2026 MMI Flags 14.8% Pharmacy Trend as Health Costs Hit $37,824

From building group health renewal rates over the past two years, the one workflow step that separates defensible filings from regulatory pushback is trend decomposition. A composite medical cost trend factor applied uniformly across inpatient, outpatient, professional, and pharmacy service categories masks the very cost drivers that plan design and utilization management target differently. The 2026 Milliman Medical Index, published May 20, makes the case for decomposition more forcefully than any prior edition: total healthcare costs for a hypothetical family of four reached $37,824, a 7.2% increase over the restated 2025 figure of $35,281 (Milliman, May 2026). Per-capita costs hit $8,460, up 7.9%, the steepest annual increase in more than a decade outside COVID-era distortions.

The divergence beneath that 7.9% headline is where pricing actuaries need to focus. Pharmacy costs surged 14.8% while outpatient facility costs rose 7.5%. Together those two categories accounted for 69% of total year-over-year cost growth. Inpatient and professional services trends ran meaningfully lower. Applying a blended 7.9% to all four categories would overstate inpatient and professional projections while understating the pharmacy and outpatient exposure that is actually accelerating. This article walks through a decomposition methodology for selecting service-category trend assumptions calibrated to the 2026 MMI and external benchmarks, with credibility-weighting techniques for small and mid-size employer groups entering the 2027 rate development cycle.

$37,824

2026 Family of Four Healthcare Cost (MMI)

14.8%

Pharmacy Cost Trend, Highest Service Category

69%

Share of Total Cost Growth: Pharmacy + Outpatient

The 2026 MMI Headline Numbers

Milliman has published the MMI annually since 2005, tracking total healthcare costs for a hypothetical family of four covered under an employer-sponsored PPO plan. The 2026 edition reports total costs of $37,824, allocated roughly as follows across the four major service categories: inpatient facility, outpatient facility, professional services, and prescription drugs. The employer share of this cost stands at 58%, with the employee bearing 42% through premiums, deductibles, copays, and coinsurance. That split has shifted meaningfully from the 61/39 employer/employee ratio Milliman reported in 2005 (Milliman MMI, 2026).

The per-capita figure of $8,460 (dividing the family cost by 4.47, Milliman's average family size adjustment) increased 7.9% year over year. For context, the PwC Health Research Institute's "Medical Cost Trend: Behind the Numbers 2026" report, drawing on a survey of 24 health plans covering 125 million group members, projects an 8.5% medical cost trend for the group market in 2026. The Mercer National Survey of Employer-Sponsored Health Plans reported employer health benefit costs rising 5.8% to $16,988 per employee in 2025 and projects acceleration into 2026 (Mercer, 2025). These three benchmarks bracket the national trend environment within which any plan-specific trend selection must be calibrated.

Decomposing the Composite Trend by Service Category

A composite trend factor treats the medical cost distribution as static. It is not. The first step in defensible trend selection is splitting the plan's historical PMPM experience into four service categories and running separate trend analyses on each.

Inpatient facility. Inpatient trends have been the slowest-growing category for several consecutive years, running in the 4% to 6% range nationally. Unit cost increases (revenue per admission) have been moderate, and utilization (admissions per thousand members) has been flat to declining as more procedures shift to outpatient ambulatory surgery centers and observation status. The MMI's inpatient component reflects this structural migration. For a plan-specific trend selection, the actuary runs exponential least-squares regression on 12 to 20 quarters of inpatient PMPM data, evaluating R-squared fit and residual patterns before choosing a regression period. If the regression trend falls within the 4% to 6% national range and the R-squared exceeds 0.85, the statistical trend typically stands without judgmental override.

Outpatient facility. This is where the cost structure is shifting. Outpatient facility costs rose 7.5% in the 2026 MMI, driven by site-of-service migration (procedures moving from inpatient to outpatient settings), increased utilization of advanced imaging and infusion therapy, and facility fee escalation at hospital-owned ambulatory centers. Outpatient spending now represents roughly 31% of total employer healthcare costs, up from approximately 25% a decade ago. This ratio shift matters for pricing actuaries because benefit leverage factors (the relationship between allowed cost changes and plan-paid cost changes) differ by service category. A plan with a $500 per-visit copay on outpatient surgery absorbs allowed-cost trend differently than one with 20% coinsurance. When the outpatient share of total costs grows, the plan-level leverage factor must be recalibrated or the projected benefit cost will be misstated.

Professional services. Physician and other professional service trends have been running 4% to 6%, similar to inpatient. The key driver here is unit cost negotiation: physician fee schedules in commercial plans are increasingly benchmarked to Medicare conversion factors with multipliers, and the 2026 Medicare physician fee schedule cut (before Congressional intervention) created downward pressure on the benchmark. Utilization trends in professional services are modestly positive, reflecting increased preventive care visits and specialist referrals post-pandemic. For most employer groups, the professional trend does not require significant deviation from the national range unless the plan has unusual specialist concentration or geographic anomalies.

Prescription drugs. The 14.8% pharmacy trend is the outlier that dominates the composite. This figure represents gross allowed costs before rebates. The MMI reports that manufacturer rebates offset roughly 31% to 33% of gross allowed drug costs, implying a net pharmacy trend in the range of 10% to 11% depending on the plan's rebate structure. Decomposing this further is essential, because the 14.8% is itself a composite of three distinct sub-trends with different drivers and different intervention points.

Parsing the 14.8% Pharmacy Surge

The pharmacy trend should be decomposed into at least three sub-components to avoid double-counting and to align trend assumptions with the plan's pharmacy benefit management strategy.

GLP-1 new-to-therapy and persistence. GLP-1 receptor agonists (semaglutide, tirzepatide) are the single largest driver of pharmacy trend acceleration. New-to-therapy enrollment continues to grow as coverage expands from diabetic to weight-management populations. Persistence rates vary significantly by indication: diabetic members maintain higher adherence than weight-management-only members, where Mercer research shows only 1 in 12 members remain on therapy at three years. For plans that already carved out GLP-1 claims in prior trend analyses, the GLP-1 sub-trend should be projected separately using the logistic adoption curve framework outlined in our GLP-1 trend factor analysis. For plans that have not yet isolated GLP-1 claims, the first step is NDC-level filtering: tag every claim line mapping to a GLP-1 molecule and separate the GLP-1 PMPM from the residual pharmacy book.

Specialty drug unit cost inflation net of biosimilar deflation. Specialty drugs outside the GLP-1 class continue to see unit cost increases in the 8% to 12% range for branded biologics, partially offset by biosimilar entry in categories like adalimumab (Humira biosimilars) and oncology supportive care. The net specialty trend depends heavily on the plan's formulary management: plans with aggressive biosimilar substitution policies can run 3 to 5 percentage points below plans on open formularies. When selecting this sub-trend, the actuary should reference the plan's PBM contract terms and the Milliman Health Trend Guidelines, which provide separate specialty and traditional drug trend benchmarks.

Traditional drug trend. Non-specialty, non-GLP-1 pharmacy (generics plus branded non-specialty) has been running 2% to 4% annually, with generic deflation partially offsetting branded price increases. This component is the most stable and the least likely to require judgmental override of the regression-derived trend. For plans with transparent PBM contracts, the net-of-rebate traditional trend may run negative in years with significant patent cliffs.

The composite pharmacy trend is reconstructed as: (GLP-1 PMPM × GLP-1 Trend) + (Specialty PMPM × Specialty Trend) + (Traditional PMPM × Traditional Trend) = Projected Total Rx PMPM. This bottom-up build avoids the distortion of applying the 14.8% gross composite to the entire pharmacy budget, which would overstate traditional drug costs and obscure the GLP-1 contribution.

The Rebate Offset: Gross vs. Net Trend Selection

The MMI's 31% to 33% rebate offset for allowed drug costs provides a national benchmark, but plan-specific rebate rates vary enormously based on PBM contract structure. Plans with transparent pass-through PBM contracts may capture 40% or more of gross drug costs as rebates, while plans on spread-pricing arrangements may realize effective rebate rates closer to 20% to 25%, with the PBM retaining the spread. The pricing actuary must select trend on a net-of-rebate basis consistent with the plan's actual contract terms. Using the MMI's 31% to 33% as the rebate assumption for a plan with a transparent contract would understate the rebate credit and overstate the net pharmacy trend. Conversely, applying a 40% rebate offset to a spread-pricing plan would understate net costs.

For 2027 ACA individual market rate filings, where the plan's rebate structure is known from the MLR reporting data, the net pharmacy trend should be derived from plan-specific allowed-to-paid ratios rather than industry benchmarks. For employer group renewals, where the PBM contract may be renegotiated at renewal, the actuary should build in an explicit rebate assumption range and sensitivity-test the renewal rate across that range.

Credibility-Weighting Plan Experience Against National Benchmarks

Small and mid-size employer groups lack the claim volume for fully credible plan-specific trend estimates. A 500-life group generates perhaps 6,000 member-months per year, yielding pharmacy claim counts that may be sufficient for aggregate trend estimation but insufficient for sub-category decomposition. The pricing actuary must blend plan-specific experience with industry benchmarks using a formal credibility framework.

The Bühlmann credibility approach assigns credibility Z = n / (n + k), where n is the plan's exposure in member-months and k is the ratio of process variance to the variance of hypothetical means, calibrated from a multi-plan dataset. For medical cost trends, k is typically estimated from national data sets: the MMI (representing actuarial cost estimates for a standardized plan), PwC's Behind the Numbers survey (24 health plans, 125 million group members), or the Milliman Health Trend Guidelines. The composite trend becomes:

Selected Trend = Z × (Plan-Specific Trend) + (1 - Z) × (Industry Trend)

For a 500-life group with 6,000 member-months and an estimated k of 15,000 (derived from the variance ratio across the PwC survey population), Bühlmann credibility is Z = 6,000 / (6,000 + 15,000) = 0.29. This means the plan's own experience receives only 29% weight, with 71% coming from the industry benchmark. For a 5,000-life group with 60,000 member-months, Z rises to 0.80, giving the plan's own data dominant weight. The k parameter is the critical judgment call, and it should be re-estimated periodically as the industry data sets update. PwC's 8.5% and the MMI's 7.9% provide the complement: the actuary selects one or blends both based on the plan's benefit design similarity to the benchmark population.

Outpatient Leverage Factor Recalibration

The shift toward outpatient spending (now 31% of employer costs, per the MMI) requires updated leverage factors when projecting benefit-level costs from allowed-cost trends. Leverage factors measure how a 1% change in allowed costs translates to a change in plan-paid costs, accounting for deductibles, copays, coinsurance, and out-of-pocket maximums.

A plan with a $2,000 individual deductible and 20% coinsurance has a higher leverage factor on outpatient services (where many claims fall near or below the deductible) than on inpatient services (where most claims exceed the deductible and hit the coinsurance layer). As outpatient's share of total costs grows, the weighted-average leverage factor across all services shifts upward, meaning that allowed-cost trend translates into a larger plan-paid cost increase than it did when inpatient dominated the mix. Pricing actuaries who use a static leverage factor calibrated to the prior year's service mix will understate 2027 plan-level costs. The fix is straightforward: recalculate leverage factors using the current year's service-category distribution and apply the updated factors to the category-specific trend projections.

Building an AI-Billing Optimization Load

The 2026 MMI identifies AI-enhanced medical billing as a cost inflator for the first time in the index's 21-year history. Provider revenue cycle management vendors have deployed machine learning tools that optimize CPT code selection, modifier usage, and charge capture, systematically shifting the coding distribution toward higher-reimbursement codes without necessarily reflecting higher clinical acuity. The effect is an increase in allowed costs per service that is not driven by unit price negotiation or utilization changes, but by coding precision improvement on the provider side.

For pricing actuaries, this creates a trend component without historical analog. The traditional trend decomposition of unit cost (price per service) and utilization (services per member) does not cleanly capture a coding-intensity shift that increases the effective price without a contractual rate change. One approach is to build an explicit load of 0.5% to 1.5% on top of the unit cost trend for professional and outpatient services, calibrated to the observed acceleration in average allowed amount per claim relative to contracted fee schedule increases. If the fee schedule increased 3% but average allowed per claim increased 4.5%, the 1.5 percentage point difference represents the coding optimization effect. This load should be monitored annually and adjusted as the AI-billing tools mature and payers develop countermeasures through pre-payment audit algorithms and code-pair editing.

Adjusting the Employer/Employee Cost Split for ACA Pricing

The MMI's 58/42 employer/employee cost split (down from 61/39 in 2005) has a specific implication for community-rated ACA pricing. Employer-sponsored plans shift costs to employees through higher deductibles and coinsurance, which affects the actuarial value of the underlying benefit design. When converting employer plan experience data into ACA metal-tier equivalents, the pricing actuary must normalize for this cost-sharing difference. A plan with a 58/42 split running at 82% actuarial value needs different trend and leverage assumptions than a Silver plan at 70% actuarial value or a Gold plan at 80%.

The benefit relativity factors used in ACA pricing (Bronze-to-Silver, Silver-to-Gold, etc.) assume a specific relationship between allowed cost trend and plan-paid cost trend at each metal level. If the employer/employee split has shifted since the relativity factors were last calibrated, the factors will misstate the cost differential across tiers. For 2027 ACA filings, pricing actuaries should verify that their induced utilization and benefit relativity assumptions reflect current cost-sharing levels rather than the 61/39 split that prevailed when many of these factors were originally developed.

Why This Matters

The 2026 MMI arrives at a moment when pricing actuaries face three simultaneous challenges: the steepest per-capita trend in over a decade, a pharmacy cost surge concentrated in a single drug class (GLP-1s) that many plans have less than two full years of mature claims data on, and a new trend inflator (AI-driven billing optimization) with no historical baseline. A composite 7.9% trend factor applied uniformly across service categories produces a rate that is directionally correct in aggregate but wrong in every component. The decomposition methodology outlined here, splitting trend by service category, further decomposing pharmacy into GLP-1, specialty, and traditional sub-trends, credibility-weighting plan experience against national benchmarks, recalibrating leverage factors for the outpatient mix shift, and building an explicit AI-billing load, produces a more transparent and defensible rate that can withstand regulatory review and client scrutiny.

Patterns we have seen in recent filing cycles suggest that state regulators are increasingly requesting service-category trend documentation rather than accepting a single composite figure. The Affordable Care Act's rate review process and ASOP No. 8 (Regulatory Filings for Health Benefits) both contemplate this level of granularity. Pricing actuaries who embed decomposed trend assumptions in their rate development workpapers now will be better positioned for the 2027 cycle.