PwC's June 2026 survey of actuaries at 27 US health plans projects a 9.0% group medical cost trend for 2027, the highest reading since 2010. Seventy percent of those plans ranked provider AI revenue optimization tools among their top three cost drivers. It is the one driver in the composite that raises reimbursement per encounter without adding clinical volume or acuity, and it compounds forward as adoption spreads.

Key Takeaways

  • 9.0% for 2027 across 27 health plans, the highest PwC reading since 2010, with provider AI revenue optimization in the top three drivers for 70% of respondents.
  • Contracted fee schedules rising 3% to 4% against average allowed per professional claim growing 7% to 10%. The gap is billing precision, not clinical intensity.
  • Forty-six percent of hospitals already use AI in revenue cycle functions and 74% use some form of automation, so the effect spreads as adoption reaches community hospitals and employed physician groups.
  • One or two systems control the entire inpatient market in 47% of US metro areas, which is the contracting base the coding increment layers on top of.
  • GLP-1 spending went from $4.34 PMPM in 2022 to $27.23 in the first quarter of 2025, a driver with genuine two-sided risk that the 9.0% composite does not separate.

The 9.0% and Its Fifth Driver

The signal was visible at the encounter level before it reached aggregate statistics. Evaluation-and-management level distributions drifted toward CPT 99215 at rates no underlying acuity shift explained, with contracted fee schedules rising 3% to 4% annually while average allowed per professional claim grew 7% to 10%. That gap is billing precision.

PwC's five structural drivers are not equally tractable. GLP-1 pharmacy has a plausible deceleration path through biosimilar entry. Behavioral health growth is supply-constrained and parity-driven. No Surprises Act adjudication creates a ratchet on out-of-network reimbursement. General fee schedule inflation reflects consolidation leverage.

Provider AI revenue optimization differs from all four. It extracts incremental reimbursement per existing encounter with no change in services rendered, and it grows as tool adoption widens rather than as patients get sicker. PwC frames the plan-side problem as one of separation: "cost management success depends on pairing disciplined contracting with advanced capabilities in coding-intensity surveillance, severity-shift monitoring, and payment integrity to distinguish changes rooted in documentation from shifts in patient complexity."

Why a Code-Layer Effect Cannot Be Managed Down

Four tool categories each add an independent increment. Computer-assisted coding parses clinical documentation and assigns the highest defensible codes, addressing a category where coding errors account for 40% to 80% of billing accuracy problems: a comorbidity in the chart but not in the ICD-10 set, or a visit coded 99213 where documentation supports 99215. Prior authorization prediction models payer adjudication logic against a commercial first-pass denial rate of 11.65%. Discharge optimization identifies documentation supporting additional covered days. Denial prevention scores claims before submission against the specific payer's denial history.

Adoption is already broad. Forty-six percent of hospitals use AI in revenue cycle functions and 74% use some form of automation, with these systems integrated into the major EHR platforms.

The increment lands on a base that payers cannot negotiate away. In 2024, one or two health systems controlled the entire inpatient market in 47% of US metro areas, and in 83% of metro areas one or two systems held more than 75% of inpatient share. At least 47% of physicians were employed by or affiliated with hospital systems, up from fewer than 30% in 2012.

That combination is what makes the standard levers ineffective. Network narrowing and tighter prior authorization operate on which providers are used and how often. The coding increment arrives through the charge per encounter at the providers already in network, so a plan can move its entire book to tier-one facilities and absorb the full effect.

The consequence for pricing is a broken estimation window rather than a higher point. A regression fitted across a period spanning tool adoption blends pre- and post-adoption claims and understates the forward run rate. Completion factors move too: vendors report 40% to 70% reductions in time from service delivery to coded claim submission, and prior authorization and denial prevention tools remove the revision loops that delayed resubmission. Where 3-month and 6-month paid percentages have shifted upward against the historical pattern in affected specialties, factors fitted on the slower window overstate IBNR, and the excess unwinds into a trend that reads lower than the truth.

The Composite Mixes a Ratchet With a Coin Flip

Decomposing the 9.0% matters because its components will not move together, and two of them are genuinely unresolved.

GLP-1 is the clearest case. Per-member per-month spending went from $4.34 in 2022 to $27.23 in the first quarter of 2025, and GLP-1 claims rose from 6.9% of total employer prescription spending in 2023 to 10.5% by 2025, with more than 85% of surveyed actuaries expecting 2027 pharmacy trend to outpace the medical composite. A plan running a 36-month regression on total pharmacy PMPM without separating GLP-1 claims is fitting a curve whose slope changes mid-window, as our GLP-1 trend factor analysis sets out.

The uncertainty runs both ways. Biosimilars to injectable semaglutide and tirzepatide are expected, and oral formulations began reaching patients in 2026 near $200 a month. The floor is structural given employer coverage commitments; the ceiling depends on pricing dynamics that are not settled. Embedding the current trajectory without weighting the biosimilar path risks overpricing 2027.

The out-of-network side ratchets rather than reverting. More than 5.1 million disputes had reached the No Surprises Act independent dispute resolution portal as of January 31, 2026, with 1.2 million filed in the first half of 2025 alone, 40% above the preceding six months. Provider-side tools now model which dispute profiles clear the qualifying payment amount, so weak disputes settle and strong ones are contested with optimized documentation. Award data then becomes a floor in in-network renegotiation.

In markets with high IDR adoption, out-of-network trend in emergency medicine and anesthesiology can run 15% to 25% above the in-network composite, which a blended selection will not show. Denial rate data by service category from the CMS prior authorization metrics release anchors the utilization management side of it.

So the filing question is not the headline. It is which of unit cost, encounter volume, coding intensity, and true morbidity change carries each point of the 9.0%, and our companion analysis on the rate filing benchmark works that split through. Attributing the coding signal to member morbidity produces the right headline number with the wrong mechanism behind it, and no answer when a regulator asks which driver is expected to revert.

Further Reading

Sources