The investigators propose an innovative new model of care in which patients identified to be at high risk of hospitalization are offered care by a physician who will direct their care both in the hospital and in clinic but is able to do so because they see patients only at high risk of hospitalization. This allows these physicians to have a panel of patients that is small enough that they can provide them with continuing ambulatory care but sick enough for those physicians to have enough of their patients hospitalized at any time to justify having the physician spend several hours each morning seeing those patients in the hospital, making the model economically viable and clinically valuable for the patient. The investigators estimate that each of the 5 physicians the investigators propose to establish in this model will serve a panel of about 200 patients in steady state with an average of 10 days of expected hospitalization and $75,000 each in Medicare spending per year, totaling $75 million annually. The investigators estimate that a 1% reduction in costs for these patients will be more than enough to cover the ongoing costs of the model the investigators propose; this is because the investigators' program reorganizes care rather than adding new forms of care.

The investigators' project has 5 aims:

Aim 1: To implement an innovative Comprehensive Care Physician (CCP) Model of multi-disciplinary team-based care at the University of Chicago Medical Center (UCMC) in which care for patients at high risk of hospitalization is led by CCPs who focus their practices on patients at high risk of hospitalization so that they can personally care for these patients both in clinic and in the hospital, with savings to Medicare shared with the AMC.

As noted above and supported by theory and data below, the key justification for this goal is based on the value of the doctor-patient relationship. Building on 15 years experience developing our hospitalist program (IRB protocol 9967) from 2 to 30 clinicians, the investigators provide detailed plans to efficiently implement the investigators' model to improve care within 6 months. To incentivize and assess cost-savings, the investigators propose:

Aim 2: Among patients who meet clinical eligibility criteria for the CCP model and express willingness to receive care in the model, to assess the utilization and quality of health care, health outcomes, and cost of care of patients randomly assigned to be offered CCP care compared to patients randomly assigned not to be offered CCP care.

Evaluation is important in the investigators' project both to ensure that the investigators meet the triple-aim and because assessment of cost savings is needed to implement the shared savings model that the investigators think is important to incentivize cost savings. The investigators recognize that the use of randomization to assess out-comes cannot be the basis for a long-term payment model in Medicare but select this approach to evaluation because the investigators think it will provide the most accurate estimate of the effects of the program. Assuming capitated payments to Accountable Care Organizations (ACOs) become a major payment approach for Medicare, cost savings from CCPs would create sustainable incentives for their use. CCPs might also work well in ACOs because selection of high risk patients into a CCP program might not be a concern if those patient came from within the ACO's covered population. Sustained incentives to use CCPs could also exist if high-risk persons selected into ACOs with CCPs, but risk adjustment was adequate. Shared savings programs with adequate risk adjustment would produce similar incentives. To address dissemination and risk adjustment, the investigators propose:

Aim 3: To use data collected by UHC from 4 Chicago-area AMCs with linked Medicare data to support evaluation and potential local and national dissemination of the CCP model

UHC (previously University HealthSystem Consortium) includes over 100 AMCs nation-wide and has exceptional data on hospital resource use to support risk-adjustment. UHC also houses laboratory and Medicare data to enhance risk-adjustment and outcomes assessment. The investigators will use this data to inform potential local dissemination and testing of our model by helping the other participating Chicago-area AMCs identify the segments of their patient populations that are frequently admitted enough to be suitable for the CCP model. If the model is successful at UCMC, and perhaps in later local dissemination and evaluation, UHC's national scope creates opportunities for dissemination of the investigators' work to almost all US AMCs. The risk-adjustment data will also allow us to evaluate alternative strategies to assess program savings to estimate shared savings that do not require randomization. Thus the investigators propose:

Aim 4: To compare the results obtained in Aim 2 to three alternative assessment strategies that: 1) compare patients in UCMC treated in the CCP model to patients in UCMC who meet CCP clinical eligibility criteria but receive usual care, 2) compare patients in UCMC who meet CCP clinical eligibility criteria to historical patients in UCMC who met CCP clinical eligibility criteria, 3) compare patients in UCMC eligible for the CCP model to current patients in 4 other Chicago-area AMCs who meet CCP clinical eligibility criteria.

The investigators propose to evaluate these alternative approaches for selection of a control group because the investigators think they might be feasible approaches to estimate savings in a shared-savings model and because they differ in potential threats to validity (e.g., 1 must address patient selection into the intervention and 2 must address time trends in costs), outcomes assessment (patient outcomes will be harder to assess in 2 and 3), and generalizability (1 requires local patients treated outside the model and 2 requires historical data, which becomes less relevant over time). All models will be adjusted with the CMS-Hierarchical Condition Category (HCC) risk adjustment model for our base analysis but the investigators will also explore other approaches to risk adjustment, such as using prior year total Medicare expenditures and patient-reported variables as risk adjusters. The investigators propose to use our randomized control group for our shared savings incentive, but to work with CMMI in our cooperative agreement to assess how these control groups and approaches to risk adjustment could be used in shared savings incentives. Patient-reported outcomes will be assessed from enrolled subjects. Finally, since workforce development is key to implementing and disseminating this model, the investigators propose:

Aim 5: To develop skills of a multidisciplinary team of providers to execute the CCP model, and expose medical students, residents and fellows, and other health professions trainees to the model.

Because the investigators are an AMC with a rich pipeline of health professions trainees and excellent record of training national leaders, the investigators are excellently situated to train clinicians in the investigators' model who have the potential to help disseminate its core principles both locally and nationally.

Inclusion Criteria:

• Medicare recipients and were hospitalized at least one time in the past year

Exclusion Criteria:

• Non-Medicare recipients and/or were not hospitalized in the past year