A Case Report: Cornerstone Health Care Reduced the Total Cost of Care Through Population Segmentation and Care Model Redesign

Abstract

Over the course of a single year, Cornerstone Health Care, a multispecialty group practice in North Carolina, redesigned the underlying care models for 5 of its highest-risk populations—late-stage congestive heart failure, oncology, Medicare-Medicaid dual eligibles, those with 5 or more chronic conditions, and the most complex patients with multiple late-stage chronic conditions. At the 1-year mark, the results of the program were analyzed. Overall costs for the patients studied were reduced by 12.7% compared to the year before enrollment. All fully implemented programs delivered between 10% and 16% cost savings. The key area for savings factor was hospitalization, which was reduced by 30% across all programs. The greatest area of cost increase was “other,” a category that consisted in large part of hospice services. Full implementation was key; 2 primary care sites that reverted to more traditional models failed to show the same pattern of savings.

Six years after the Affordable Care Act put the spotlight on value-based care and accountable care organizations (ACOs), there is still a shortage of case studies illustrating the practical steps required to implement value-based care and a dearth of publicly accessible reports providing detailed data on how successfully ACOs achieve the Triple Aim¹ of reduced costs, improved quality of care, and better outcomes for patients.² The recent publication of 2 trials of patient-centered medical homes (PCMHs), a foundational element of an ACO, in Pennsylvania that yielded conflicting results,^3,4 and the accompanying editorial,⁴ point out the need for specific information on which elements of design and implementation contribute to the success or failure of an effort.⁵

Providers need to understand which specific tactics work, which do not, and why, so that the industry as a whole can learn from the successes and failures of other health care organizations.

In the belief that the short-term solution is for individual organizations to make available detailed data on their own experience with value-based care, Cornerstone Health Care would like to share details about an ACO that it successfully implemented in 2013.

Overview

Cornerstone is a multispecialty medical group in North Carolina with more than 330 physicians and advanced practice professionals, and 80 practice locations in a 12-county region that utilizes 18 hospitals operated by 6 separate health systems. Cornerstone prospered as a fee-for-service (FFS) enterprise despite being one of the lowest cost providers in the region. The viewpoint within the organization was that Cornerstone is progressive: it was one of the nation's first adopters of electronic medical records and had achieved National Committee for Quality Assurance (NCQA) level-3 recognition with multiple PCMHs for primary care. But at the core, the organization was pursuing a productivity model, and by 2010 it because apparent that the initiatives Cornerstone had undertaken were insufficient to either fulfill the commitment to the Institute of Medicine's Six Aims for health care or to maintain independence and economic viability in a rapidly consolidating health care market.⁶ After evaluating the options, Cornerstone committed to an accelerated transformation from a FFS model to value-based care across its Medicare, Medicaid, and commercially insured patients.

Cornerstone was familiar with the approaches taken by many other health care providers: use care coordinators to manage transitions of care,⁶ direct patients to appropriate lower-cost care providers, and address “low-hanging fruit” through organization-wide efforts to reduce unnecessary or redundant diagnostics⁷ and encouraging the use of generic drugs. This approach can, and often does, modestly reduce the cost of care, at least in the short term. But it does little to improve long-term outcomes, keep patients healthier, or prevent the progression of risk factors to chronic disease—the areas in which accountable care can have its greatest and most sustainable impact.

Cornerstone believed that deeper change was essential. In addition to introducing new resources, central elements of patient care delivery were completely redesigned: staffing, care team roles,⁸ policies and procedures, physical layout of practices, and patient engagement methods—with a relentless focus on improving patient outcomes. The approach was to develop new models of care for targeted patient populations: those who in the traditional FFS system were at high risk of poor health outcomes and high costs.

For the first round of implementation, 5 care models were created. Two were based on disease-related criteria: one for cancer patients⁹ and one for patients with late-stage congestive heart failure (CHF). A third care model focused on Medicare-Medicaid dual-eligible patients, who had proven difficult to treat—or sometimes reach—because of social and behavioral health factors. The final 2 were designed for high-opportunity primary care populations, which together represented approximately 15%–20% of the patients. One of these targeted the highest-risk complex patients, many of whom were frail and elderly. The second focused on patients with 5 or more chronic diseases. It was believed that these patients would particularly benefit from a coordinated, team-based approach to care.

After 1 year—a brief period in a health care reform program—Cornerstone was able to measure the effectiveness of 4 of the 4 programs in its patient population. (The oncology program could not be included for reasons that will be explained.) The transformation program as a whole yielded a nearly 13% decrease in total claims spend on the ACO population (compared to prior year) (Table 1), and as much as a ∼19% decrease in specific population segments. Savings relative to predicted costs (assuming cost trend growth) were even greater. Each of the 4 evaluable models achieved a similar level of savings. The organization believes that the magnitude of these results, after just 1 year of operation, supports the care model transformation approach. By sharing these results in some detail, Cornerstone demonstrates both the opportunities created by discretely described population-specific care models and some of the pitfalls when their implementation is derailed.

Table 1.

Percentage of Costs Attributed to Different Service Areas, Year Prior to Study

	Cornerstone heart function (%)	Cornerstone life care (%)	Cornerstone care outreach clinic (%)	Cornerstone personalized primary care practices (fully implemented) (%)	Aggregate for fully implemented programs (%)	Cornerstone personalized primary care practices that did not fully adopt model (%)
Inpatient hospital	70	54	47	54	55	47
Outpatient hospital	4	12	17	7	9	13
Home health	5	8	11	8	8	7
SNF	8	9	5	7	8	7
Office	5	5	7	9	7	9
ED	1	2	4	3	2	3
Injectable drugs	2	2	0	3	2	5
Laboratory	1	2	2	2	2	3
DME	1	2	1	2	2	1
Advanced imaging	0	1	1	2	1	2
ASC	1	1	0	1	1	1
Rehab (PT/OT)	0	1	1	1	1	0
Other	1	3	3	2	3	2

ASC, ambulatory surgery center; DME, durable medical equipment; ED, emergency department; OT, occupational therapy; PT, physical therapy; SNF, skilled nursing facility.

Five Care Models—Description of the Interventions

All of the care models share certain features because certain approaches such as nurse navigators and integrated behavioral medicine appeared to work in many settings. But each population had its own unique needs, so the care models were customized accordingly.

As already explained, 2 programs targeted high-risk patients with particular diseases:

Cornerstone Heart Function Clinic was designed for patients with late-stage (Class III and Class IV) CHF. Although many health systems have programs for CHF, most focus on preventing the readmission of previously hospitalized patients.¹⁰ Cornerstone sought to proactively identify patients at high risk of hospitalization and keep them out of the hospital by focusing on reducing exacerbations, improving quality of life, and providing palliative care. Multiple triggers¹¹ were used to identify patients for the program, including antecedent hospital admission, rapid health changes, cardiologist recommendation, and primary care referral. The Heart Function Clinic was physically located in an existing cardiology practice and staffed by 3 cardiologist “champions,” who were assigned to the clinic on a rotating basis. A pharmacist and psychologist¹² were embedded in the clinic. The team also included registered nurse (RN) health navigators, a nutritionist and a full-time nurse practitioner—all of whom worked in close coordination with the cardiologists.¹³

Cornerstone Oncology focused on a problem identified during the analysis phase: a large share of the total costs generated by cancer patients came not from cancer treatment itself, but from comorbidities such as diabetes and heart disease. To address these issues, a full-time internist was embedded in the oncology practice, treating the whole patient rather than just the tumor. Also incorporated was RN care navigation specifically for patients with lung cancer or breast cancer to help them cope with the medical and life challenges unique to these conditions. For example, breast cancer patients were assisted with the prosthesis process and finding and fitting wigs. Other features of the program include web-based access to patient support groups, on-site pharmacy, behavioral medicine, and palliative care resources. As with the CHF model, these elements were embedded within an existing oncology practice.

The other 3 care models focused not on specific diseases, but on patients meeting discrete criteria placing them at high risk of experiencing decreased health status with resulting increases in utilization and cost:

Cornerstone Care Outreach Clinic (CCOC). This program was designed for patients eligible for both Medicare and Medicaid (dual eligible) insurance, as well as low-income patients insured by Medicaid or Medicare only. As has been well documented, a high burden of complex physical illnesses and psychiatric disease leads to very high costs in this patient population.¹⁴ The clinic was structured to provide intensive case-managed care with increased time for individual patient visits, clinical and behavioral health team meetings centered on patient panel management, a physician-led team experienced in complex care, and improved integration with community-based resources (substance abuse treatment, mental health services, and other social services). The CCOC provided fully integrated mental health care to patients on site. Based on the IMPACT collaborative care model for treatment of depression,¹⁵ there was comprehensive screening for depression using the Patient Health Questionnaire-9¹⁶ screening tool, substance abuse, and other mental illnesses, with structured treatment protocols based on patient registries and regular tracking of data and outcomes.^17,18 Staffed by a licensed clinical social worker (who integrates counseling and social work roles) and a 0.2 full-time psychiatrist, the clinic has achieved remission rates for depression at and above those of the IMPACT studies.¹⁹ The clinic focused resources to overcome identified barriers to care such as low literacy, lack of transportation, and poor access to health education to enable better patient outcomes.

Cornerstone Life Care (CLC) provided comprehensive, coordinated care for highly complex patients with multiple late-stage chronic conditions—the 3% of the organization's patients with the highest level of risk, and the sickest and most frail population. Using the Impact Pro (Optum, Eden Prairie, MN) risk stratification scores, high-risk patients living within a 20-mile radius of the clinic were identified and offered care in the CLC clinic. Patients were served in a stand-alone facility designed to provide easy access for the frail and elderly. A team, consisting of a lead internal medicine physician, nurse practitioner, clinical pharmacist, RN patient navigator, and social worker, collaborated closely with other caregivers, specialists, home health agencies, and palliative care agencies. The program focused on improving patient quality of life, functionality, and self-care ability, while reducing cost and unnecessary health care utilization.

At the time of patient enrollment in the program, a comprehensive health assessment and review of medical records was performed to identify gaps in care and other opportunities for improving care. A clinical pharmacist evaluated all medications, including prescriptions, over-the-counter medications, and supplements, and then streamlined therapy to reduce drug-drug, drug-disease interactions and potential overdoses. The pharmacist managed the medications to achieve goals set by the physician and patient, often intensifying therapy based on patient physiology, potential side effects, and care preferences. Given the intensity and needs of this population, the care team was responsible for fewer patients and there was an emphasis on patient navigation and proactive health status monitoring. Patients maintained touch with their principal primary care provider but were managed by the team at CLC.

Cornerstone Personalized Primary Care (CPPC) was designed to provide enhanced primary care to patients with 5 or more chronic diseases. These patients are more medically stable than those in the CLC; therefore, the panel size for the care team was larger and the approach focused more on stabilization of the patient's condition to prevent deterioration. In this model, a general internist led a team that included a geriatric nurse practitioner, an RN health navigator, and a medical assistant–with access to behavioral medicine and retail pharmacy-coordinated health coaching through a unique partnership with Rite Aid providing access to health coaching in the community.²⁰ Extensive care plan usage, a daily huddle, and extended-hours access are key attributes of the model. CPPC is embedded in internal medicine clinics, which themselves reside in ambulatory care centers that are open 7 days a week. Each of these centers has outpatient imaging, infusion, lab, and urgent care services on site, providing convenient lower-cost alternatives to hospital and emergency room sources of these services.

CPPC was implemented at 4 locations. In 2, the implementation went as planned; in 2 it did not. One was initially built on the CPPC model but over the course of the year reverted to a pattern of practice more consistent with traditional FFS; the other added the services of a navigator but never implemented the high-intensity team approach. For purposes of the analysis, only the full implementations are included, although the performance of the less-than-full implementations will be shown and discussed.

Methods

This is a retrospective evaluation of the changes in cost for Medicare, dual-eligible Medicare and Medicaid, and commercially insured Medicare Advantage patients for whom Cornerstone Health Care had risk or results-sharing arrangements with a payer who could provide paid claims data. Although the same care measures were applied to all patients regardless of payer, commercial patients were excluded from the analysis because sufficiently detailed data on them could not be obtained from the payers–even those in accountable care contract arrangements.

Data collection

Patient identification

For the purposes of the analysis, all patients cared for by Cornerstone Health Care and members of the 4 programs noted were identified. The analysis was then limited to those who were enrolled in the fully implemented transformation programs between August 2013 and May 2014 for at least 3 months and for whom there was at least 1 year's worth of claims records for the 12 months preceding the launch of the transformation program. Because of the life cycle of cancer treatment, most patients in the oncology program were diagnosed within the year studied. As a result, there were no meaningful comparators and the entire program was omitted from the analysis. The programs that could be analyzed enrolled a total of 2060 patients. Of those, 261, including both traditional Medicare (Medicare Shared Savings Program [MSSP]) and United Healthcare Medicare Advantage patients, were included in this analysis.

Calculation of costs

Claims were tracked at a patient level for the year before and the year after enrollment. Paid claims for the entire available enrollment period (from 3 to 12 months) were summed, and if the enrollment period was less than a year, claims were annualized for each patient, using a program-level annualization factor. Claims were analyzed for each care model using 13 standard categories of utilization, such as inpatient hospital care, advanced imaging, injectable drugs, among others.

The claims reflect actual costs to payers. Program costs to Cornerstone, including additional personnel and informatics tools are much more difficult to calculate and allocate across programs and are not considered here.

The true savings generated by a care model are not just savings in real claims compared to the previous year, but savings in real claims compared to what they would have been without the intervention. To estimate the anticipated costs of care for the study period, the Impact Pro tool, which predicts costs based on each individual's demographic and risk factors, was used. The tool uses a member's clinical episodes of care, prior use of health care services, prescription drugs, and lab results as markers of their future health risk and creates markers of risk that can be both predictive and provide clinical insights into why a patient is high risk. The tool predicts both future expenditures and calculates the probability of 1 or more hospitalizations in a methodology delineated as Symmetry Episode Treatment Groups (Optum, Eden Prairie, MN). Because the predictions generated by the Impact Pro tool include cost for pharmacy, and the study claims data did include complete Medicare Part D pharmacy costs, the “before” costs were upwardly adjusted by including an additional 9.7% for medications. (The claims data supplied by Centers for Medicare & Medicaid Services for participants in MSSP ACOs does not include the full cost of Part D medications. Only the amount paid by the patient is included in the claims provided. Because of this lack of data, the total cost of care for these patients could not be calculated. The Impact Pro predicted total cost of care includes pharmacy costs. Observed costs were increased to account for pharmacy costs and to match the Impact Pro predicted costs based on Centers for Disease Control and Prevention data.)²¹

Results

Overall results

Table 2 shows the results in actual dollars for each of the 4 models studied and for the 2 internal medicine practices that did not fully convert to the new model for comparison. Note that in every instance among the intervention groups the spend in the study year was less than the prior year's spend and less than the estimated spend for the study year. The spend for the 2 sites that did not fully implement the changes was higher than the previous year, but much less than that forecast for the implementation year.

Table 2.

Summary Program-Level Results

	Cornerstone heart function	Cornerstone life care	Cornerstone care outreach clinic	Cornerstone personalized primary care practices (fully implemented)	Aggregate for fully implemented programs	Cornerstone personalized primary care practices that did not fully adopt model
No. of patients in study sample	21	56	25	159	261	169
Average per person spend in year before study	$40,284	$51,289	$23,616	$18,574	$30,771	$20,644
Average per person spend in study year (actual)	$34,754	$45,815	$19,805	$16,051	$26,876	$24,448
Predicted average per person spend in study year	$38,583	$63,374	$35,900	$26,631	$38,224	$32,754.52
Aggregate spend in year before study	$845,957	$2,872,164	$590,403	$2,953,340	$7,261,863	$3,488,755
Aggregate spend in study year (actual)	$729,840	$2,565,657	$495,130	$2,552,148	$6,342,775	$4,131,713
Predicted aggregate spend in study year	$810,246	$3,548,963	$897,500	$4,246,625	$9,503,334	$5,384,795
% change (before actual)	−13.7%	−10.7%	−16.1%	−13.6%	−12.7%	18.4%
% difference of actual to predicted	−9.9%	−27.7%	−44.8%	−40.0%	−33.3%	−23.3%

As shown in Table 3, total medical costs for the patients included in the new models were reduced by 12.7% compared to the year before enrollment, and inpatient hospital costs were reduced by 29.4%. The reduction in inpatient costs is a key driver of overall savings because inpatient costs accounted for 47% to 70% of total medical costs in the year before the study period. All programs had 2- to 3-fold increased costs in the “Other” category. On a percentage basis, the increases in cost in this category went up 2- to 3-fold as well (Table 4). Previous analyses of this category indicate that it consists primarily of hospice services. The study team thinks this means that each of the programs reviewed is more effectively incorporating palliative care and the use of advance care planning into their care delivery.

Table 3.

Percent Change in Spend by Service Area (Prior to Actual)

	Cornerstone heart function (%)	Cornerstone life care (%)	Cornerstone care outreach clinic (%)	Cornerstone personalized primary care practices (fully implemented) (%)	Aggregate for fully implemented programs (%)	Cornerstone personalized primary care practices that did not fully adopt model (%)
Overall program	−13.7	−10.7	−16.1	−13.6	−12.7	18.4
Inpatient hospital	−32.5	−13.3	−23.3	−44.9	−29.4	19.9
Outpatient hospital	16.1	−19.1	−26.7	40.2	−0.6	−12.4
Home health	71.5	−19.9	−0.6	18.6	5.2	17.0
SNF	−4.2	−28.9	−98.1	91.3	14.3	114.8
Office	49.6	17.4	18.4	−8.6	6.5	0.4
ED	−24.6	16.7	8.1	−28.4	−8.4	−8.2
Injectable drugs	68.0	−33.5	−57.0	−50.9	−30.9	−7.2
Laboratory	−8.4	35.8	−5.4	−26.0	−2.2	−22.5
DME	106.0	−36.1	38.6	−31.8	−24.8	−2.8
Advanced imaging	−3.5	−38.9	−47.3	−19.2	−27.4	−26.8
ASC	4.9	−34.7	992.4	−43.1	−25.2	−51.8
Rehab (PT/OT)	−100.0	−88.0	−78.5	−47.3	−72.5	206.6
Other	91.1	121.5	105.1	209.0	152.7	111.0

ASC, ambulatory surgery center; DME, durable medical equipment; ED, emergency department; OT, occupational therapy; PT, physical therapy; SNF, skilled nursing facility.

Table 4.

Percentage of Costs Attributed to Different Service Areas, Study Year

	Cornerstone heart function (%)	Cornerstone life care (%)	Cornerstone care outreach clinic (%)	Cornerstone personalized primary care practices (fully implemented) (%)	Aggregate for fully implemented programs (%)	Cornerstone personalized primary care practices that did not fully adopt model (%)
Inpatient hospital	55	53	43	34	45	47
Outpatient hospital	6	11	15	11	11	9
Home health	10	7	14	11	10	7
SNF	8	7	0	15	10	13
Office	8	7	10	9	8	7
ED	1	2	6	2	2	2
Injectable drugs	5	1	0	2	2	4
Laboratory	1	2	2	2	2	2
DME	1	2	2	1	1	1
Advanced imaging	0	1	1	1	1	1
ASC	1	0	1	1	1	0
Rehab (PT/OT)	0	0	0	0	0	1
Other	3	7	7	9	7	4

ASC, ambulatory surgery center; DME, durable medical equipment; ED, emergency department; OT, occupational therapy; PT, physical therapy; SNF, skilled nursing facility.

This analysis implicitly assumes that unit costs (the cost of an admission or a clinic visit) are stable from year to year. However, Medicare and commercial plans have annual cost escalators built in such that identical admissions are slightly more costly each year. Additionally, with the study population primarily comprised of pre-selected high-risk individuals, a trajectory of increasing utilization and therefore rising costs may be reasonably predicted.

Care model-specific results

The goal of each new care model is to improve and maintain the health and functional status of the patients involved by improving their access to evidence-based medical care, reducing emergency department and hospital admissions, and improving the patient's engagement in care (and thereby adherence with treatment). The result of achieving these goals is a decrease in utilization of expensive resources, with some increase in other, more appropriate areas, and an overall reduction in health care spending, or at least a moderation of the rate of rise of spending. Demonstration of the improvement in health and functional status for many medical conditions may require at least 3 years of clinical data and relatively large numbers of patients to show reductions in infrequent but severe complications. However, changes in utilization and cost may be demonstrable in less than a year; these are reported here.

Cornerstone Heart Function

The principal contributor to cost in CHF management is recurrent hospitalization, following a visit to the emergency department for shortness of breath, and subsequent transfer to a nursing home after discharge. During the baseline year, inpatient hospital costs accounted for 70% of total costs for CHF and skilled nursing facility (SNF) costs were 8%. Through early proactive outreach, addressing social issues such as diet and transportation, better recognition and treatment for depression, elimination of polypharmacy, and improving health literacy, this program reduced the cost of hospitalization by almost one third and of SNF care by approximately 4% after 1 year. Table 3 shows the cost impact by category of utilization (for all care models). Principle added costs were for ambulatory treatments including durable medical equipment (DME), office visits, home health, injectable drugs (primarily for intravenous diuretics administered in a physician office), and “other” (primarily hospice). Overall costs for CHF were reduced by 13.7%.

CLC

Overall, this program reduced the cost of care by 10.7%, the smallest percentage decrease among the programs that produced savings. However, because the cost per patient was the highest among the 4 groups studied (Table 2), the absolute dollars generated in savings were the greatest of any. Inpatient hospital costs decreased only 13% (Table 3), but there also were substantial decreases in the costs of most categories: advanced imaging, ambulatory surgical centers, home health, outpatient hospital, injectable drugs, rehab, and SNF. Individually these categories contribute much less to costs than does inpatient hospital use, but cumulatively they had a substantial impact, in total amounting to an estimated $4496.

Emergency department, laboratory and “other” costs did rise for this population. As expected, the cost of office visits increased by 17%. The goal of this model is to create a primary care experience that prevents the need for more expensive venues and forms of care. But it is worth noting that the costs in this analysis reflect those claims paid by Medicare and a commercial payer, not the cost to Cornerstone to deliver the service. Office visits in this model are much more expensive than in traditional models, requiring extra time, enhanced numbers of personnel, and expertise in the management of frail, highly medically complex patients.

CCOC

Many of the 25 patients in this program have social, psychological, and literacy issues making them challenging to treat effectively in an office setting. Office visit costs for this group rose by 18% during the study year, in part reflecting greater success in getting patients to show up for scheduled appointments. Another factor for this increase was the smaller panel size of patients being managed by the clinical team. This group saw reductions in inpatient hospital, outpatient hospital, rehab, and SNF (Table 3). Overall costs were reduced by 16.1%. The dramatic increases in spend for ambulatory surgery centers and for DME reflect the impacts of 1 or 2 patients requiring services in a very small group (Table 3).

CPPC

As enhanced PCMH, these 2 internal medicine practices cared for the largest number of patients (159) and accounted for the most money spent on health care of any of the 4 care systems (Table 2). They achieved a 13.6% reduction in total spend, coming mainly from inpatient hospitalization but including all other categories except outpatient hospital, home health, SNF, and other (Table 3).

The comparison group for the CPPC sites consists of 2 internal medicine clinics with similar characteristics that failed to convert to the new model. These 2 sites served an approximately an equal number of patients to that of the personalized sites (169 patients vs. 159). However, their base year performance was an average of 11.1% ($2070) greater expense than that of the study sites, and increased 18.4% overall during the study year (Table 3). They experienced rises in many categories of spend, including inpatient hospital, home health, SNF, rehabilitation, and other. Although the dramatic increase in expenditures for rehabilitation and other are noteworthy, the actual dollars involved are far less than those for inpatient hospitalizations (comparison of data in Tables 5 and 6).

Table 5.

Cost Per Service Area Prior to Study Year, Per Person Per Year

	Cornerstone heart function	Cornerstone life care	Cornerstone care outreach clinic	Cornerstone personalized primary care practices (fully implemented)	Aggregate for fully implemented programs	Cornerstone personalized primary care practices that did not fully adopt model
Overall program	$40,284	$51,289	$23,616	$18,574	$30,771	$20,644
Inpatient hospital	$28,351	$27,816	$11,045	$10,030	$17,051	$9,673
Outpatient hospital	$1704	$5990	$4003	$1268	$2851	$2594
Home health	$2008	$4019	$2698	$1526	$2446	$1532
SNF	$3032	$4618	$1249	$1297	$2372	$1472
Office	$1960	$2642	$1656	$1618	$2067	$1820
ED	$482	$929	$1058	$497	$710	$582
Injectable drugs	$995	$832	$56	$599	$696	$1019
Laboratory	$462	$799	$417	$426	$562	$554
DME	$224	$1120	$261	$285	$506	$299
Advanced imaging	$119	$422	$336	$291	$343	$318
ASC	$283	$262	$19	$181	$212	$225
Rehab (PT/OT)	$122	$327	$135	$93	$165	$87
Other	$541	$1511	$683	$464	$792	$467

ASC, ambulatory surgery center; DME, durable medical equipment; ED, emergency department; OT, occupational therapy; PT, physical therapy; SNF, skilled nursing facility.

Table 6.

Cost per Service Area in Study Year, per Person per Year

	Cornerstone heart function	Cornerstone life care	Cornerstone care outreach clinic	Cornerstone personalized primary care practices (fully implemented)	Aggregate for fully implemented programs	Cornerstone personalized primary care practices that did not fully adopt model
Overall program	$34,754	$45,815	$19,805	$16,051	$26,876	$24,448
Inpatient hospital	$19,132	$24,115	$8471	$5525	$12,044	$11,600
Outpatient hospital	$1978	$4845	$2932	$1777	$2833	$2273
Home health	$3443	$3218	$2683	$1809	$2573	$1792
SNF	$2904	$3281	$23	$2482	$2712	$3162
Office	$2931	$3103	$1960	$1478	$2200	$1826
ED	$363	$1085	$1144	$356	$651	$534
Injectable drugs	$1672	$553	$24	$294	$481	$946
Laboratory	$424	$1085	$394	$315	$549	$429
DME	$461	$716	$361	$195	$380	$291
Advanced imaging	$115	$258	$177	$235	$249	$233
ASC	$297	$171	$205	$103	$158	$108
Rehab (PT/OT)	—	$39	$29	$49	$45	$268
Other	$1034	$3346	$1402	$1434	$2000	$984

ASC, ambulatory surgery center; DME, durable medical equipment; ED, emergency department; OT, occupational therapy; PT, physical therapy; SNF, skilled nursing facility.

Discussion

Much of the health care reform dialogue has focused on payment reform rather than delivery system reform, yet actual improvement in health care outcomes and reductions in the cost are driven by changes in the way health care is delivered day to day. Certainly policy makers and payers must ultimately link new payment models to innovations in care that demonstrably achieve the Institute for Healthcare Improvement Triple Aim. Because a population can be sectioned into subgroups based on medical, behavioral, economic, cultural, and other factors affecting their health and needs for care, an approach focusing on unique patient populations and organizing resources appropriately may be the most effective ways to actually improve care delivery.

Precedents exist for each of the models Cornerstone employed and many of these have generated similar results in individual trials. However, Cornerstone has employed them together in an integrated way and prospectively enrolled patients into each one according to their predicted need. As will be discussed, the results are directionally compatible with these previous reports, lending some credence to this study's admittedly preliminary conclusions. A strength of this approach is to more precisely target the resources employed to the needs of specific patients, thus facilitating both potential reductions in the costs of care and improved quality.

For Cornerstone, the most striking thing about the preliminary results of the primary care programs was the difference in performance between the fully implemented locations and the locations that remained tied to the FFS model. The 2 fully implemented locations saved nearly 14% of payer costs compared to the previous year, while the 2 less-than-fully implemented locations observed a cost increase of approximately 18%. The difference is even more striking when hospitalization costs are examined; the fully implemented locations reduced hospitalization costs by 45%, whereas the other 2 locations actually increased them by 20%.

It should be noted that all of the primary care facilities in the transformation program had already received NCQA's level-3 PCMH designation. The 2 less-than-fully implemented locations used a bolt-on solution, adding the services of patient navigators alone to the care team. Yet the physicians quickly reverted to productivity-based care, and the rest of the model was never deployed. The sample size for the primary care focused care transformation programs is small; however, the study team comes away with the belief that precisely targeted resources focused on specific high-need patient populations may drive significant sustained improvement in controlling unnecessary high-cost utilization above what Cornerstone had already achieved as a low-cost provider in the region with all of its primary care practices designated PCMHs.³

This care model redesign approach is not limited to the specific models initially developed. Additional care models were developed subsequently, including a nephrology medical home, a care model for chronic obstructive pulmonary disease (COPD) patients, new care models for patients with diabetes in an endocrinology clinic, a memory clinic focused on a family-centered approach to care for patients with dementia, and a pharmacy hub designed to provide pharmacist-led comprehensive medication management. If similar results can be demonstrated with care models for these other patient populations, the likelihood of broad generalization of this methodology could have substantially positive impact for the health care delivery system. Further analyses that study the costs of these care models relative to the improvement in total costs and outcomes will permit improvement in payment models that can further accelerate innovation in care redesign.

Limitations

The information presented here should be understood as a case study rather than a more statistically rigorous standardized study. The study team is retrospectively reporting pre- and postintervention results in clinical settings using a statistically validated standard risk and cost prediction tool. The more rigorous statistical study methodologies that might be possible in an academic, grant-supported setting simply were not feasible in this private practice setting. The innovations necessary for delivery system transformation will not occur in randomized controlled study settings alone, but in practices across the spectrum of the delivery system, so the team believes that reporting this experience has value for the wider health care system as all stakeholders learn from each other what does and does not work to improve cost, quality, and outcomes in patient care delivery.

There are a number of known data limitations. Paid claims data were not available for all individuals enrolled in the programs and these individuals were excluded from the analysis; the total claims history for most individuals was less than 3 years in duration; enrollment in the transformation programs was highly selective and occurred over an extended period of time; the enrollment periods for individuals ranged from 3 months to 12 months and required annualization for comparative analysis; and interventions were not standardized.

Limitations to this case report are inherent in its real-world origin. The study team did not have access to data for commercial populations so the results may not be applicable for patients in those insurance categories. It is possible in the Medicare Advantage population of the study data that other interventions by the health plans contributed to the results, although this would not be the case for the MSSP population. Because of the limitations of the claims data used, the actual numbers studied were a relatively small sample, and limit the ability to provide fastidious statistical analysis. Clinical data science can challenge these results with Hawthorn effects, regression to the mean concerns, sample size concerns, and standard actuarial analyses, but the predictive cost tool for the specific patients with whom Cornerstone intervened has been a statistically validated tool used in the real-world payer community for many years, which should offer some assurance that this analysis has relevance.

Although this study is not able to demonstrate a direct causal relationship with these interventions, Cornerstone does have data from the same time period showing an overall improvement in cost of care for its patients in the MSSP and United Medicare Advantage programs. In 2014, Cornerstone received shared savings from the MSSP, with the 6 highest quality scores in the nation and fourth lowest cost basis that received shared savings. Cornerstone also received shared savings from the United Medicare Advantage and has been able to demonstrate a continued very lost cost of care with this payer. In 2015, Cornerstone was part of the CHESS MSSP. Preliminary subgroup data from that contract demonstrate that the cost of care for Cornerstone did not rise. Cornerstone is now participating in a NextGen ACO for 2016 and a full-risk Medicare Advantage contract where its preliminary performance is positive. In the analysis of this case report the study team did not report clinical results that support why Cornerstone saw utilization drop, but some of that analysis has been done for these patients and other care models that Cornerstone is designing. Examples include a significant reversal of Grade 3 chronic kidney disease in patients in the CLC, a decrease in CHF exacerbations in the heart function clinic, and a decrease in COPD patient readmission rate from 12% to 6% after implementation of the COPD care model.

Cornerstone's core approach has been to proactively identify patients who would benefit from new care models. The fact that costs for these patients have been effectively reduced persuades the study team that the patients enrolled in the programs met that criterion. But how many additional patents could have benefited similarly? This is unknown. Cornerstone has advanced considerably in the ability to model and predict the trajectory of patient health status and costs, but much remains to be learned. Similarly, if the organization is to effectively assess value-based health care, it must be better able to predict the costs patients would have generated under conventional health care. Such a system seems to be part of the Centers for Medicare & Medicaid Services' plans for the Shared Savings Program, and the study team encourages the agency to bear in mind the needs of individual delivery networks and health plans as it assembles data for its reimbursement system.

Challenges

It is important to stress that this study's results reflect a real-world launch, with a host of decisions and compromises. Cornerstone did not have risk-sharing contracts with all of itsr payers, much less capitation-based payment models. This imposed some limitations on the services that could be offered. For example, transportation could not be provided to patients in need of it (an important feature of several value-based care delivery models). Virtual medicine could not be utilized to the extent the organization would have liked; nor could co-pays be eliminated for office visits, specialty referrals, and mental health care—an important step in aligning patient incentives and behavior. The data the study team worked with was incomplete and often too old to be very useful. The efforts to build a new culture achieved respectable results for a single year; but they need to go further. Also inefficiencies were experienced with lower volume clinics in the start-up phase, yet Cornerstone was still able to achieve savings during the ramp-up period. The organization continues to evolve the care models as experience teaches what interventions work.

Conclusion

This experience in the first year further persuades Cornerstone of one of its starting-point principles: the migration of the payment model from fee-for-service to fee-for-value is necessary, but insufficient. Payment reform must be accompanied by material changes in how care is delivered. Cornerstone believes that its focused care model transformation approach can serve as a powerful initial step for organizations embracing population health.

Footnotes

Author Disclosure Statement

The authors declared no conflicts of interest with respect to the research, authorship, and/or publication of this article. The authors received no financial support for the research, authorship, and/or publication of this article.

References

Berwick

, Nolan

, Whittington

. The triple aim: care, health and cost. Health Aff (Millwood), 2008; 27:759–769.

Oberlander

, Laugesen

. Leap of faith—medicare's new physician payment system. N Engl J Med, 2015; 373:1185–1187.

Friedberg

, Schneider

, Rosenthal

, Volpp

, Werner

. Association between participation in a multipayer medical home intervention and changes in quality, utilization, and costs of care. JAMA, 2014; 311:815–825.

Friedberg

, Rosenthal

, Werner

, Volpp

, Schneider

. Effects of a medical home and shared savings intervention on quality and utilization of care. JAMA Intern Med, 2015; 175:1362–1368.

Jackson

, Wilson

Jr . Does PCMH “Work”?—The need to use implementation science to make sense of conflicting results. JAMA Intern Med, 2015; 175:1369–1370.

Xing

, Goehring

, Mancuso

. Care coordination for Washington State Medicaid enrollees reduced inpatient hospital costs. Health Aff (Millwood), 2015; 34:653–661.

Durand

, Lewin

, Berkowitz

. Medical-imaging stewardship in the accountable care era. N Engl J Med, 2015; 373:1691–1693.

Shaw

, McDuffie

, Hendrix

, Edie

, Lindsey-Davis

, Williams

Jr . Effects of nurse-managed protocols in the outpatient management of adults with chronic conditions. Washington, DC: Department of Veterans Affairs, 2013.

Sprandio

J D

. Oncology patient-centered medical home. J Oncol Pract, 2012; 8:47s–49s.

10.

Heidenreich

, Albert

, Allen

, et al. Forecasting the impact of heart failure in the United States: a policy statement from the American Heart Association. Circ Heart Fail, 2013; 6:606–619.

11.

Greiner

, Hammill

, Fonarow

, et al. Predicting costs among Medicare beneficiaries with heart failure. Am J Cardiol, 2012; 109:705–711.

12.

Bekelman

, Plomondon

, Carey

, et al. Primary results of the patient-centered disease management (PCDM) for heart failure study. JAMA Intern Med, 2015; 175:725–732.

13.

Doddamani

, Newman

, Adams

, Berger

, Graf

. An integrated approach to comprehensive heart failure care. Chronic Dis Intr, 2014; 1:2.

14.

Medicare Payment Advisory Commission and Medicaid and CHIP Payment and Access Commission. Beneficiaries Dually Eligible for Medicare and Medicaid: Data Book. 2015. www.medpac.gov/docs/default-source/data-book/january-2015-medpac-and-macpac-data-book-beneficiaries-dually-eligible-for-medicare-and-medicaid.pdf?sfvrsn=0. Last accessed September 1, 2016 .

15.

Gilbody

, Bower

, Fletcher

, Richards

, Sutton

. Collaborative care for depression: a cumulative meta-analysis and review of longer-term outcomes. Arch Intern Med, 2006; 166:2314–2321.

16.

Kroenke

, Spitzer

, Williams

. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med, 2001; 16:606–613.

17.

Hegel

, Imming

, Cyr-Provost

, Hitchcock-Noel

, Arean

, Unutzer

. Role of allied behavioral health professionals in a collaborative stepped care treatment model for depression in primary care: project IMPACT. Families Systems Health, 2002; 20:265–277.

18.

Arean

, Hegel

, Vannoy

, Fan

, Unuzter

. Effectiveness of problem-solving therapy for older, primary care patients with depression: results from the IMPACT project. Gerontologist, 2008; 48:311–323.

19.

Unützer

, Katon

, Callahan

, et al. Collaborative care management of late-life depression in the primary care setting: a randomized controlled trial. JAMA, 2002; 288:2836–2845.

20.

Health Dialog. Rite Aid Health Alliance. www.healthdialog.com/solutions/rite-aid-health-alliance. Last accessed September 1, 2016 .

21.

US Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics. Health, United States, 2013: With Special Feature on Prescription Drugs. 2014. www.cdc.gov/nchs/data/hus/hus13.pdf. Last accessed September 1, 2016 .