Medicare Underpayment for Diabetes Prevention Program: Implications for DPP Suppliers
Amanda S. Parsons, MD; Varna Raman, MBA; Bronwyn Starr, MPH; Mark Zezza, PhD; and Colin D. Rehm, PhD
To date, one of the most successful and scalable interventions to reduce diabetes among high-risk individuals is the Diabetes Prevention Program (DPP). In a randomized controlled trial, the DPP lifestyle intervention, which emphasizes modest weight loss and increased physical activity, significantly reduced the risk of diabetes.1
In its original form, DPP could not be sustainably scaled by community organizations.2
The trial included 1-on-1 coaching with a healthcare professional, supervised exercise classes, and substantial monetary study incentives and tools (eg, grocery vouchers). To reduce program costs, estimated at approximately $1400 per participant at the time, several changes were made. Group sessions replaced 1-on-1 sessions and the community became the primary place of delivery.3
Additionally, incentives were removed and physical activity shifted from private/coach-led to memberships at local fitness facilities or simply by encouraging physical activity. These changes reduced the intensity of the intervention while slightly affecting efficacy and became the basis for the CDC’s National DPP (NDPP). A recent study found that among more than 10,000 participants in NDPP, the median weight loss was 4.2%, less than the 6% from the original trial.4
To date, more than 1750 organizations are seeking or have received recognition as an NDPP supplier.5
In 2016, CMS announced that Medicare would offer reimbursement for DPP beginning in 2018. In November 2017, a final rule detailing the payment structure was released.6
Because little is known about the actual costs of DPP, CMS considered a number of factors in its payment methodology, including looking at similar covered services. CMS referenced Medicare reimbursement of $10 per patient for 30 minutes of group education and training for patient self-management by a nonphysician health professional for 5 to 8 patients (Current Procedural Terminology code 98962).6
Although strides have been made to make DPP sustainable and scalable, implementing the program may be cost-prohibitive if reimbursement levels are insufficient. Therefore, it is important to understand the cost inputs of DPP to create adequate reimbursement and to help prospective suppliers understand the implementation costs. This study aims to describe the various cost inputs in a real-world large-scale DPP implementation and compare them with the CMS reimbursement rates using the experience of Montefiore Health System (MHS), a large healthcare system in Bronx, New York. MHS implemented DPP in partnership with the YMCA of Greater New York from 2011 to 2015 and on its own since 2015.7
To date, the Montefiore DPP has served more than 1350 patients.7,8
To estimate the costs of implementing the DPP in a single year (2016), we used a microcosting strategy. Key informant interviews, program data, and supply orders were used to estimate the costs of delivering DPP. First, we identified fixed costs associated with DPP delivery. These fixed costs included the annual salary of a full-time DPP coordinator. The coordinator is responsible for conducting outreach, confirming patient eligibility by briefly reviewing the electronic health record (EHR), providing administrative and clerical support to the program, and teaching some classes. Additional fixed costs were staff salaries, including 25% of the salary of a manager, 5% of that of a project manager, and 5% of a data analyst to manage data and CDC reporting.
Semivariable expenses included DPP class instruction, which depended on the job classification of the lifestyle coach (eg, health educator or community health worker) and whether the class was taught during normal working hours or not. For classes taught on weekends or evenings, coaches were given “session pay” (ie, overtime). Based on key informant interviews, each hour of classroom instruction required an additional 1.75 hours to remind patients about classes, travel to the class, set up the classroom, conduct make-up sessions, and document attendance, weight, and physical activity. For health educators, who taught a majority of classes, the cost for teaching plus the additional time was $94.48 per session.
Additional variable expenses included the printed materials for the core and maintenance curriculum guides ($22.34/unit and $19.92/unit, respectively). We accounted for the costs associated with facilitator guides, training DPP instructors, and the costs of clinical quality scales. The costs of becoming a DPP master trainer, which is required to train certified instructors, were also included. Lastly, the costs of incentives, such as pedometers, water bottles, and salad bowls, were calculated by reviewing previous orders.
For program revenue, we used data from the 2016 program year, when 322 patients began attending 1 of 22 class cycles. Attendance at each class was measured and average weight loss was calculated to estimate revenue per the CMS rules.7
The final revenue thresholds provided by CMS are described in Figure 1A
. Although about one-third of MHS’ DPP patients receive Medicare, we purposely calculated program revenue as if all patients were being reimbursed at the proposed CMS rates, in order to assess the programmatic implications of those rates. Up-front costs and cross-subsidies across the health system were not estimated; therefore, cost estimates should be interpreted as underestimates. Because reimbursement is dependent on outcomes, we conducted secondary analyses by calculating revenues assuming that our outcomes had been in line with those of average NDPP suppliers.4
Overall, the costs of implementing DPP at MHS were $177,976, or $553 per participant attending 1 or more sessions in 2016. Figure 1B
shows the breakdown of costs per participant. Twenty-eight percent of costs ($153/participant) were for direct instruction, whereas 24% were for telephonic outreach and orientation, as well as confirming patient eligibility ($133/participant). Filling a class with 15 to 20 participants requires placing about 30 in each class and reaching out to 60 to 70 eligible patients. The curriculum guide (5.1% of total costs), staff training and teaching guides (3.4%), and student incentives (1.7%) were modest costs. About 5.6% of the costs were for data management and reporting, and 17.4% were for direct supervision and staff management. An additional 15% of the costs were for other program coordination activities (eg, scheduling classes/rooms, sending materials to sites).
Based on our program experience, the average number of classes attended per patient was 6.4. Among all individuals, including those who attended only a single session and, therefore, had no follow-up weights, 14.3% of patients lost at least 5% of their initial weight. Fifty percent of participants attended 4 or more sessions and 33% attended 9 or more (Figure 2
). Based on these attendance and weight loss outcomes, MHS would receive $34,625, or $108 per patient, which would cover 19.5% of costs.
Additional analyses modeled reimbursement assuming that our outcomes had been aligned with national outcomes presented by CDC.4
If MHS had achieved these outcomes, the reimbursement would be $61,270, or $190 per patient, covering 34.4% of costs.
Based on our program experience, Medicare reimbursement would cover just 19.5% of program costs. This shortfall is only partially explained by the relatively poorer outcomes of our program. Compared with other DPP implementations, our program had poorer attendance, driven mostly by a large number of drop-outs between sessions 1 and 2. A recent evaluation of NDPP data found that the median number of sessions attended was 14 and that 86.6% attended 4 or more sessions.4
Comparatively, MHS’ mean attendance was 6.4 sessions and the median was 4, due to a large number of drop-outs after the first class (among those attending ≥4 sessions, median attendance was 12 sessions). Half of the patients dropped out before class 4 compared with 13.4% nationally.
These differences in attendance outcomes may be driven by geographic and/or demographic differences in attendance and/or differences in program delivery. For instance, some programs hold a “class zero” orientation session prior to the first session, which reduces early drop-outs. Implications for poorer attendance in certain patient populations mean that DPP suppliers who serve them have to expend additional resources conducting outreach, sending reminders, and providing meaningful incentives. It is critical to note that the intensity and length of DPP may not be feasible for many patients, including those with inconsistent work schedules and complex family demands. Given that nationally only one-third of patients reached the 5% weight loss threshold across all of the implementations, it is imperative that CMS recognize this in its pay-for-performance methodology. Even if our program had outcomes identical to the national CDC-reported data, there would still be a considerable shortfall in revenue of about $363 per patient (34.4% of program costs would be covered).
Notably, in these national data, Hispanic and non-Hispanic black participants had lower median attendance (11 and 13 sessions, respectively) than non-Hispanic white participants (16 sessions).4
Furthermore, outcomes among Hispanic (median weight loss of 3.0%) and non-Hispanic black (2.6%) participants were worse than for non-Hispanic whites (4.4%) nationally. At MHS, more than 90% of DPP participants are Hispanic or non-Hispanic black, compared with 23.8% nationally.4
This major demographic difference may be an additional explanation for the poorer outcomes among MHS participants.
It is important to quantify the true costs of delivering programs like DPP, which sit outside of the traditional clinical encounter. Program costs do not decrease when patients do not attend classes. Instructors are still paid, the encounter is still documented, materials are still printed, and the program is still managed. Therefore, the CMS payment, which underpays for both attendance and performance, will be insufficient for most, if not all, DPP suppliers. Furthermore, given the strong correlation between attendance and weight loss, the payment structure is paying for the same outcome twice. Lastly, so few patients achieve 9% weight loss that this payment benchmark has little utility.
DPP program costs are significant and most are not variable, because activities such as program coordination, outreach, EHR documentation, instruction, and reporting all occur whether or not participants attend classes. Placement- and instruction-related costs account for just over half (52%) of costs. Another 38% of costs are for program coordination, staff management, and data reporting; 9% are for instruction materials for coaches and patients; and 2% are for patient incentives. Because of significant fixed/semivariable program costs, DPP is not well suited for reimbursement based on a threshold number of classes or largely skewed toward pay-for-performance for weight loss thresholds that exceed common experience. Rather, a baseline of fee-for-service reimbursement on a per class basis with a potential pay-for-performance bonus for each percentage point of weight loss would align payments with costs. This payment structure also acknowledges the diabetes risk reduction with weight loss less than the 5% threshold.9
For our program to break even at current levels of performance with regard to attendance and weight loss, we would need $88.71 per patient per session attended. If we achieved the national outcomes, break-even revenue would be $42.28 per session attended.
It is important to analyze, document, and disseminate the costs associated with DPP implementation to inform prospective suppliers and payers and to advocate for appropriate reimbursement. As healthcare systems wrestle with the relative value of prevention versus treatment, it is important for NDPP suppliers to quantify outcomes beyond weight loss. Additional value can be demonstrated by documenting reductions in glycated hemoglobin, blood pressure, and/or cholesterol and, more importantly, long-term outcomes like diabetes incidence. It is important for CMS to consider the real-world costs of program delivery when setting reimbursement rates.
The authors wish to thank Nicole Hollingsworth, Melinda Marquez, and Elizabeth Spurrell-Huss for contributing their input and expertise to the estimation of program costs.