Guidance for Developing Payment Models for COMPASS Collaborative Care Management for Depression and Diabetes and/or Cardiovascular Disease Introduction Within the COMPASS (Care Of Mental, Physical, And Substance use Syndromes) collaborative care model, we have shown improvements in clinical care over baseline care. As part of our cooperative agreement, we are to propose payment model(s). This guide explains why existing fee-for-service billing codes are not sufficient or appropriate for collaborative care management services, such as developed in the COMPASS model. We offer principles and examples for how to customize shared savings or a per-member-permonth care management payment with a pay-for-performance component to support collaborative care management services. We selected these two payment models, because they are feasible in the current environment and are aligned with the national movement towards linking fee-for-service to quality and creating alternative payment models. These principles are intended to be used by medical groups and private and public payers to inform the development of their payment models. We limited the guiding principles for each payment model to one to two pages for ease of use. Also, we recognize that the details will depend on the local environment and will be negotiated between medical groups and payers. We also provide our current understanding of the impact of COMPASS on costs both costs of implementation and healthcare services. From the background on current payment models and current financial impact, we provide some conclusions. Page 1 of 11
Why Traditional Fee-For-Service Payment Models that are Limited to Existing CPT Billing Codes Do Not Support Collaborative Care Management Although medical groups may want to consider the billing codes listed below when thinking about how to sustain collaborative care management, there are significant limitations: There is not one billing code that perfectly fits the evidence-based bundle of collaborative care management services in COMPASS. The below list of existing codes can be knitted together and then tested, but relying on the existing CPT codes forces organizations to adapt the healthcare delivery model to fit the CPT codes as opposed to adhering to the fidelity of the evidence-based model. The basic challenge is that the prescriptive CPT codes are designed for specific, discrete services. Different payers have different requirements for which provider types and type of primary care settings are authorized to bill for the discrete, prescriptive service. The G0444 billing code for annual depression screening does not cover the evidencebased portion of the United States Preventive Services Task Force s (USPSTF) recommendation that follows the portion about screening: staff-assisted depression care supports to assure accurate diagnosis, effective treatment, and follow-up. This is concerning, because depression screening alone is not effective nor recommended. And even though the Affordable Care Act (ACA) states the preventive services must be recommended by a physician or other licensed practitioner within the scope of their practice under State law and a 2013 regulatory change by CMS states that State Medicaid Offices can submit a State Plan Amendment to allow non-licensed practitioners to provide the USPSTF-recommended services if recommended by a physician or other licensed practitioner, most state Medicaid offices have not submitted these amendments. The complexity of the fee-for-service billing system means that each medical group would need to: (1) establish a list of potential billing codes; (2) determine which codes (if any) are reimbursed by their specific payers and which provider types are authorized to bill each payer for the service; (3) create a reimbursement grid that identifies which services are reimbursable by each payer and provider type; and (4) create a new billing workflow to implement and test the strategy devised in the reimbursement grid. However, even if each organization goes through these intensive and time-consuming steps, it is highly likely that the revenue generated will not be sufficient to sustain the costs of COMPASS collaborative care management, and key components will be completely un-funded, including the systematic case review meetings. Thus, to sustain evidence-based behavioral healthcare services, we need a payment model that: does not solely rely on the traditional fee-for-service billing codes, helps drive organizations and providers toward the desired outcomes, is appropriately risk-adjusted to prevent cream skimming and adverse selection, and allows primary care teams to decide how to best configure and train their team to adhere to the fidelity of the evidence-based care delivery model as opposed to a billing code. Page 2 of 11
Examples of Billing Codes for Consideration: Physician Office Visits for Evaluation and Management o 99201 through 99205 for initial visits with new patients and 99211-5 for followup visits with established patients, depending on the service complexity and duration Telephone Evaluation and Management o 98966 through 98968 for non-physicians or 99441 through 99443 for physicians, depending on duration Online Evaluation and Management o 98969 for non-physicians or 99444 for physicians Office Consultation o 99241 through 99245 Medical Team Conference o 99366 through 99368 Annual Depression Screening o G0444 Health and Behavior Assessment and Intervention o 96150 through 96155 CMS pays for non-face-to-face care management services (99490) provided to Medicare beneficiaries with two or more chronic conditions that last at least 12 months or until the death of the patient and that place the patient at significant risk of death, acute exacerbation/decompensation, or functional decline. The rate is for at least 20 minutes of non-face-to-face care management services in the 30-day billing period. Page 3 of 11
Principles for Shared Savings Payment Models that Support Collaborative Care Management 1. Include the following set of quality measures to support evidence-based collaborative care management models. We stress the importance and value of using these measures as a single set of measures. Depression Remission at Six Months: Adult patients age 18 and older with major depression or dysthymia and an initial PHQ-9 score > 9 who demonstrate remission at six months defined as PHQ-9 score less than 5 (NQF 0711). Depression Improvement on or after 120 Days: Adult patients age 18 and older with major depression or dysthymia and an initial PHQ-9 > 9 that occurred at least 120 days ago whose most recent PHQ-9 is 5 or more points less than the initial PHQ-9 or whose most recent PHQ-9 < 10. Screening for Clinical Depression and Follow-Up Plan: Percentage of patients aged 12 years and older screened for clinical depression on the date of the encounter using an age appropriate standardized depression screening tool, and, if positive, a follow-up plan is documented on the date of the positive screen (NQF 0418 and PQRS 134). Shared Savings contracts that do not include the depression remission or improvement measure with the depression screening measure run the risk of paying for services that have not been shown to improve quality or contain costs. Although the USPSTF recommends depression screening, they only recommend depression screening when staff-assisted depression care supports are in place to assure accurate diagnosis, effective treatment, and follow-up. This clause is referring to the over 70 randomized controlled trials that support collaborative care management for depression. 2. Include the depression remission measure as a pay-for-reporting measure during the first three years and then as a pay-for-performance measure. The medical groups will need time to implement an infrastructure for systematic follow-up care and PHQ-9 re-measurement, which are necessary for the depression remission measure. 3. Follow Medicare s method to calculate quality scores in recognition that shared savings are intended to support overall quality of care (including but not limited to depression). This method: (1) enables ACOs to earn quality points for each measure on a sliding scale based on the percentile level of performance, (2) totals the points earned for measures in each domain, (3) divides by the total points available for that domain to produce an overall domain percentage score, and (4) averages the percentage score for each domain to generate an overall quality score to assess the amount of shared savings or losses. 4. Include an advanced payment model in the shared savings contract, especially for smaller accountable care organizations (ACO). For example, Medicare s Advanced Payment ACO Model provides an upfront, variable payment and monthly payments depending on the ACO s size to enable small ACOs to invest in the infrastructure that is needed to bear financial risk and provide population-based care. Also, Medicare s Next Generation ACO Payment Model includes four alternative payment mechanisms, and one of the four methods includes Page 4 of 11
traditional fee-for-service plus a $6 per-member-per-month payment that will be fully retrieved by CMS during the reconciliation process. Principles for Care Management Per-Member-Per Month Payment and Pay-For-Performance Components that Support Collaborative Care Management Experience to Date and Promising Results in Payment Strategies To date, large-scale implementations of collaborative care management have used varied payment approaches. Strategies have ranged from per-member-per-month, fully capitated payments, as used in Kaiser Permanente and Veterans Administration services, to case-rate payments that augment other primary care billing, as used in Minnesota s DIAMOND program. DIAMOND In the DIAMOND program, a per-member-per-month care management fee covers the psychiatric caseload review and the care management services as a bundled set of services. Payment for the monthly fee is contingent on being trained and certified as a DIAMOND medical group with standardized processes and training. MHIP Pay-for-performance strategies have been implemented on a limited basis but with promising results. In Washington State Mental Health Integration Program (MHIP), care managers are funded via a monthly payment to support a caseload. MHIP introduced a pay-for-performance component in 2008, in which 25 percent of the total payment is tied to effective treatment. Performance is assessed on a number of quality indicators including timely follow-up with patients, demonstration of improved patient outcomes, and systematic consultation and treatment adjustment for patients who are not improving. Since the pay-for-performance component was introduced, the effectiveness of the program has substantially improved. Figure 1 below shows significant reductions in median time-toimprovement, based on a study of 8,000 depressed adults served in 29 community health clinics participating in MHIP after pay-for-performance was implemented. The population in this study included patients covered by Medicaid and patients who were not insured. Page 5 of 11
New York State Medicaid More recently, the New York State Medicaid program has planned and implemented a supplemental monthly case rate Medicaid payment for collaborative care management of depression. Essential elements of New York State s program requirements include: trained and state-certified depression care managers; primary care providers trained in screening and collaborative, stepped care; use of a state-approved patient care registry; and designated psychiatrists who provide caseload supervision and consultation and have access to the care registry. Similar to the DIAMOND program, the New York State Medicaid supplemental monthly case rate is intended to cover the both collaborative care management services and psychiatric caseload review and consultation. Psychiatrists will provide at least weekly caseload-focused consultation based on the registry either on-site or from a remote location via phone or video. The New York State Collaborative Care Depression Program is also developing a pay-forperformance component. Principles for Per-Member-Per Month Payment and Pay-For-Performance Components Per-member-per month payment strategies for COMPASS and other collaborative care management models should cover two key components that are generally not effectively reimbursable under usual fee-for-service payment models:! collaborative care management services and! caseload review and consultation. Psychiatric and other case review functions that do not involve face-to-face patient contact could be reimbursed on a per-patient basis. However, it is more efficient and scalable to bill case Page 6 of 11
review and consultation services as a component of a bundled monthly rate that is a function of the size of the patient caseload that is supervised. Pay-for-performance strategies should be considered as a component of any case rate payment model for collaborative care management, in ways that reinforce overall program goals to improve patient outcomes on a timely and efficient basis. Pay-for-performance elements may vary over time, reflecting the phase and maturity of the program as well as primary clinical concerns of the target patient population. Specific learnings from Washington State s MHIP suggest the following key principles of payfor-performance models: Develop measurement and transparency collaboratively: o The pay-for-performance measures were initially developed in 2008 by a Steering Committee, which included representation from the payor, community partners, and Community Health Centers (CHCs) participating in the MHIP. Incorporating feedback from participating CHCs early on in the process of defining measurements supported and encouraged buy-in from providers and executive directors. Additionally, real-time, interactive reports on performance were made available to participating CHCs, which could be aggregated by individual provider, clinic, and CHC to facilitate self-monitoring and data transparency. CHC-level performance was shared across all participating CHCs to encourage sharing of best practices, which aided achievement of measures and reduced barriers. Re-evaluate and adjust measurement: o Initial targets for each measure were reevaluated on an annual basis to determine ongoing effectiveness. If the majority of participating CHCs exceeded the target for a specific performance measure in a given year, the target was reevaluated and incremental adjustments made as appropriate. For example, once the majority of participating CHCs exceeded the target for demonstration of improved patient outcomes, the target was increased to drive performance towards even better outcomes. Measures were also monitored on a regular basis to ensure they were defined appropriately and were meaningful to improving quality of care. In scenarios where either or both of these criteria were not met, the measurement was adjusted or retired. Use measurement to improve value: Initially, the pay-for-performance component accounted for 25% of the MHIP funding. However, in subsequent years, the portion of funding tied to performance increased significantly. This reinforced the need for measurement transparency and the provision of ongoing technical assistance for participating CHCs to maintain buy-in and promote the achievability of the performance measures. Page 7 of 11
COMPASS Financial Impact We assessed the financial impact of COMPASS care on three dimensions: the costs of preparing clinics to deliver COMPASS care (preparation costs); the costs of providing COMPASS care to patients (delivery costs); and the impact of COMPASS on patient s medical care costs (medical costs). We briefly describe each and then assess the net financial impact of COMPASS by bringing the results of all three financial dimensions together. Preparation Costs Eight partner sites were asked to self-report the costs associated with preparing to deliver COMPASS care using a provided template by the evaluation team. The template asked for employee hours and expenditures in defined categories as shown in Table 1. Partners were asked to respectively complete the template after the preparation activities were complete and to estimate costs over the period from which they first committed to participating in COMPASS (but no sooner than Aug 1, 2012) through the first 3 months of providing COMPASS care (but no later than July 31 st ). Thus, the period for reporting was all allowed to vary within a 12-month window to reflect partners actual time period for preparing their delivery systems for COMPASS care and making adjustments to delivery systems through the first 3 months of providing care. Because many preparation costs were at the partner-level rather than the care-site level, preparation costs were reported by partner not by care site. The average cost across all partners was $302,000, with median, min and maximum total costs of $236,000, $38,000, and $670,000, respectively. Preparing for new roles and adjusting the care delivery system each accounted 35% of these costs followed by strategic decision making and coordination. Over the long-run, these costs could be spread over many thousands of patients, but spread only over the 3,912 patients enrolled by the end of the grant period, the preparation costs averaged $617 per patient enrolled. Table 1. COMPASS preparation costs Average of eight partners ($) Average as percent of total Median ($) Lowest in expenditure category ($) Highest in expenditure category ($) Defining and hiring for new positions 105,000 35% 65,000 3,000 412,000 Preparing the delivery system, including planning work flow and training existing staff 74,000 25% 51,000 14,000 190,000 Creation or adopting a registry and other information technology 34,000 11% 20,000 0 144,000 Strategic decision making and coordination 83,000 27% 42,000 0 204,000 Other 6,000 2% 3,000 0 21,000 Total 302,000 100% 236,000 38,000 669,000 Care Delivery Costs The partners were also asked to provide a one-month snap shot of care delivery costs during the month of June 2014 using a similar template with cost categories shown in Table 2. We Page 8 of 11
requested care delivery costs by care site, however, only three of eight partners were able to disaggregate costs to individual care sites. Therefore, care delivery costs are also presented at the partner-level. We divide care delivery cost by the number of patients enrolled in COMPASS care during the month, because delivery costs are more strongly associated with the number of patients cared for in comparison to preparation costs. The average one-month per-patient cost across all partners was $409, with mean, min and maximum total costs of $404, $220, and $659, respectively. Care manager time accounted for about half of these costs. Consulting physician time, registry and IT, and strategic decision making & care coordination each accounted for 10%-13%. With an average of 6 months of enrollment in COMPASS care, costs were higher than anticipated at $2,400 per patient. Table 2. COMPASS care delivery per patient Medical Costs Average of eight partners ($) Average as percent of total Median ($) Lowest in expenditure category ($) Highest in expenditure category ($) Care manager time 213 52% 211 89 448 Consulting physician time 56 13% 35 14 132 Registry maintenance and other information technology 15 5% 14 4 33 Ongoing decision making & system care coordination 59 13% 23 0 316 Supervision 41 10% 13 0 135 Other staff costs 17 3% 0 0 90 Other significant costs 8 3% 0 0 62 Total 409 100% 404 220 659 We collected self-reported counts of emergency department visits (ED), inpatient stays (IP), outpatient visits (OP) and prescription drugs for 12 months pre and 12 months post COMPASS enrollment using patient phone surveys. Three of eight partners were also able to provide utilization data with near complete capture directly from claims or other accounting databases. We had planned to assess self-report bias by comparing self-reported utilization to observed utilization in these three sites, and to use the results to correct self-reported data from patients of all eight partners. Complete or partially complete surveys were obtained for 665 COMPASS enrollees and administrative data provided some utilization data on 1,324 patients across the three sites able to provide such data. However, only 110 patients met the following four criteria necessary to compare self-reported utilization and utilization from administrative data: complete or partially complete baseline survey; complete or partially complete 12-month follow-up survey; 12 months of pre-enrollment in administrative data; and 12 months post-enrollment in administrative data. This sample was not large enough in number to obtain reliable estimates of self-report bias for the costly but less common utilization categories of ED visits and IP stays. Therefore, we report results from the survey without correction for self-report bias with the expectation that self-report bias is similar pre-enrollment and post-enrollment. Page 9 of 11
We compared utilization among patients with a completed baseline survey, and we compared pre- and post-enrollment in an intention to treat analysis (ITT) by carrying forward self-reported utilization from the pre period to the post period for patients who did not have a complete follow-up survey. Table 3 shows the difference in utilization between the 12 months pre and post. Emergency department visits and inpatient stays decreased, but only the decrease in ED visits is statistically significant. Outpatient visits and prescription drug use increased, but neither increase was statistically significant. We estimated cost for each type of utilization as the average paid amounts among Medicare enrollees from the Medical Expenditure Panel Survey. We included amounts paid by all payers for Medicare enrollees. The costs of ED visits and inpatient stays include payments to both facilities and physicians. Using these paid amounts, we estimate that medical costs decline by $682 per patient in the first 12 months. When this reduction is extrapolated to all 3,912 patients enrolled before the end of the grant, medical costs would be reduced by $2,729,108. Table 3. Change in per-person encounters and costs by type of care over 12 months post-compass enrollment compared to 12 pre-enrollment, intention to treat (ITT) analysis Change in utilization counts per person among survey respondents Change in medical costs** per person among survey respondents Predicted change in medical costs** for all COMPASS patients n = 665 n = 4,000 Emergency Department -0.158* -$137 -$546,203 Inpatient Stays -0.042 -$595 -$2,379,061 Outpatient Visits 0.183 $42 $169,745 Prescription Drugs 0.081 $7 $26,411 All utilization 0.063 -$682 -$2,729,108 *Statistically significant at.05 level or greater based upon Wilcoxon signed rank test. **Based upon change in encounters from ITT analysis as presented in Table 2 and average cost per utilization type for Medicare enrollees in the Medical Expenditure Panel Survey, all sources of payment (Emergency Department = $863, Inpatient = $14,186, Outpatient = $232, Prescription drug = $82). We also summarized pre- and post-utilization among the patients (N=993) from the three partners that were able to provide administrative utilization data. To use the largest sample, we also employed an ITT analysis by carrying forward pre-enrollment counts to the post enrolment period for any patient who did not have 12 months of follow-up eligibility in administrative data. Rather than savings, these data show an increase cost of $204 per patient. There is a trend towards a decrease in inpatient stays, but only the increase in outpatient visits is statistically Page 10 of 11
significant. In the sample of patients that represent only 25% of COMPASS enrollees, 95% are from two implementation partners that may not be representative of the other sites. Summary of financial impact The total cost of providing COMPASS care, including both preparation and care delivery costs, would exceed $2,500 per patient, even when preparation costs are spread over patients that would be enrolled over ten years. If medical cost savings continue to be realized at the rate estimated by ITT analysis of survey data in the first 12 months ($682 per patient), then four years would be required to offset the full costs of COMPASS care. This estimate of savings may be conservative because the ITT analyses produce conservative results and because greater savings have been observed in subsequent follow-up years in collaborative care experiments. At the same time, our evaluation sample size was not sufficient to show statistically significant reductions in inpatient stays during this time period, and administrative data for patients from three sites indicate that medical care costs may be higher in the first year rather than lower. Conclusions Although it would be helpful to have additional data and more time to determine the full impact of COMPASS on cost, there is enough known about the limitations of the existing billing codes in fee-for-service and promising alternative payment models to make recommendations that a shared savings model or care management per-member-per month payment with pay-forperformance components can be used to support the provision of the evidence-based components of collaborative care management. An important caveat would be that patients may need to be in the same insurance product and/or provider for the ROI to be cost neutral and/or eventually cost savings. Important features of a new payment models for collaborative care management include tying a significant portion of the payment to performance and bundling the core components of the evidence-based collaborative care management model. To provide more insight into costs impacts, we recommend longer periods of data collection and larger samples. September 2015 Page 11 of 11