Medicare Spending Trend Is Not One Number: Part B, Part D, and MA Each Require Separate Actuarial Assumptions

The 2026 Medicare Trustees Report puts total Medicare benefit payments at $1.2 trillion in 2025, but that figure is doing a poor job of describing the pricing environment health actuaries face (KFF, June 2026). Part B owns 48% of spending, governed by the physician fee schedule and outpatient payment systems. Part D owns 15%, reshaped by the IRA redesign that shifted catastrophic-phase liability from the federal government to plans and now running toward $346 billion by 2035 on a 6.7% annual growth trajectory. The easiest way to miss a Medicare trend assumption is to blend these categories into a single PMPM and apply one adjustment factor.

Part B at 48%: Physician Fee Schedule Volatility and the Skin Substitute Problem

Part B's spending share has grown steadily as outpatient hospital care, physician-administered drugs, and advanced diagnostic imaging have taken volume from inpatient settings. The $481 billion in combined Part A and Part B traditional Medicare spending in 2025 sits behind a Part B sub-allocation where physician fee schedule services account for 15% of the Part A/B total, outpatient hospital services account for 16%, and inpatient hospital care accounts for 33% (KFF, June 2026). For MA plan pricing actuaries, Part B is the primary exposure because MA bids are benchmarked to per-capita FFS Part A/B spending by county, and Part B movements flow directly into benchmark rate changes.

The skin substitute episode is the sharpest demonstration of Part B volatility that health actuaries have seen in recent memory. Part B spending on skin substitute products rose from $252 million in 2019 to more than $10 billion in 2024, a nearly 40-fold increase in five years, driven almost entirely by pricing dynamics and launch prices on new skin substitute products rather than by volume growth in the underlying wound care population (CMS, 2025). CMS reset reimbursement at $127.28 per square centimeter in the CY 2026 physician fee schedule final rule, effective January 1, 2026, and projects gross FFS program savings of $19.6 billion in 2026 alone from that one policy change. Without the reset, the Part B monthly premium would have run approximately $11 per month higher than it did.

The mechanism matters for MA bids: benchmark rates incorporate both the direction of spending and the volatility of specific line items within Part B. A plan that built its 2025 bid on Part B trend estimates calibrated to the skin substitute trajectory would have been pricing against a spending base that CMS subsequently compressed by roughly 90%. The lesson is not that skin substitutes are a unique event. It is that Part B contains physician-administered drugs, infusion therapy, and outpatient procedure volumes that are each governed by payment policies with independent reform cycles. Medigap pricing is exposed to the same volatility, since Medigap wraps Part B cost-sharing on traditional Medicare and any acceleration in Part B outpatient spending flows through to claims without a prior authorization backstop.

Part D at 15%: What the IRA Redesign Did to Plan Liability

Part D spent $181 billion in 2025 (CMS Office of the Actuary, 2026). The Trustees project that figure reaches $346 billion by 2035, a 6.7% compounded annual growth rate that is nearly 2 percentage points faster than the 4.8% trajectory in last year's report. The upward revision is attributed explicitly to GLP-1 adoption and high-cost specialty drug pipeline growth. That acceleration happened on top of a structural redesign that fundamentally redistributed plan liability starting in 2025.

Under the IRA Part D restructuring, the catastrophic phase shifted from a regime where the federal government bore 80% of costs above the annual out-of-pocket threshold to one where plans bear 60%, manufacturers contribute 20% through the Manufacturer Discount Program, and CMS provides 20% reinsurance. The financial signal is direct: the national average monthly Part D plan bid jumped from $64.28 to $179.45 in 2025, a near-tripling, as plans priced in the catastrophic-phase liability they had previously ceded to the government (CMS, 2025). The annual out-of-pocket cap, $2,000 in 2025, moves to $2,100 in 2026 with inflation indexing thereafter.

For 2027 bid calibration, two additional pressures interact. First, CMS completed a second round of drug price negotiation under the IRA, placing Ozempic, Rybelsus, and Wegovy on the negotiation list with maximum fair prices effective 2027. The negotiated discounts reduce unit cost but bring GLP-1 medications into the "selected drug" category, where plan reinsurance from CMS is set at 40% rather than the standard 20% in the catastrophic phase. Second, the July 2026 GLP-1 Bridge program gives Medicare Part D enrollees access to GLP-1 medications at $50 per monthly supply through December 2027, compressing plan premium revenue on that population while the underlying utilization trend continues to accumulate. Health actuaries building 2027 bids need separate assumptions for selected-drug GLP-1 volume, non-selected specialty drug catastrophic exposure, and the revenue impact of the bridge program's below-market premium implication.

Decomposing the 14% MA Payment Premium Over FFS

Medicare Advantage now enrolls 35 million beneficiaries, 55% of all eligible Medicare participants, up 1.1 million year over year (KFF, February 2026). Total MA payments reached $534 billion in 2025, representing 53% of all Part A and Part B payments even though MA enrollment is roughly 55% of the eligible population. That arithmetic is the origin of the 14% per-enrollee payment premium that KFF and MedPAC have cited in successive reports.

MedPAC's March 2026 report to Congress puts the $76 billion overpayment figure alongside a clean decomposition: favorable selection accounts for approximately 11 percentage points of the excess, meaning MA enrollees spend about 11% less than FFS enrollees with similar risk scores would, and coding intensity accounts for approximately 4 percentage points, meaning MA risk scores run about 4% above what the same enrollee population would generate under FFS diagnostic coding patterns (MedPAC, March 2026). The residual is mechanical benchmark inflation: MA payments are set as a percentage of county-level FFS benchmarks, and benchmarks that include high-cost spending episodes pull MA payments up even for counties with lower average acuity.

The favorable selection component is the harder one to reduce through policy, because it reflects actual differences in who enrolls in MA, not gaming of the risk adjustment system. MA enrollees tend to be younger within the Medicare eligible age band, more likely to be in Medicare for age rather than disability, and more likely to be enrolled through former employer group coverage where the choice architecture favors managed products. CMS has progressively addressed coding intensity through the V28 model transition, the statutory minimum 5.90% coding intensity adjustment applied in both 2026 and 2027, and the 2027 rule change excluding diagnoses from unlinked chart review records. The V28 model went to 100% weight in CY 2026, and for CY 2027 CMS chose to retain the 2024 model calibration (using 2018 diagnosis data and 2019 expenditure data) rather than implement the updated model calibrated with 2023 and 2024 data that was proposed in the advance notice. That decision was net positive for MA plans, but it means the coding intensity wedge between MA and FFS has not fully closed.

Rebate mechanics add the third actuarial layer. When a plan's estimated costs fall below the county benchmark, the plan retains a share of the difference as a rebate, averaging nearly $2,400 per enrollee in 2026, and is required to deploy rebates toward supplemental benefits, reduced cost-sharing, or premium reductions (MedPAC, March 2026). The average MA enrollee receives $2,664 per year above estimated costs for Medicare-covered services. That rebate pool is what funds the dental, vision, hearing, and other supplemental benefits that drive enrollment growth. Actuaries pricing MA bids need to track benchmark levels by county, projected risk scores under V28 and the coding intensity adjustment, and the sustainable supplemental benefit budget implied by the resulting rebate margin, all as separable inputs rather than a single enrollment-weighted average.

The 2027 Stress Grid for Health Actuaries

The four Medicare trend components interact in ways that create compounding risk for plans bidding in 2027. The table below outlines the primary stress scenarios an actuary should consider as separate assumption sets, along with the channels through which each flows to financial results.

Assumption component	Base case (2027 bid)	Adverse scenario	Primary exposure
Part B utilization trend	4.5% to 5.5% ex-skin substitute reset	6.5%+ driven by physician-administered drug volume or new outpatient payment policy	MA benchmarks, Medigap claims, Part B premium
Risk score normalization (V28 + coding intensity adjustment)	Modest compression; 5.90% statutory coding intensity floor maintained	CMS shifts to 2023-calibrated model in 2028, accelerating compression by 200+ bps	MA bid margin, plan profitability, exit decisions
Part D catastrophic liability (GLP-1 and specialty)	6% to 7% trend on selected drugs post-negotiation; 60% plan catastrophic share	GLP-1 adherence ramp above bid assumptions; oral formulations launch 2027 and expand eligible population	Part D plan bids, MA Part D buy-down, premium stabilization
Supplemental benefit compression and plan-exit selection	Benefit reduction of 3% to 5% value-equivalent as rebate pools compress	Disproportionate exit of healthy enrollees following benefit reductions; adverse selection into remaining plans	MA risk pool quality, MA MLR, 2028 bid assumptions

The plan-exit selection row deserves emphasis because it is the hardest to quantify and the most consequential if it materializes at scale. When an MA plan reduces supplemental benefits or raises cost-sharing, the enrollees most likely to switch away are the healthiest, lower-cost members who enrolled primarily for the supplemental benefit rather than for care coordination. The sicker, higher-cost members, for whom the plan's network and care management are more valuable, tend to stay. If exit selection runs at 1.5 times the ratio seen in the broader Medicare population, a 5% decline in enrollment can increase the remaining pool's risk score by 2 to 3 percentage points without any change in actual medical trend, compressing MA margins in the following year's bid cycle before the actuary has sufficient data to reprice.

The 3 million MA beneficiaries who disenrolled when plans exited or contracted in 2026 provide a real-world data point on this dynamic. Plans that contracted to improve margins saw their remaining membership skew toward higher-acuity users; the plans that absorbed the returning FFS beneficiaries saw enrollment growth without a corresponding risk-score increase, because the switchers had been coded under their prior plan's V28 coding intensity rather than FFS diagnostic patterns.

Why the Actuarial Assumptions Cannot Be Averaged Together

Health actuaries working on MA bids, Medigap rate filings, or stop-loss pricing for Medicare beneficiaries have always needed to separate Part A and Part B trend. What has changed since 2023 is that Part D and MA benchmarks have both moved into assumption-set territory where the inputs are no longer correlated with general medical trend in a stable way. Part D trend is now driven by IRA negotiation cycles, GLP-1 adherence ramps, and catastrophic-phase plan liability, none of which tracks a simple PMPM inflation factor. MA benchmark trend is driven by per-capita FFS spending projections that are themselves being affected by skin substitute policy resets, physician fee schedule conversions factor movements, and the mix of inpatient versus outpatient services for which CMS payment policy has been actively changing.

The Trustees' aggregate $1.2 trillion figure, and the 6.7% Part D growth rate that the Trustees now publish explicitly as an increase from last year's 4.8%, are useful for understanding program-level sustainability and the Q2 2033 HI trust fund depletion trajectory. They are not useful as bid inputs. For bid purposes, the Part B assumption needs a baseline FFS trend, a physician fee schedule conversion factor assumption, and an explicit carve-out for any Part B categories with pending CMS payment policy changes. The Part D assumption needs a selected-drug unit cost path, a GLP-1 adherence ramp consistent with the bridge program's July 2026 launch, and a separate catastrophic reinsurance recovery assumption. The MA benchmark assumption needs a county-level FFS trend, a risk score path under the retained 2024 calibration model plus the 5.90% coding intensity adjustment, and a supplemental benefit cost projection that closes within the resulting rebate budget. Merging these four separate actuarial problems into a single trend factor is the most reliable way to miss the 2027 cycle.