Health plan acquisition is an expensive and risky investment. Acquiring a new individual member costs between $500 and $1,500. A group enrollment involves broker fees, implementation costs, and administrative setup that can reach several thousand dollars per enrolled employee. The economics of that investment depend entirely on whether the members who enroll are priced correctly relative to the medical costs they will generate. When they are not, the cost does not surface in the acquisition budget. It surfaces in the first-year Medical Loss Ratio, months later, when the damage is already done.
The fundamental problem in member enrollment is information asymmetry. Members and brokers know significantly more about expected utilisation than the health plan does at the time of enrollment. Individuals enrolling through open enrollment during a period of active health concern are systematically more likely to be high-cost than individuals enrolling from a position of good health. Groups that move their insurance business are more likely to be doing so because their current insurer has priced them correctly — reflecting their actual experience — than because they are genuinely seeking a better deal. The health plan that does not account for this asymmetry in its risk scoring will consistently attract worse-than-average risk from competitive bidding.
Where the enrollment decision breaks down
The core enrollment decisions — plan selection guidance, prospective risk scoring, and onboarding outreach — are all made with less information than is theoretically available. Application data captures demographics, geography, and stated health status. For group enrollments, prior carrier claims experience may be available but is often incomplete or presented in summary form that obscures the distribution of risk within the group.
What application data does not capture are the signals that are most predictive of first-year cost: prior pharmacy utilisation patterns, recent laboratory results, prior authorization history, and the behavioral indicators of chronically ill members who are actively managing their conditions. These signals are available for members who have prior coverage history — which is the majority of the enrollment population in a mature market. The question is whether the health plan has built the infrastructure to incorporate them into the enrollment decision.
Plan selection guidance compounds the problem. Generic plan recommendations based on demographic bands push members toward plan designs that may not match their actual utilisation profiles. A member with high pharmacy utilisation enrolling in a low-premium, high-deductible plan faces significant out-of-pocket exposure that may be manageable in a year of good health and unmanageable in a year of active utilisation. That member disenrolls at the first opportunity, replacing themselves with another unknown risk. The churn cost — acquisition expense plus first-year MLR variance — is substantially higher than the cost of a better plan recommendation at the point of enrollment.
The economic case for better risk scoring
Each 0.01 point of RAF score inaccuracy in Medicare Advantage represents approximately $80 to $100 per member per year in revenue impact. On a population of 100,000 Medicare Advantage members, the difference between accurate prospective risk scoring and inaccurate risk scoring compounds into material revenue or margin variance. The same dynamic applies in commercial insurance through its effect on the first-year MLR: a cohort whose risk was underestimated by 10% at enrollment will generate 10% more medical cost than the pricing model assumed, absorbed entirely as margin erosion.
The prospective risk model that improves these outcomes uses a richer feature set than application data alone. Prior pharmacy claims are among the most predictive features available: medication profiles reveal chronic conditions that may not be documented in the application, specialist visit patterns reveal active management of complex conditions, and recent authorization history reveals pending high-cost events. Integrating these signals into the enrollment risk score produces a materially better predictor of first-year cost — and a basis for plan selection guidance that matches the member to the plan design most likely to serve them well through their actual utilisation pattern.
Risk-stratified onboarding changes the first-year economics further. A newly enrolled member identified as high-risk at enrollment can be engaged proactively with care management outreach, pharmacy management support, and primary care attribution before they generate the high-cost event that would otherwise be the first signal of their risk profile. A preventable hospital admission in the first year costs between $12,000 and $20,000. The care management contact that prevents it costs a fraction of that. The window to make that contact is the period between enrollment and first utilisation — and it only exists if the risk identification happened at enrollment rather than after the first claim was processed.
The technology dimension
Prospective risk scoring at enrollment draws on data from multiple sources that large health plans hold across their claims, pharmacy, and clinical data estates. For plans running their core claims and member data on IBM Z, the prior claims, pharmacy, and authorization history required to build the enrollment risk score is resident on the same platform. A model deployed via IBM Machine Learning for z/OS can score incoming enrollment applications against the full member data estate in real time, returning a risk tier and plan recommendation alongside the eligibility determination — within the enrollment processing workflow, without requiring data extraction or external API calls.
What success looks like
The metrics for a prospective enrollment risk programme are first-year MLR by enrolled cohort compared to pricing assumption, disenrollment rate in the first year, and care gap closure rate among high-risk newly enrolled members. The programme baseline should be established before deployment by retrospectively comparing the predicted risk scores of prior cohorts against their actual first-year cost performance. The gap between predicted and actual is the opportunity. Closing it through better prospective scoring is the investment.
