered Medicare Expenditures, And Increased Hospice Use.

The Medicare Care Choices Model (MCCM) tested a new option for eligible Medicare beneficiaries to receive conventional treatment for terminal conditions along with supportive and palliative care from participating hospice providers. Using claims data, we estimated differences in average outcomes from enrollment to death between deceased MCCM enrollees and matched comparison beneficiaries who received usual services covered by original Medicare. Enrollees were 15 percentage points less likely to receive an aggressive life-prolonging treatment at the end of life and spent more than five more days at home. MCCM also reduced net Medicare expenditures by 13 percent, decreased inpatient admissions by 26 percent, reduced outpatient emergency department visits by 12 percent, and increased hospice use by 18 percentage points. Although the Centers for Medicare and Medicaid Services did not expand the model, given concerns about generalizability, these results provide evidence that MCCM is a promising approach to transforming care delivery at the end of life.

Effects of the Million Hearts Model on Myocardial Infarctions, Strokes, and Medicare Spending: A Randomized Clinical Trial.

Importance: The Million Hearts Model paid health care organizations to assess and reduce cardiovascular disease (CVD) risk. Model effects on long-term outcomes are unknown. Objective: To estimate model effects on first-time myocardial infarctions (MIs) and strokes and Medicare spending over a period up to 5 years. Design, Setting, and Participants: This pragmatic cluster-randomized trial ran from 2017 to 2021, with organizations assigned to a model intervention group or standard care control group. Randomized organizations included 516 US-based primary care and specialty practices, health centers, and hospital-based outpatient clinics participating voluntarily. Of these organizations, 342 entered patients into the study population, which included Medicare fee-for-service beneficiaries aged 40 to 79 years with no previous MI or stroke and with high or medium CVD risk (a 10-year predicted probability of MI or stroke [ie, CVD risk score] ≥15%) in 2017-2018. Intervention: Organizations agreed to perform guideline-concordant care, including routine CVD risk assessment and cardiovascular care management for high-risk patients. The Centers for Medicare & Medicaid Services paid organizations to calculate CVD risk scores for Medicare fee-for-service beneficiaries. CMS further rewarded organizations for reducing risk among high-risk beneficiaries (CVD risk score ≥30%). Main Outcomes and Measures: Outcomes included first-time CVD events (MIs, strokes, and transient ischemic attacks) identified in Medicare claims, combined first-time CVD events from claims and CVD deaths (coronary heart disease or cerebrovascular disease deaths) identified using the National Death Index, and Medicare Parts A and B spending for CVD events and overall. Outcomes were measured through 2021. Results: High- and medium-risk model intervention beneficiaries (n = 130 578) and standard care control beneficiaries (n = 88 286) were similar in age (median age, 72-73 y), sex (58%-59% men), race (7%-8% Black), and baseline CVD risk score (median, 24%). The probability of a first-time CVD event within 5 years was 0.3 percentage points lower for intervention beneficiaries than control beneficiaries (3.3% relative effect; adjusted hazard ratio [HR], 0.97 [90% CI, 0.93-1.00]; P = .09). The 5-year probability of combined first-time CVD events and CVD deaths was 0.4 percentage points lower in the intervention group (4.2% relative effect; HR, 0.96 [90% CI, 0.93-0.99]; P = .02). Medicare spending for CVD events was similar between the groups (effect estimate, -$1.83 per beneficiary per month [90% CI, -$3.97 to -$0.30]; P = .16), as was overall Medicare spending including model payments (effect estimate, $2.11 per beneficiary per month [90% CI, -$16.66 to $20.89]; P = .85). Conclusions and Relevance: The Million Hearts Model, which encouraged and paid for CVD risk assessment and reduction, reduced first-time MIs and strokes. Results support guidelines to use risk scores for CVD primary prevention. Trial Registration: ClinicalTrials.gov Identifier: NCT04047147.

Effect of the Million Hearts Cardiovascular Disease Risk Reduction Model on Initiating and Intensifying Medications: A Prespecified Secondary Analysis of a Randomized Clinical Trial.

Importance: The Million Hearts Cardiovascular Disease (CVD) Risk Reduction Model pays provider organizations for measuring and reducing Medicare patients' cardiovascular risk. Objective: To assess whether the model increases the initiation or intensification of antihypertensive medications or statins among patients with blood pressure or low-density lipoprotein (LDL) cholesterol levels above guideline thresholds for treatment intensification. Design, Setting, and Participants: This prespecified secondary analysis of a cluster-randomized, pragmatic trial included primary care and cardiology practices, health care centers, and hospital-based outpatient departments across the US. Participants included Medicare patients who were enrolled into the model in 2017 by participating organizations and who were at high risk and at medium risk of a myocardial infarction or stroke in 10 years. Patient outcomes were analyzed for 1 year postenrollment (through December 2018) using an intent-to-treat design. Analysis began November 2019. Interventions: US Centers for Medicare & Medicaid Services paid organizations for risk stratifying Medicare patients and reducing CVD risk among high-risk patients through discussing risk scores, developing individualized risk reduction plans, and following up with patients twice yearly. Main Outcomes and Measures: Initiating or intensifying statin or antihypertensive therapy within 1 year of enrollment, measured in Medicare Part D claims, and LDL cholesterol and systolic blood pressure levels approximately 1 year after enrollment, measured in usual care and reported to Centers for Medicare & Medicaid Services via a data registry (data complete for 51% of high-risk enrollees). The study's primary outcome (incidence of first-time myocardial infarction and stroke) is not reported because the trial is ongoing. Results: A total of 330 primary care and cardiology practices, health care centers, and hospital-based outpatient departments and 125 436 Medicare patients were included in this analysis. High-risk patients in the intervention group had a mean (SD) age of 74 (4.1), 15 213 (63%) were male, 21 657 (90%) were receiving antihypertensive medication at baseline, and 16 558 (69%) were receiving statins. Almost all (21 791 [91%]) high-risk intervention group patients had above-threshold systolic blood pressure level (>130 mm Hg), LDL cholesterol level (>70 mg/dL), or both. Patients in the intervention group with these risk factors were more likely than control patients (8127 [37.3%] vs 4753 [32.4%]; adjusted difference in percentage points, 4.8; 95% CI, 2.9-6.7; P < .001) to initiate or intensify statins or antihypertensive medication. Centers for Medicare & Medicaid Services did not pay for CVD risk reduction for medium-risk enrollees, but initiation or intensification rates for these enrollees were also higher in the intervention vs control groups (12 668 [27.9%] vs 7544 [24.8%]; adjusted difference in percentage points, 3.1; 95% CI, 1.9-4.3; P < .001). Among high-risk enrollees with clinical data approximately 1 year after enrollment, LDL cholesterol level was slightly lower in the intervention vs control groups (mean [SD], 89 [31.8] vs 91 [32.1] mg/dL; adjusted difference in percentage points, -1.8; 95% CI, -2.9 to -0.6; P = .002), as was systolic blood pressure (mean [SD], 133 [15.7] vs 135 [16.4] mm Hg; adjusted difference in percentage points, -1.7; 95% CI, -2.8 to -0.6; P = .003). Conclusions and Relevance: In this study, a pay-for-performance model led to modest increases in the use of CVD medications in a range of organizations, despite high medication use at baseline.

Estimating the incidence of unintended births and pregnancies at the sub-state level to inform program design.

OBJECTIVES: Unintended (mistimed or unwanted) pregnancies occur frequently in the United States and have negative effects. When designing prevention programs and intervention strategies for the provision of comprehensive birth control methods, it is necessary to identify (1) populations at high risk of unintended pregnancy, and (2) geographic areas with a concentration of need. METHODS: To estimate the proportion and incidence of unintended births and pregnancies for regions in Missouri, two machine-learning prediction models were developed using data from the National Survey of Family Growth and the Missouri Pregnancy Risk Assessment Monitoring System. Each model was applied to Missouri birth certificate data from 2014 to 2016 to estimate the number of unintended births and pregnancies across regions in Missouri. Population sizes from the American Community Survey were incorporated to estimate the incidence of unintended births and pregnancies. RESULTS: About 24,500 (34.0%) of the live births in Missouri each year were estimated to have resulted from unintended pregnancies: about 25 per 1,000 women (ages 15 to 45) annually. Further, 40,000 pregnancies (39.7%) were unintended each year: about 41 per 1,000 women annually. Unintended pregnancy was concentrated in Missouri's largest urban areas, and annual incidence varied substantially across regions. CONCLUSIONS: Our proposed methodology was feasible to implement. Random forest modeling identified factors in the data that best predicted unintended birth and pregnancy and outperformed other approaches. Maternal age, marital status, health insurance status, parity, and month that prenatal care began predict unintended pregnancy among women with a recent live birth. Using this approach to estimate the rates of unintended births and pregnancies across regions within Missouri revealed substantial within-state variation in the proportion and incidence of unintended pregnancy. States and other agencies could use this study's results or methods to better target interventions to reduce unintended pregnancy or address other public health needs.

Parent Partners: Evaluation of a Novel Peer-Support Intervention for the Caregivers of Children Hospitalized for Behavioral Health Conditions.

OBJECTIVE: Nearly 10% of all hospitalized children have a primary behavioral health diagnosis, but the effectiveness of treatment can be limited by caregivers' challenges navigating the behavioral health system. In this study, we assessed a novel peer-support intervention ("parent partners") designed for the caregivers of children admitted to an inpatient psychiatric unit. METHODS: We used a mixed-methods approach including 1) document review and interviews to assess implementation and 2) a difference-in-differences analysis using claims for Medicaid-enrolled children admitted to the intervention inpatient psychiatric unit and matched comparison children admitted to other inpatient psychiatric units to assess the impacts on health care use after discharge. RESULTS: Ninety-six percent of caregivers who were offered the intervention engaged with a parent partner. The primary challenges to implementation were accommodating the needs of parent partners to address behavioral health crises among their own children and initial limited engagement from behavioral health clinicians. The intervention leaders reported success in addressing these through adjustments to staffing policies, training parent partners in engagement with clinicians, and incorporating parent partners into team rounds. We did not find a statistically significant difference in follow-up outpatient behavioral health visits (adjusted treatment to comparison difference +3% [90% CI = -2%, +9%]), readmissions (+5% [-33%, +43%]), or behavioral health ED visits (-15% [-44%, +14%]). CONCLUSIONS: This novel intervention was implemented successfully, and although our study did not find statistically significant impacts on health care use after discharge, the findings for ED visits are suggestive of benefits. Parent peer support in inpatient settings warrants additional investigation.

Effects of a Behavioral Health and Chronic Illness Care Intervention on Patient Outcomes in Primary Care Practices in the Dakotas.

From 2012 to 2015, Sanford Health, a large health care system, integrated behavioral health services and chronic condition care management in some of its primary care practices in the Dakotas and rural Minnesota. Using difference-in-differences analyses for fee-for-service Medicare beneficiaries attributed to 22 participating practices and 91 matched comparison practices, we found that the program increased the receipt of four recommended diabetes care processes by 8.6% (p=.048) and, by slowing the increase in emergency department (ED) visits, reduced them by 4.9% (p=.07) relative to the comparison group. However, the findings are mixed: the program did not affect hospital admissions, readmissions, or Medicare spending. In addition, the program increased admissions for ambulatory care-sensitive conditions by 13.6% (p=.07) relative to the comparison group. Sanford's program provides a concrete example of how to incorporate behavioral health services in primary care in underserved areas with some positive results on quality-of-care processes and ED utilization.

Estimating the hospital costs of inpatient harms.

OBJECTIVE: To estimate the additional hospital costs associated with inpatient medical harms occurring during an index inpatient admission and costs from subsequent readmissions within 90 days. DATA SOURCE: 2009 to 2011 Healthcare Cost and Utilization Project's State Inpatient Databases from 12 states. STUDY DESIGN: We compare hospital costs incurred by patients experiencing a specific harm during their hospital stay to the costs incurred by similar patients who did not experience that harm. DATA EXTRACTION: We extracted records for adult patients admitted for a reason other than rehabilitation or mental health, were at risk of a harm, and were admitted for less than a year. PRINCIPAL FINDINGS: The costliest inpatient harms, such as surgical site infections and severe pressure ulcers, are associated with approximately $30 000 in additional index stay costs per harm. Less costly harms, such as catheter- or hospital-associated urinary tract infections and venous thromboembolism, can add $6000 to $13 000. Birth and obstetric traumas add as little as $100. CONCLUSIONS: Our analysis represents rigorous estimates of the hospital costs of a variety of inpatient harms; these should be of interest to health care administrators and policy makers to identify areas for cost savings to the health care system.

Effects of a community-based care management model for super-utilizers.

OBJECTIVES: Medicare, Medicaid, and commercial plans have all explored ways to improve outcomes for patients with high costs and complex medical and social needs. The purpose of this study was to test the effectiveness of a high-intensity care management program that the Rutgers University Center for State Health Policy (CSHP) implemented as an adaptation of a promising model developed by the Camden Coalition of Healthcare Providers. STUDY DESIGN: We estimated the impact of the program on 6 utilization and spending outcomes for a subgroup of beneficiaries enrolled in Medicare fee-for-service (n = 149) and a matched comparison group (n = 1130). METHODS: We used Medicare claims for all analyses. We used propensity score matching to construct a comparison group of beneficiaries with baseline characteristics similar to those of program participants. We employed regression models to test the relationship between program enrollment and outcomes over a 12-month period while controlling for baseline characteristics. RESULTS: A test of joint significance across all outcomes showed that the CSHP program reduced service use and spending in aggregate (P = .012), although estimates for most of the individual measures were not statistically significant. Participants had 37% fewer unplanned readmissions (P = .086) than did comparison beneficiaries. Although we did not find statistically significant results for the other 5 outcomes, the CIs for these outcomes spanned substantively large effects. CONCLUSIONS: Although these findings are mixed, they suggest that adaptations of the Camden model hold promise for reducing short-term service use and spending for Medicare super-utilizers.

Rural hospital transitional care program reduces Medicare spending.

OBJECTIVES: To evaluate impacts of a telephonic transitional care program on service use and spending for Medicare fee-for-service beneficiaries at a rural hospital. STUDY DESIGN: Observational cohort study. METHODS: Patients discharged from Atlantic General Hospital (AGH) with an AGH primary care provider were assigned a nurse care coordinator for 30 days. The nurse reviewed the patient's conditions, assessed needs for transition support, conducted weekly telephone calls (beginning 24-72 hours after discharge) to monitor adherence to treatment plans, and scheduled follow-up appointments. Using claims data, we evaluated impacts on service use and spending using a difference-in-differences design with a matched comparison group. RESULTS: The intervention reduced Medicare spending in the 6-month period after discharge by 30.8%, or $1333 per beneficiary per month (90% CI, -$2078 to -$589), which was partly driven by a 39.4% reduction in spending for inpatient claims (difference, -$729; 90% CI, -$1234 to -$225). There were no statistically significant changes in the 14-day ambulatory care follow-up rate, 30-day unplanned readmission rate, number of inpatient admissions, or number of emergency department visits, although this may be due to modest statistical power to detect effects. CONCLUSIONS: The estimated $5.4 million in savings from this intervention more than offset the costs of the $1.1 million funding for the award. Although other studies have found that care transitions programs can improve outcomes, this study was unique in the size of the impacts relative to the low-touch intervention and the location in a small rural healthcare system.

The Impact of a Health Information Technology-Focused Patient-centered Medical Neighborhood Program Among Medicare Beneficiaries in Primary Care Practices: The Effect on Patient Outcomes and Spending.

BACKGROUND: The Center for Medicare & Medicaid Innovation (CMMI) tests new models of paying for or delivering health care services and expands models that improve health outcomes while lowering medical spending. CMMI gave TransforMED, a national learning and dissemination contractor, a 3-year Health Care Innovation Award (HCIA) to integrate health information technology systems into physician practices. This paper estimates impacts of TransforMED's HCIA-funded program on patient outcomes and Medicare parts A and B spending. RESEARCH DESIGN: We compared outcomes for Medicare fee-for-service (FFS) beneficiaries served by 87 treatment practices to outcomes for Medicare FFS beneficiaries served by 286 matched comparison practices, adjusting for differences in outcomes between the 2 groups during a 1-year baseline period. We estimated impacts in 3 evaluation outcome domains: quality-of-care processes, service use, and spending. RESULTS: We estimated the program led to a 7.1% reduction in inpatient admissions and a 5.7% decrease in the outpatient emergency department visits. However, there was no evidence of statistically significant effects in outcomes in either the quality-of-care processes or spending domains. CONCLUSIONS: These results indicate that TransforMED's program reduced service use for Medicare FFS beneficiaries, but also show that the program did not have statistically significant favorable impacts in the quality-of-care processes or spending domains. These results suggest that providing practices with population health management and cost-reporting software-along with technical assistance for how to use them-can complement practices' own patient-centered medical home transformation efforts and add meaningfully to their impacts on service use.

The Efficacy of Using Financial Incentives to Change Unhealthy Behaviors Among a Rural Chronically Ill and Uninsured Population.

PURPOSE: To measure the effects of weight loss, medication compliance, and physical activity financial incentives delivered by health providers to uninsured patients. DESIGN: Full factorial (orthogonal) randomized design. SETTING: Primary care clinic in rural Mississippi. PARTICIPANTS: A total of 544 uninsured adult patients with diabetes or hypertension. All patients were enrolled in the clinic's care management program. INTERVENTION: Patients randomly received 0, 1, 2, or 3 financial incentives. Incentives for weight loss, medication compliance, and physical activity were awarded quarterly over 1 year. MEASURES: Weight loss, medication compliance, physical activity, and 6 clinical measures related to diabetes and hypertension. ANALYSIS: Cross-section and panel data regression models were used to compare outcomes for those who received incentives to those who did not receive incentives and to measure interaction effects. RESULTS: Effects of the weight loss incentive were positive but statistically insignificant (-3.7 lb; P = .106), while medication compliance was high in both the treatment and control groups (+2.37 percentage points; P = .411), and physical activity take-up was very limited. Effects on clinical outcomes, as well as interaction effects between incentives, were mixed and generally statistically insignificant. CONCLUSION: This study found little to no evidence that these financial incentives had beneficial effects on the incentivized behaviors in this setting. Likewise, the study found no effects on clinical outcomes nor any systematic evidence of interaction effects between 2 and 3 incentives.

Association Between Extending CareFirst's Medical Home Program to Medicare Patients and Quality of Care, Utilization, and Spending.

Importance: CareFirst, the largest commercial insurer in the mid-Atlantic Region of the United States, runs a medical home program focusing on financial incentives for primary care practices and care coordination for high-risk patients. From 2013 to 2015, CareFirst extended the program to Medicare fee-for-service (FFS) beneficiaries in participating practices. If the model extension improved quality while reducing spending, the Centers for Medicare and Medicaid Services could expand the program to Medicare beneficiaries broadly. Objective: To test whether extending CareFirst's program to Medicare FFS patients improves care processes and reduces hospitalizations, emergency department visits, and spending. Design, Setting, and Participants: This difference-in-differences analysis compared outcomes for roughly 35 000 Medicare FFS patients attributed to 52 intervention practices (grouped by CareFirst into 14 "medical panels") to outcomes for 69 000 Medicare patients attributed to 42 matched comparison panels during a 1-year baseline period and 2.5-year intervention at Maryland primary care practices. Main Outcomes and Measures: Hospitalizations (all-cause and ambulatory-care sensitive), emergency department visits, Medicare Part A and B spending, and 3 quality-of-care process measures: ambulatory care within 14 days of a hospital stay, cholesterol testing for those with ischemic vascular disease, and a composite measure for those with diabetes. Interventions: CareFirst hired nurses who worked with patients' usual primary care practitioners to coordinate care for 3656 high-risk Medicare patients. CareFirst paid panels rewards for meeting cost and quality targets for their Medicare patients and advised panels on how to meet these targets based on analyses of claims data. Results: On average, each of the 14 intervention panels had 9.3 primary care practitioners and was attributed 2202 Medicare FFS patients in the baseline period. The panels' attributed Medicare patients were, on average, 73.8 years old, 59.2% female, and 85.1% white. The extension of CareFirst's program to Medicare patients was not statistically associated with improvements in any outcomes, either for the full Medicare population or for a high-risk subgroup in which impacts were expected to be largest. For the full population, the difference-in-differences estimates were 1.4 hospitalizations per 1000 patients per quarter (P = .54; 90% CI, -2.1 to 5.0), -2.5 outpatient ED visits per 1000 patients per quarter (P = .26; 90% CI, -6.2 to 1.1), and -$1 per patient per month in Medicare Part A and B spending (P = .98; 90% CI, -$40 to $39). For hospitalizations and Medicare spending, the 90% CIs did not span CareFirst's expected impacts. Hospitalizations for the intervention group declined by 10% from baseline year to the final 18 months of the intervention, but this was matched by similar declines in the comparison group. Conclusion and Relevance: The extension of CareFirst's program to Medicare did not measurably improve quality-of-care processes or reduce service use or spending for Medicare patients. Further program refinement and testing would be needed to support scaling the program more broadly to Medicare patients.

Effects of Medicaid disease management programs on medical expenditures: Evidence from a natural experiment in Georgia.

In recent decades, most states' Medicaid programs have introduced disease management programs for chronically ill beneficiaries. Interventions assist beneficiaries and their health care providers to appropriately manage chronic health condition(s) according to established clinical guidelines. Cost containment has been a key justification for the creation of these programs despite mixed evidence they actually save money. This study evaluates the effects of a disease management program in Georgia by exploiting a natural experiment that delayed the introduction of high-intensity services for several thousand beneficiaries. Expenditures for medical claims decreased an average of $89 per person per month for the high- and moderate-risk groups, but those savings were not large enough to offset the total costs of the program. Impacts varied by the intensity of interventions, over time, and across disease groups. Heterogeneous treatment effect analysis indicates that decreases in medical expenditures were largest at the most expensive tail of the distribution.