For the fourth year in a row, commercial payers are bracing for continued increases in total healthcare cost trends. According to the most recent MarketScan® Semiannual Norms1, overall PMPY trends between 2024 and 2025 were 12.1% for the active commercial population with a 5% increase in facility inpatient spend and a 19.8% increase in prescription drugs. A recent report by PwC indicates costs are projected to climb another 8.5% in 2026. This marks a more than 10-year high as new therapeutics—especially GLP1 drugs, oncology treatments, and immunologic agents—continue to exert significant inflationary pressure, while at the same time, hospitals are shifting costs to commercial payers.
While the healthcare landscape continues to shift under the weight of multiple forces, one constant remains: a small percentage of members drive a disproportionate share of total spend. MarketScan data1 shows that just 1% of members account for nearly 30% of total costs. Understanding who will fall into this category next year—not just who was high-cost last year—has become essential.
These persistent pressures create an urgent need for new strategies. Yet a core challenge emerges when examining year over year volatility: MarketScan data2 indicates more than half of individuals categorized as high-cost claimants in one year will not remain in that category the following year. This makes it difficult for payers to rely on retrospective data alone.
To address this, organizations are increasingly turning to AI and machine learning. By training predictive models on large, longitudinal datasets—such as the MarketScan® Commercial Claims and Encounters database—payers can uncover patterns that traditional analytics miss. These models learn from millions of historical medical claims, pharmacy records, and utilization patterns to identify subtle indicators of rising risk long before major costs appear.
Failing to reach emerging high-risk individuals leads to preventable increases in cost, while engaging the wrong population wastes limited resources. The question becomes: What if you could identify rising risk or future high-cost members before major claims occur? AI-driven prediction makes that possible.
1. Addressing limited resources
Most organizations do not have unlimited care management capacity. Outreach teams—care managers, social workers, nurses, benefit coordinators—must prioritize effectively. Machine learning models help by identifying not only who is likely to become high-cost next year, but also who is most actionable.
Traditional workflows that rely on last year’s high-cost claimants often capture individuals with one-time events or well managed chronic conditions. AI-based models can distinguish between transient high spend and emerging risk, enabling earlier intervention and more efficient resource allocation.
2. Risk stratification and prioritized outreach
Once predictive insights are available, the next challenge is operationalizing them.
Effective models should help answer:
Who is at the highest risk of increased cost, and who can be deprioritized?
Which service settings are driving predicted risk?
How soon will members become high risk?
What are the key features and drivers of risk?
How actionable is each high-risk member?
What are their predicted costs?
Machine learning models excel at producing these granular, forward-looking insights. They forecast risk and provide the context needed to guide outreach sequencing, care management workflows, and targeted interventions.
3. Understanding what drives future costs
Predicting future high-cost claimants requires understanding the complex interplay of clinical, behavioral, and socioeconomic factors. Diagnoses alone are insufficient. AI models should incorporate:
Sociodemographic and socioeconomic indicators
Recency and severity of major events (e.g., inpatient stays)
Pharmacy utilization, especially specialty medications
Behavioral health needs
Historic patterns such as ER use and provider relationships
Disease severity and progression
Machine learning identifies which combinations of these factors most strongly predict future cost escalation—often revealing relationships too complex for manual analysis.
Adopting AI-driven predictive modeling is not just a clinical or operational improvement—it’s a financial strategy. When predictive models accurately identify rising risk members before major cost events occur, organizations can redirect care management resources toward members who are most likely to benefit. The result is a measurable return on investment.
Below is an illustrative example of how ROI can materialize for a commercial payer with 250,000 covered lives:
In the wake of inflationary pressures, limited outreach resources, and increasingly complex patient needs, benefits and care managers must shift from reactive strategies to proactive ones. Traditional retrospective analysis has value, but AI-powered predictive models enable payers to:
Forecast high-cost claimants before major events occur
Allocate resources where they will have the greatest impact
Reduce preventable hospitalizations
Tailor interventions to the specific drivers of each member’s predicted risk
Truven’s Risk of Rising Cost model delivers a powerful, AI-driven view into which members are most likely to experience significant cost escalation. Built on the depth of the MarketScan dataset, the model uses over 400 clinical, behavioral, socioeconomic, and utilization features to detect early signals of risking risk that traditional retrospective methods miss.
Predictive modeling, when paired with actionable care management strategies, offers a path to mitigating rising trends in a continuously shifting healthcare landscape. AI doesn’t replace human expertise—it amplifies it, helping organizations intervene earlier, act smarter, and ultimately improve outcomes for both members and the bottom line.
Have questions about AI and how it can be utilized with healthcare data? Connect with our team today to learn more.
1 Merative MarketScan® Semiannual Employer Norms, Commercial
Current: Paid data Jul 2024 - Jun 2025
Previous: Paid data Jul 2023 - Jun 2024
2 Merative MarketScan® Commercial Incurred 2023 Paid through March 2024
3 Analysis based on 2022 analysis of MarketScan® Commercial claims from 23M lives