Intensive Care Management of a Complex Medicaid Population: A Randomized Evaluation

ABSTRACT

Objectives: Care management programs are employed by providers and payers to support high-risk patients and affect cost and utilization, with varied implementation. This study sought to evaluate the impact of an intensive care management program on utilization and cost among those with highest cost (top 5%) and highest utilization in a Medicaid accountable care organization (ACO) population.

Study Design: Randomized controlled quality improvement trial of intensive care management, provided by a nonprofit care management vendor, for Medicaid ACO patients at 2 academic centers.

Methods: Patients were identified using claims, chart review, and primary care validation, then randomly assigned 2:1 to intervention and control groups. Among 131 patients included in intent-to-treat analysis, 87 and 44 were randomly assigned to the intervention and control groups, respectively. Patients in the intervention group were eligible to receive intensive care management in the community/home setting and, in some cases, home-based primary care. Patients in the control group received standard of care, including practice-based care management. Prespecified primary outcome measures included total medical expense (TME), emergency department (ED) visits, and inpatient utilization.

Results: Relative to controls, patients randomly assigned to receive intensive care management had a $1933 smaller increase per member per month in TME (P = .04) and directionally consistent but nonsignificant reductions in ED visits (17% fewer; P = .40) and inpatient admissions (34% fewer; P = .29) in the 12 months post randomization compared with the 12 months prerandomization.

Conclusions: Our study results support that targeted, intensive care management can favorably affect TME in a health system–based high-cost, high-risk Medicaid population. Further research is needed to evaluate the impact on additional clinical outcomes.

Am J Manag Care. 2022;28(9):430-435. https://doi.org/10.37765/ajmc.2022.89219

Takeaway Points

This randomized controlled quality improvement trial of high-risk, high-cost patients in an academic health system’s Medicaid accountable care organization (ACO) showed that intensive care management was associated with a statistical reduction in total medical expenditure and a non–statistically significant reduction in utilization in the 12 months post randomization relative to the 12 months prerandomization.

• Patients randomly assigned to intensive care management had a $1933 smaller increase per member per month in total medical expense compared with controls (P = .04).
• A non–statistically significant reduction in utilization was observed: 17% fewer emergency department visits (P = .40) and 34% fewer inpatient admissions (P = .29).
• The results of this study suggest that intensive care management can affect spending and utilization among patients in a health system–based Medicaid ACO.

As health systems, payers, and providers adopt value-based care, care management programs are foundational to reducing unnecessary utilization and improving clinical outcomes¹ by longitudinally managing patients identified based on clinical complexity and/or utilization.^2-4 There have been varied approaches to the design, implementation, and evaluation of care management programs. Many studies are limited by lack of an optimal control population, leading to potential confounding due to regression to the mean among patients with high utilization.^5-8 The few published randomized evaluations also vary in terms of implementation and evaluation.^9-11

Mass General Brigham (MGB) is a large academic health system and early leader in value-based care, with more than 700,000 patients currently in various value-based arrangements, primarily accountable care organization (ACO) contracts that include commercial, Medicare, Medicaid, and self-insured employed populations. MGB has matured a portfolio of programs designed to support patients across a variety of care settings.¹² MGB’s practice-based care management program, the integrated care management program (iCMP), has been shown to reduce utilization and spend among a Medicare population.¹³

In 2017, MGB entered a pilot Medicaid ACO arrangement with Massachusetts Medicaid (MassHealth). Utilizing Massachusetts Delivery System Reform Incentive Payment funding, iCMP Patients Linked to Urgent Support (iCMP PLUS) was created. This program provides intensive, community-based, multidisciplinary primary care beyond the traditional primary care practice for patients identified as having the highest cost and highest risk based on utilization and medical, social, and/or behavioral health complexity. For this program, MGB subcontracted Commonwealth Care Alliance (CCA), a nonprofit, Massachusetts-based organization with regional and national recognition, for services to complex dual-eligible patients.^14-17 We conducted a randomized controlled quality improvement evaluation of the early phase of the iCMP PLUS to determine its impact on acute care utilization (inpatient admission, observation status, emergency department [ED] visit) and total medical expense (TME), particularly in the context of an academic health system and outside nonprofit vendor collaboration.

METHODS

Study Setting and Design

Based in Massachusetts, MGB is composed of 2 academic tertiary hospitals, Massachusetts General Hospital (MGH) and Brigham and Women’s Hospital (BWH), and multiple community hospitals and practices. MGB has participated in value-based care since 2006 and currently has 125,000 Medicaid ACO members.¹⁸

At the time the iCMP PLUS program was launched, the Medicaid ACO included 35,000 members, and 9189 adult members were aligned to primary care practices affiliated with MGH and BWH. A total of 772 members were identified as high risk based on utilization and clinical characteristics.^13,19 This cohort was narrowed to those with TME in the top 5% of members in the year prior to randomization, while also including lower-cost members with at least 10 ED visits in the preceding year, resulting in a total of 479 members identified as high risk, high cost. Chart review was conducted by a central team of physician reviewers who evaluated the electronic health record (EHR; primary care provider [PCP] notes, specialist notes, medication lists, problem lists, discharge summaries, no-show visit rates) for clinical characteristics indicative of high medical, social, and/or behavioral health comorbidities; unclear or limited engagement with the attributed PCP; and high likelihood of benefiting from the proposed program, resulting in a cohort of 222 patients. The central reviewers provided recommendations regarding whether 1 of 2 intensive care management approaches—either a Wrap model, in which patients continue to receive primary care from their existing PCP, or the ambulatory intensive care unit (AICU) model, in which patients receive primary care from a CCA physician—would be most appropriate based on the patient’s engagement with traditional practice-based primary care. Patients were excluded if they lacked any inpatient admissions or ED visits in the prior 3 months, had high fixed medication costs (ie, infusion therapy), were primarily treated by an oncologist, received primarily pediatric care, or had missing EHR information. The final step involved PCP validation; PCPs reviewed their list of patients, as well as data on utilization, and identified 166 patients as candidates for intensive care management.

Identified patients were randomly assigned 2:1 to receive the intervention (n = 107) or standard care (n = 59). Standard care included access to MGB’s practice-based primary care and traditional care management program, iCMP.^13,19,20 PCPs and traditional care management staff were asked to not enroll intervention patients in traditional care management for 12 months following randomization to allow iCMP PLUS to enroll patients.

The MGB Institutional Review Board (IRB) reviewed the details of the trial and deemed it quality improvement and thus exempt from IRB review.

Intervention

The intensive care management intervention was provided by a multidisciplinary team with expertise in medical, behavioral, and social service provision, with the option for home-based services and/or home-based primary care. The care team included nurse practitioners (NPs), physicians (primary care and psychiatry), social workers, behavioral health and substance abuse specialists, and health outreach workers. Prior to enrollment, patients received significant outreach by CCA health outreach workers to enhance engagement, which included phone calls, clinic visits, home visits, and visits to nonhome, nonclinic sites (eg, homeless shelters, coffee shops) to meet their new care team. Patients in both the Wrap and AICU models received intensive care coordination, which included (1) comprehensive initial assessment, (2) a detailed patient-centric care plan highlighting plans to address gaps in care and engagement barriers, (3) CCA NP management, (4) social services and behavioral health support, (5) access to mobile integrated health and psychiatric crisis stabilization units (CSUs), (6) joint visits between the patient and primary care and specialist providers, and (7) chronic disease management follow-up (eg, wound care, diabetes, blood pressure). Patients were discharged from the program (6 of 87 intervention patients) when they met care goals, chose to be discharged, or did not follow up consistently with care delivered (1 patient). Figure 1 provides an overview of the care team members and details the unique program features. The mean panel size per NP was 70 patients. Patients had access to a CCA brick-and-mortar clinic, although home visits provided by the health outreach worker, NP, physician, behavioral health specialist, or other team member were more common. In addition, care was provided in various locations based on the patient’s needs (ie, patient’s home, homeless shelter, ED, other community setting). Finally, there were myriad efforts to integrate CCA and MGB, including credentialing of CCA staff at MGB, ensuring access to EHRs, and education of MGB providers about CCA offerings.

Measures and Data Sources

The prespecified primary outcome measure was TME for services delivered both inside and outside the MGB system, including all facility inpatient and outpatient, professional, ambulance, and laboratory costs for medical, surgical, and behavioral health services, but excluding retail pharmacy costs. Secondary measures included all-cause acute inpatient admissions, ED visits, and observation stays in any location. Outcome measures were developed using Medicaid claims and membership data provided by MassHealth. Claims and associated costs for substance abuse treatment services were not provided.

Patient characteristics, including age, gender, race/ethnicity, primary language, marital status, educational attainment, and baseline risk score (Adjusted Clinical Groups [ACG] version 11.2), were obtained from Medicaid member files and EHR data housed within the MGB Enterprise Data Warehouse.

Program Costs

Mean programmatic costs include clinical and operational staff costs, and the annual programmatic cost was $583,000 during the study period. For patients in the intervention group, these costs included utilization of mobile integrated health, CSUs, and other ancillary supports such as transportation and medication management tools. The costs included a monthly per-member per-month (PMPM) cost and per diem costs for CSUs. PMPM costs extended to not only those enrolled but also those identified for enrollment but not yet enrolled, as many patients required continued outreach attempts. The program costs were agreed upon prior to the start of the pilot.

Study Cohort

The analysis compared patients randomly assigned to the intervention with patients randomly assigned to control, using the date MGB released detailed patient data to the intensive care management program (June 1-26, 2017) as the randomization date. Control patients were assigned the median intervention patient randomization date (June 21, 2017) as their randomization date. Analytic cohorts were limited to patients with a minimum of 6 months of Medicaid enrollment in both the year prior to and the year post randomization date to ensure stable estimates. Of the 107 patients assigned to iCMP PLUS, 87 had sufficient Medicaid enrollment to be included in the analysis, and 44 of the 59 patients assigned to control had sufficient Medicaid enrollment to be included in the analysis. Of the 87 intervention patients included in the analysis, 66 had been assigned by their PCP to Wrap and 21 had been assigned by their PCP to AICU.

Analysis

Descriptive statistics were used to compare patient characteristics, health care utilization, and costs in the year prior across cohorts. We compared baseline utilization and costs by month relative to randomization date to ensure similar patterns across cohorts in the year prior. Difference-in-difference models were used to estimate the difference in prerandomization to postrandomization changes in outcomes for intervention patients relative to controls, adjusting for patient gender, primary language, and ACG rank probability of high cost, and weighting for the number of months of Medicaid eligibility. TME outcomes were modeled using linear models with random intercepts. Utilization outcomes were modeled using negative binomial models with log links and random intercepts. Least squared estimates of prerandomization and postrandomization TME and utilization were derived from models. Engagement rates in the year post randomization were developed using programmatic data provided by CCA and the iCMP teams, and patients were considered to be engaged if they were actively involved in defining their care plan.

RESULTS

The mean age of patients identified as appropriate for intensive care management was 48.7 years; the majority were women (62%), English speakers (69%), and not currently married (85%); 32% were Black, 21% were Hispanic, and 32% were White. The mean TME for selected patients was $5978.30 PMPM in the year prior. Baseline characteristics of patients in the intervention and control cohorts were similar (Table 1).

Intent-to-Treat Analysis

Patients randomly assigned to receive intensive care management had a $1933 smaller increase (P = .04) in TME PMPM 12 months prerandomization to 12 months post randomization than patients randomized to standard care (Figure 2). Although not statistically significant, intervention patients had 17.3% lower growth in ED visits (P = .40) and 34.2% lower growth in inpatient admissions (P = .29) but 47% larger growth in observation stays (P = .33) 12 months prerandomization to 12 months post randomization (Table 2). In the year post randomization, engagement rates among patients randomly assigned to iCMP PLUS were significantly higher than engagement rates among patients randomly assigned to iCMP (73.8% vs 25.4%; P = .0008) (Table 3).

DISCUSSION

This randomized controlled quality improvement trial showed that intensive care management of high-risk Medicaid ACO patients resulted in a mean $1933 smaller increase in TME per month compared with those randomly assigned to usual care—annualized to $23,196 savings per member to the state’s Medicaid program. This study is unique in that it illustrates the impact of a distinctive home-/community-based intensive care management program on a health system–based high-risk Medicaid population, providing multidisciplinary team-based care inclusive of primary care to meet their particular needs in the most appropriate setting. The partner vendor, CCA, has a unique clinical model that meets patients’ needs, beginning with social determinants of health, to address care delivery barriers.

There are relatively few published randomized controlled trials of care management.^9-11 The Camden Coalition randomized 800 Medicare and Medicaid patients with high utilization to usual care compared with care management and showed no benefit with respect to 6-month readmission.⁹ CareMore’s program, which randomly assigned 253 Medicaid patients to usual care or care management, showed statistically significant reductions in TME and inpatient utilization.¹⁰ These differences may be attributable to intervention design, including identification of patients more likely to have persistently high costs and differences in the engagement and enrollment of patients and the measurement period for outcomes. The approach outlined in this study involves more intensive, continuous, and holistic care management, including home-based care tailored to patient needs and focused on the highest-risk patients.

We hypothesize that the TME reduction was driven partly by curtailing utilization affected by ED provider practice change. For example, all intervention patients received a comprehensive acute care plan that ED physicians could review on patient arrival, mitigating repetitive workup and admission. Similarly, providers’ awareness that intervention patients would have follow-up by an outpatient care team may have resulted in deferring workup, admission, and future ED visits.

There are multiple possible drivers of the favorable effects on cost and utilization, likely due to curtailing rising costs in the intervention cohort. First, our multistep approach to patient selection relied on (1) claims-based identification, (2) clinical characteristics, and (3) detailed chart review to delineate health drivers and opportunities to reduce preventable utilization, as well as primary care validation as to who would most likely benefit from intensive, longitudinal care management.^13,19,20 This methodical approach to patient selection, ensuring that those with the highest need were included, likely contributed to the findings. Second, our program leveraged NPs as the primary quarterbacks of care coordination and enhanced provision of community-based primary care, supported by community health workers, registered nurses, social workers, physicians, and behavioral health specialists. In addition to the longitudinal care provided by the behavioral health specialists, the 24/7 access to CSUs allowed patients to avoid an ED visit and/or inpatient psychiatric admission. Third, the CCA teams were resourced to meet the patients where they most needed support—telephonically or in person, and in a traditional practice or in the patient’s home, homeless shelter, ED, or other community setting. The ability to intensively engage across sites of care was critical to facilitate program enrollment and engagement. Finally, by selecting patients for whom baseline TME was high ($5953 and $6015 PMPM for the intervention and control groups, respectively) and potentially impactable, there was an ability to demonstrate sizeable reductions that offset the cost of the intervention.

Another novel element of this program was its implementation in a large health system with experience in value-based care that contracted clinical and care management services to another provider group with specialized expertise. As organizations targeting specific patient segments mature across the country, such as those focused on end-stage renal disease, dual-eligible or Medicaid beneficiaries, older patients, and other complex populations, these modular collaborations may be increasingly common.^21-23 Success of these models relies on effective integration. To facilitate integration, CCA providers had complete access to MGB’s EHR and were MGB credentialed. Similarly, protocols and procedures were shared. Biweekly meetings were held to discuss program operations, and monthly meetings were held to review clinical cases in real time, with ongoing communication between MGB and CCA to troubleshoot operational challenges.

We examined the impact of patient engagement and found a very significant increased rate of patient engagement among patients in intensive care management. We believe that the increased efforts and success in engaging patients may be a major driver of success in this program.

Since the initial pilot period, our programmatic model has evolved in several important ways. First, we honed our approach to patient selection using an enhanced algorithm that leverages claims and clinical data, and we currently do not rely on intensive chart review. Utilization thresholds for programmatic inclusion are now significantly higher than in the earlier phases of this program. Second, MGB and CCA have worked to further define optimal team structures and staffing ratios. Third, MGB and CCA have improved communication workflows and protocols between the care team and PCPs. Cultivating the relationship between the care team and primary care is important to ensure the trust that is crucial to long-term program success. This takes significant time, good faith, and effort from both sides.

Our evaluation has several policy implications. First, we identified a segment of the Medicaid-only population that phenotypically appeared closer to a patient seen in a dual-eligible program, rather than the average Medicaid patient. This finding suggests that state and federal policy makers should consider further incentivization of care management beyond existing limited payments.²⁴ Enhanced care management payments would incentivize health systems, payers, and providers to increase investments in care management, particularly intensive care management that is highly resourced, to offset TME in high-risk patients. Because of the significant up-front investment required to launch care management programs, provider groups may be wary of investing in resources, even if they may ultimately result in cost savings. The Center for Medicare & Medicaid Innovation has the ability to pilot varied value-based care approaches, and a model that incorporates incentivization of intensive care management would align with its overall goals of testing payment models.

Limitations

There are several key limitations to highlight. This pilot evaluation was limited by a small sample size, which affected the ability to detect statistical significance despite meaningful impact on metrics of utilization. Sample size was in part affected by the removal of 35 patients post randomization (20 from the intervention group and 15 from the control group) due to lack of sufficient claims data. The program was implemented at 2 academic medical centers and did not include community practices, which limits generalizability. In addition, a key element of the intervention was providing support for patients with substance use disorders; however, because of privacy regulations, we did not have access to claims related to substance use. This study also did not capture additional metrics evaluating quality of care, which would be helpful for future research to address. Finally, there was no formal analysis of patient or provider experience with the program, which would have contributed to a better understanding of impact beyond utilization and cost.

CONCLUSIONS

The findings of this study suggest that an intensive care management program focused on high-risk Medicaid patients can favorably affect expenditure and utilization and that academic medical centers can partner effectively with nonprofit community-based primary care organizations. Care management components that are tailored to meet the needs of patients who have the highest impactable utilization are important elements of an intensive care management program. Further study in diverse settings is warranted to confirm the findings of this pilot trial with promising results.

Author Affiliations: Population Health Management, Mass General Brigham (JSR, JG, MV, LN, EM, MLM, CV), Somerville, MA; Harvard Medical School (JSR, MLM, SKC, CV), Boston, MA; The Mongan Institute, Massachusetts General Hospital (JG, LN, CV), Boston, MA; Commonwealth Care Alliance (LWT), Boston, MA; Massachusetts General Hospital (JSR, EW, SKC), Boston, MA; Brigham and Women’s Hospital (LWT), Boston, MA; CVS Health (SKC), Woonsocket, RI.

Source of Funding: Massachusetts Delivery System Reform Incentive Payment grant provided by MassHealth to Mass General Brigham as part of Mass General Brigham’s Medicaid accountable care organization program participation.

Author Disclosures: Dr Rowe is now employed by agilon health, which helps providers participate in value-based care. Dr Tishler is employed by Commonwealth Care Alliance and was involved in the development and implementation of the iCMP PLUS intervention. Dr Chaguturu is employed by CVS Healthcare. 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 (JSR, LWT, EW, SKC); acquisition of data (JSR, JG, LN, CV); analysis and interpretation of data (JSR, JG, MV, LN, CV); drafting of the manuscript (JSR, JG, EM, MLM, LWT, CV); critical revision of the manuscript for important intellectual content (JSR, MV, EM, MLM, LWT, EW, CV); statistical analysis (JG, LN, CV); provision of patients or study materials (LWT, SKC); administrative, technical, or logistic support (MV, EM, MLM, EW, CV); and supervision (SKC, CV).

Address Correspondence to: Christine Vogeli, PhD, Massachusetts General Hospital, 100 Cambridge St, Boston, MA 02114. Email: CVOGELI@mgh.harvard.edu.

REFERENCES

1. Hong CS, Abrams MK, Ferris TG. Toward increased adoption of complex care management. N Engl J Med. 2014;371(6):491-493. doi:10.1056/NEJMp1401755

2. Hong CS, Siegel AL, Ferris TG. Caring for high-need, high-cost patients: what makes for a successful care management program? The Commonwealth Fund. August 2014. Accessed May 6, 2021. https://www.commonwealthfund.org/sites/default/files/documents/___media_files_publications_issue_brief_2014_aug_1764_hong_caring_for_high_need_high_cost_patients_ccm_ib.pdf

3. Anderson GF, Ballreich J, Bleich S, et al. Attributes common to programs that successfully treat high-need, high-cost individuals. Am J Manag Care. 2015;21(11):e597-e600.

4. Hasselman D. Super-utilizer summit: common themes from innovative complex care management programs. Center for Health Care Strategies. October 2013. Accessed May 6, 2021. https://www.chcs.org/media/FINAL_Super-Utilizer_Report.pdf

5. Johnson TL, Rinehart DJ, Durfee J, et al. For many patients who use large amounts of health care services, the need is intense yet temporary. Health Aff (Millwood). 2015;34(8):1312-1319. doi:10.1377/hlthaff.2014.1186

6. Sabik LM, Bazzoli GJ, Carcaise-Edinboro P, Chandan P, Harpe SE. The impact of integrated case management on health services use and spending among nonelderly adult Medicaid enrollees. Med Care. 2016;54(8):758-764. doi:10.1097/MLR.0000000000000559

7. Waxmonsky JA, Giese AA, McGinnis GF, et al. Colorado Access’ enhanced care management for high-cost, high-need Medicaid members: preliminary outcomes and lessons learned. J Ambul Care Manage. 2011;34(2):183-191. doi:10.1097/JAC.0b013e31820f64be

8. Xing J, Goehring C, Mancuso D. Care coordination program for Washington State Medicaid enrollees reduced inpatient hospital costs. Health Aff (Millwood). 2015;34(4):653-661. doi:10.1377/hlthaff.2014.0655

9. 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

10. Powers BW, Modarai F, Palakodeti S, et al. Impact of complex care management on spending and utilization for high-need, high-cost Medicaid patients. Am J Manag Care. 2020;26(2):e57-e63. doi:10.37765/ajmc.2020.42402

11. Bell JF, Krupski A, Joesch JM, et al. A randomized controlled trial of intensive care management for disabled Medicaid beneficiaries with high health care costs. Health Serv Res. 2015;50(3):663-689. doi:10.1111/1475-6773.12258

12. Bakshi S, Schiavoni KH, Carlson LC, et al. The essential role of population health during and beyond COVID-19. Am J Manag Care. 2021;27(3):123-128. doi:10.37765/ajmc.2021.88511

13. Hsu J, Price M, Vogeli C, et al. Bending the spending curve by altering care delivery patterns: the role of care management within a Pioneer ACO. Health Aff (Millwood). 2017;36(5):876-884. doi:10.1377/hlthaff.2016.0922

14. Klein S, Hostetter M. In focus: redesigning primary care for those who need it most. The Commonwealth Fund. March 2016. Accessed May 6, 2021. https://www.commonwealthfund.org/publications/2016/mar/focus-redesigning-primary-care-those-who-need-it-most

15. Klein S, Hostetter M, McCarthy C. The “One Care” program at Commonwealth Care Alliance. The Commonwealth Fund. December 8, 2016. Accessed June 2, 2021. https://www.commonwealthfund.org/publications/case-study/2016/dec/one-care-program-commonwealth-care-alliance

16. Master R. Realizing the promise of integrated care for the ‘dual eligible.’ Health Affairs. October 2012. Accessed June 2, 2021. https://www.healthaffairs.org/do/10.1377/hblog20121022.024416/full/

17. Iezzoni LI, Dorner SC, Ajayi Y. Community paramedicine—addressing questions as programs expand. N Engl J Med. 2016;374(12):1107-1109. doi:10.1056/NEJMp1516100

18. About population health. Partners HealthCare. Accessed May 6, 2021. https://populationhealth.partners.org/population-health-overview/

19. Vogeli C, Spirt J, Brand R, et al. Implementing a hybrid approach to select patients for care management: variations across practices. Am J Manag Care. 2016;22(5):358-365.

20. Haime V, Hong C, Mandel L, et al. Clinician considerations when selecting high-risk patients for care management. Am J Manag Care. 2015;21(10):e576-e582.

21. Our starting point: forging a new path towards value. Somatus. Accessed June 3, 2021. https://somatus.com/our-starting-point/

22. About our mission, team & leadership. Cricket Health. Accessed June 3, 2021. https://www.crickethealth.com/about/

23. Clinical model. CareMore Health. February 9, 2021. Accessed June 3, 2021. https://www.caremore.com/About/Clinical-Model.aspx

24. Chronic care management services. CMS. July 2019. Accessed June 3, 2021. https://www.cms.gov/outreach-and-education/medicare-learning-network-mln/mlnproducts/downloads/chroniccaremanagement.pdf