Introduction

The Medical Loss Ratio (MLR) is a crucial metric in the health insurance industry that helps ensure insurance companies allocate a fair portion of premium revenue toward patient care rather than administrative costs or profits. Each percent that the MLR is incorrectly predicted could lead to millions of dollars of losses and missed revenue opportunities. Furthermore, Regulatory bodies, such as the Centers for Medicare & Medicaid Services (CMS), require insurers to submit MLR reports annually. Actuaries play a pivotal role in preparing these reports, ensuring compliance, accuracy, and financial soundness.

At Genhealth AI, we specialize in cost prediction for healthcare plans and providers, we assist insurers and actuaries in optimizing their MLR calculations through advanced data modeling and forecasting. We apply a novel approach to predicting medical expenses by using our Large Medical Model (LMM) which is trained on 140M patients across the US. The LMM can predict event level costs, procedures, meds, etc. and we can build up the population level expenses by adding up all the individual events. This approach enhances the accuracy of cost predictions far beyond any existing methodology and allows us to improve MLR predictions by orders of magnitude, ensuring better compliance and financial planning.

Major Steps in Creating an MLR Forecast

1. Gather Necessary Financial Data

Our MLR report starts with compiling relevant financial information including the following:

Membership Data: Number of covered members per month.

Claims Expenses: Total incurred claims, including claims paid and those still outstanding.

Quality Improvement Expenses: Costs incurred to improve healthcare quality, such as disease management programs or wellness initiatives.

Premium Revenue: Total earned premiums from policyholders.

Adjustments: Taxes, licensing, and regulatory fees that can be deducted.

We transform the data into a healthcare event sequence format which is used to drive predictions. We feed those sequences into our Large Medical Model which can consume a unified view of data from multiple sources, including claims data, EHR records, and in house databases.

2. Use AI to Forecast Expenses

Depending on the population size and variance in data, we may fine tune or further train our base model. This allows our AI to both learn from the 140M patients that GenHealth already has, but to also better conform to the care and demographic patterns of the customer’s data. Once we have this fine tuned model or decide to use the base model, we run inference on each individual. For each patient in the historical data, we feed in their full claims and medical history into GenHealth’s Large Medical Model. In addition to members that have historical data, plans and providers often get new patients that they have never seen before. For these patients we also simulate their futures, but prompt the model with limited data usually only related to the plan and geography.

At the end of this training and forecasting, we have new sequences that play out the future for each patient. Typically we run 64 futures for each individual and average these predictions to create a probabilistic distribution for each patient. Finally the predicted cost tokens are summed up to provide an aggregate view of the population’s expenses over time.

3. Ensure Data Accuracy and Integrity

Actuaries play a vital role in ensuring the accuracy, integrity, and compliance of financial data by validating expense allocations, properly reporting incurred but not reported (IBNR) claims, and adhering to CMS regulatory definitions, particularly CMS regulatory definitions.

A critical part of this process is identifying and addressing gaps in data that and predictions could impact MLR calculations, such as incomplete incurred claims reporting or misclassified administrative costs. To ensure precision, we also benchmark the incoming data and resulting predictions against the 140M patient population’s raw data set and standard stats based actuarial models.

Traditional actuarial reviews often rely on manual audits, which can miss subtle patterns in claims data. Our approach uses AI to scan billions of patient interactions, identifying anomalies such as duplicate claims, missing QIA expenses, or over-reported administrative costs. This allows insurers to correct inaccuracies early, ensuring compliance with CMS definitions and reducing the risk of financial penalties.

4. Estimating Plan Bids, Benchmarks, and Rebate Determination

In addition to the expenses, the crucial part of MLR reporting is the revenue. In the healthcare insurance landscape, two distinct types of rebates play a crucial role in regulating costs and ensuring fairness—one tied to Medical Loss Ratio (MLR) compliance and another linked to Medicare Advantage (MA) plan bids and benchmarks (which may vary by state and county) which ultimate help calculate revenue.

The MLR rebate acts as a penalty, requiring insurers to refund policyholders if they fail to allocate a sufficient portion of premium revenue toward healthcare services. In contrast, the Medicare Advantage rebate serves as an incentive, rewarding insurers that submit cost-efficient bids below the CMS benchmark by providing additional funds to enhance benefits for enrollees. While one ensures compliance, the other promotes competitive pricing and value-driven care.

undefined As part of those bids and benchmark rates we also can determine the member premium which may be $0 if the bid amount is below the benchmark rate.

We combine all of these to calculate the per member per month total revenue which ultimately becomes the Adjusted Premium Revenue for the health plan.

5. Calculate the MLR Ratio

The MLR is calculated by dividing the MLR numerator, which consists of health plan spending on incurred claims (excluding administrative non-claims costs) and quality improvement activities (QIA), by the MLR denominator, which is the total premium revenue minus applicable taxes and fees.

While traditional actuarial models primarily rely on historical trends, they often struggle to capture emerging cost drivers and patient-specific variations. Our LMM adapts in real time, incorporating shifting patient risk profiles, evolving treatment costs, and utilization trends. This allows insurers to generate more precise cost projections, improving their ability to meet MLR thresholds while maintaining financial stability.

Medicare Advantage Rebate (Plan Bid vs. Benchmark)

A key aspect of MLR calculation is assessing plan bids against the CMS benchmark for the designated region. This comparison helps determine whether a plan qualifies for rebates or requires members to cover additional premiums.

Step 1: Compare Plan Bids to the CMS Benchmark

Each Medicare Advantage (Part A & B) and Part D sponsor submits a bid amount, which represents their projected cost of providing covered services. This bid is then evaluated against the benchmark amount, which is the maximum CMS is willing to reimburse in that region. The benchmark is derived from:

• Plan-Bid Component: The adjusted benchmark based on the weighted average of bids in the region.
• Statutory Component: CMS's estimate based on traditional Medicare costs before adjustments for quality bonuses.

Step 2: Determine Rebate or Additional Premium Requirements

• If Medicare Advantage (MA) plan’s bid is below the benchmark, the plan qualifies for a rebate, which is paid by CMS to the plan sponsor (the insurer). The rebate amount is determined using a formula that factors in the rebate percentage, number of enrolled members, their risk scores, the CMS benchmark, and any quality bonus adjustments. The rebate percentage varies based on the plan’s star rating and quality performance, with higher-rated plans receiving a larger rebate.

However, insurers cannot retain the rebate as profit. Instead, they are required to use the rebate funds to directly benefit enrollees. These benefits can take the form of lower monthly premiums, reduced out-of-pocket costs (such as deductibles and copayments), or additional covered services like dental, vision, or wellness programs. By structuring the system this way, CMS ensures that cost savings from efficient plan management are passed down to beneficiaries rather than increasing insurer profits.

• If an MA plan’s bid is above the CMS benchmark, CMS will only cover up to the benchmark amount, and the enrollee must pay the difference in the form of an additional premium. This structure creates a financial incentive for insurers to submit competitive bids and efficiently manage healthcare costs while maintaining quality standards.

Medical Loss Ratio (MLR) Rebate – Ensuring Compliance

The Affordable Care Act (ACA) imposes MLR thresholds to ensure insurers allocate a fair share of premium revenue toward actual healthcare services and quality improvement:

• 80% for individual and small group markets (e.g., a small tech startup with 50 employees purchasing a group health plan).
• 85% for large group markets (e.g., a national retail chain with 10,000 employees providing insurance to its workforce).

If an insurer does not meet these thresholds, they must issue rebates to policyholders. Accurate cost predictions are essential in this process because underestimating costs can lead to unexpected shortfalls, forcing insurers to pay unplanned rebates, while overestimating costs may result in excessive premiums, making plans less competitive.

Our Large Medical Model (LMM) enhances MLR compliance by accurately predicting cost distributions, reducing the risk of non-compliance and unexpected financial shortfalls. By adjusting for regional cost variations at a patient-event level, it ensures that rebate calculations precisely reflect actual healthcare utilization patterns. Additionally, our model optimizes risk-adjustment methodologies, improving revenue planning and strengthening regulatory compliance. It proactively identifies spending inefficiencies that could lead to MLR shortfalls, allowing insurers to adjust premium allocations before exceeding compliance thresholds. By forecasting spending trends at an individual patient level, insurers can fine-tune their healthcare investments, reducing rebate liabilities while improving overall plan sustainability.

Looking to the Future: How Genhealth AI Transforms MLR Compliance

Creating an MLR report requires a meticulous approach to financial data, regulatory compliance, and precise calculations. Actuarial forecasting plays a vital role in this process by ensuring accuracy, helping insurers meet legal obligations, and protecting consumers. With the advent of transformer based neural network models like the one we employ in our LMM, insurers can confidently submit MLR reports that exceed the classical methods and improve financial stability.

We have found, traditional actuarial methods alone often struggle to capture real-time cost fluctuations, patient-specific variations, and regional pricing trends—leading to higher rebate liabilities, pricing inefficiencies, and compliance risks. This is where Genhealth AI’s Large Medical Model (LMM) revolutionizes the process.

By integrating AI-driven cost predictions, real-time risk stratification, and automated data validation, our LMM transforms MLR forecasting into a proactive, precision-driven process. Insurers gain a competitive edge by:

• Reducing unexpected rebate penalties with real-time forecasting that identifies spending inefficiencies before they impact MLR compliance.
• Enhancing bid competitiveness by dynamically adjusting bids based on evolving cost trends and regional benchmarks.
• Improving financial sustainability by optimizing premium allocation, ensuring regulatory compliance while maintaining market competitiveness.
• Automating MLR reporting and compliance checks, significantly reducing administrative burden and audit risks.

With Genhealth AI’s LMM, insurers move beyond reactive compliance and gain strategic control over their financial planning. By making data-driven, patient-level decisions, they can ensure competitive pricing, optimize rebate distributions, and enhance the overall quality of care—all while staying ahead of regulatory requirements.

The Future of MLR Compliance is Here

By leveraging AI-powered patient level predictions, insurers can streamline MLR reporting, mitigate compliance risks, and uphold industry best practices with greater efficiency than ever before. The integration of advanced AI into actuarial processes doesn’t just improve MLR forecasting—it transforms it, paving the way for a more sustainable, compliant, and financially stable future.