AI Analytics

Technical writing

CMS Medicare Advantage: Plan Bids, Star Ratings, and the Federal Dataset Behind Private Medicare

· AI Analytics
Federal DataCMSMedicare AdvantageHealthcare

Medicare Advantage — the private-plan alternative to traditional fee-for-service Medicare — crossed a historic threshold in 2024 when more than half of all Medicare beneficiaries enrolled in a private plan for the first time. Behind that enrollment milestone sits one of the most elaborate federal data ecosystems in American healthcare: annual landscape files, monthly enrollment counts, county-level benchmark rates, star rating scorecards, and partially disclosed bid data that collectively reveal how the government pays private insurers to deliver Medicare benefits to 33 million Americans.

What Medicare Advantage Is

Medicare Advantage (Part C) allows Medicare beneficiaries to receive their Parts A and B benefits through a private health insurance plan rather than through the traditional fee-for-service (FFS) Medicare program. A beneficiary enrolled in an MA plan pays premiums to a private insurer instead of directly to the federal government, receives care through the plan's provider network, and typically gains access to supplemental benefits — dental, vision, hearing aids, over-the-counter allowances — that traditional Medicare does not cover.

The statutory authority for MA comes from the Balanced Budget Act of 1997, which created the Medicare+Choice program as a way to introduce private-plan competition into Medicare. The Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (MMA) renamed the program Medicare Advantage, substantially increased payment rates to attract insurer participation, and added the Part D prescription drug benefit that most MA plans now bundle into MA-PD (Medicare Advantage with Prescription Drug coverage) offerings.

As of 2024, approximately 33 million Medicare beneficiaries — representing roughly 51% of all Medicare eligibles — are enrolled in MA plans. That majority threshold marks a structural shift: MA has moved from an alternative to traditional Medicare to the dominant form of Medicare coverage in the United States. The implications for provider contracting, drug formulary design, federal spending, and health equity are significant, and they all trace back to the payment and performance data that CMS publishes.

The CMS MA Data Ecosystem

CMS publishes MA data through several distinct channels, each serving different analytic purposes.

Plan Finder data is the public-facing layer, published annually on medicare.gov. Beneficiaries use it to compare plans available in their county, view premiums, search drug formularies, and see star ratings. The underlying data is structured and downloadable, making it a starting point for plan-level benefit comparison research.

Landscape files are comprehensive plan-level data files released annually by CMS showing every MA and Part D plan available in every county — premiums, cost-sharing structures (deductibles, copays, out-of-pocket maximums), star ratings, plan types (HMO, PPO, PFFS, SNP), and supplemental benefit availability. The landscape files are the workhorse dataset for researchers studying plan design, geographic variation in plan availability, and premium trends. CMS releases them at cms.gov/medicare/health-drug-plans, typically in October ahead of the Annual Enrollment Period (AEP).

Monthly enrollment data provides plan-level and county-level enrollment counts updated each month. CMS publishes these through the Medicare Advantage and Part D Contract and Enrollment Data page, giving researchers a near-real-time view of how enrollment shifts within and across plans and geographies during and after the AEP.

Contract and Enrollment data aggregates enrollment by contract, plan, state, and county on a quarterly basis. This is the standard input for market concentration analyses, state-level insurer market share calculations, and longitudinal enrollment trend studies.

Bid data is partially disclosed. CMS publishes county-level benchmark rates — the maximum payment rate against which plan bids are compared — in the Medicare Advantage Ratebook and Supplementary Rate Development Data files released each spring. Individual plan bid amounts remain confidential as proprietary business information, but the spread between a plan's premium and the benchmark (the rebate) can be inferred from publicly available premium and benefit data.

The Star Ratings System

CMS rates MA-PD plans on a 1–5 star scale each fall, with ratings based on performance data from the prior measurement year. The annual star ratings serve two functions simultaneously: they inform beneficiary plan selection during the AEP, and they determine which plans receive Quality Bonus Payments (QBPs) from CMS.

The rating methodology incorporates more than 40 measures organized into five domains:

Plans with overall ratings of 4 stars or higher receive QBPs — an additional payment from CMS that increases county-level benchmarks by 5% for 4-star plans and 5% for 5-star plans in qualifying counties (double bonuses apply in certain low-enrollment counties). In the 2024 plan year, approximately 72% of MA beneficiaries were enrolled in plans rated 4 stars or higher, a figure that reflects both genuine quality improvement and, some researchers argue, measurement issues in how the ratings are constructed.

The financial stakes of star ratings are substantial. A large MA insurer with millions of enrollees can gain or lose hundreds of millions of dollars in annual revenue depending on whether its plans clear the 4-star threshold. This creates strong incentives for plans to invest in the specific activities — outreach for preventive care, member experience programs, medication adherence interventions — that the star rating measures reward.

The Benchmark and Bidding System

The financial architecture of MA rests on the relationship between county-level benchmarks and plan bids. CMS calculates a benchmark for each county each year, representing the maximum per-member-per-month (PMPM) payment rate the government will use as a reference for that market. The benchmark is derived from a blend of local MA payment rates and traditional Medicare FFS spending for that county's beneficiary population, with adjustments for geographic wage differentials and other factors.

Each MA plan submits an annual bid to CMS representing its estimated cost to provide Medicare-covered services to an average-risk beneficiary in its service area. The bid is expressed as a PMPM dollar amount.

When a plan's bid falls below the benchmark, the difference generates a “rebate.” CMS returns a percentage of that rebate (currently 50–70% depending on the plan's star rating) to the plan, which is then required to use those rebate dollars to offer supplemental benefits (dental, vision, hearing, OTC allowances, fitness memberships) or to reduce member premiums. This is why many MA plans can offer $0 premiums and rich supplemental benefits in counties with generous benchmarks — the benchmark-to-bid spread funds those extras. When a plan's bid exceeds the benchmark, the plan must charge beneficiaries the difference as a premium.

Quality Bonus Payments layer on top of this structure. For plans rated 4 or 5 stars, the effective county benchmark is increased by 5%, which expands the potential rebate and therefore the supplemental benefits or premium reductions a plan can offer. In markets with generous benchmarks and 5-star plans, the combination can produce very rich benefit packages at zero premium — which in turn drives enrollment toward high-rated plans in competitive markets.

Risk Adjustment and the HCC Model

CMS does not pay MA plans a flat PMPM rate for every enrollee. Instead, payments are risk-adjusted using the CMS-HCC (Hierarchical Condition Category) model. The HCC model assigns each beneficiary a risk score based on their documented diagnoses from the prior year, with higher scores for beneficiaries with more serious or costly conditions. A beneficiary with advanced heart failure, chronic kidney disease, and diabetes will carry a risk score several times higher than a healthy beneficiary of the same age, and the plan receives a correspondingly higher payment for that member.

The HCC model creates a financial incentive for MA plans to document as many diagnoses as possible for their enrolled population — a practice known as risk score upcoding. Plans conduct retrospective chart reviews, health risk assessments (HRAs), and in-home visits specifically to identify diagnoses that may not have been captured in claims data but can legitimately be coded and submitted to CMS to increase risk scores.

The excess payment problem this creates has been documented repeatedly. The HHS Office of Inspector General (OIG) has published multiple analyses finding that MA plans receive $10–30 billion or more per year in excess payments attributable to risk score differences between MA enrollees and demographically comparable traditional Medicare beneficiaries. The Government Accountability Office (GAO) has flagged the same issue in multiple reports, noting that MA enrollees consistently show higher risk scores than comparable FFS beneficiaries, and that the difference cannot be fully explained by actual health status differences.

CMS has responded with periodic recalibrations of the HCC model and changes to the risk adjustment data validation (RADV) audit program, which requires plans to return payments for diagnoses that cannot be supported by medical records. The 2023 RADV final rule significantly expanded the financial exposure for plans with unsupported diagnosis codes, triggering substantial industry opposition and litigation.

Enrollment Trends and Market Structure

MA enrollment has grown almost continuously since the program was restructured under MMA 2003. From roughly 5 million enrollees in 2003, the program has grown to approximately 33 million in 2024 — a more than sixfold increase over two decades. Annual enrollment growth has averaged 8–10% in recent years, driven by aggressive marketing, rich supplemental benefits, $0-premium plans, and the aging of the baby boom generation into Medicare eligibility.

The market has consolidated substantially. The top three MA insurers by enrollment control roughly 60% of all enrolled beneficiaries:

Market concentration varies substantially by geography. Some counties — particularly rural markets — are served by only one or two MA plans, while dense metropolitan markets may offer 30 or more plan options during the AEP. The CMS landscape files make it possible to measure plan availability and market concentration at the county level for every year since the modern MA program began.

The CMS MA Public Data Files

Researchers working with MA data have access to a rich set of public files from CMS:

The Prior Authorization Controversy

MA plans have faced sustained regulatory and congressional scrutiny for their use of prior authorization — a requirement that providers obtain insurer approval before delivering certain services or prescribing certain drugs. Prior authorization is rare in traditional FFS Medicare, where the fee schedule defines covered services and claims are paid without pre-approval. MA plans, as managed care entities, routinely apply prior authorization to post-acute care (skilled nursing facilities, home health), inpatient stays, durable medical equipment, certain drugs, and high-cost procedures.

The Kaiser Family Foundation has published analyses finding that MA plans deny 6–7% of prior authorization requests overall, compared with a near-zero denial rate in traditional Medicare for the same services. More troubling, the OIG's 2022 report found that MA plans denied 13% of prior authorization requests that appeared to meet Medicare coverage criteria — meaning the coverage criteria were satisfied but the plan still denied the request. For beneficiaries who do not appeal denials, these decisions can result in foregone care that traditional Medicare would have covered.

CMS responded by issuing new prior authorization transparency rules, effective in 2024 and 2025, that require MA plans to report detailed prior authorization data — the number of requests received, approved, denied, and overturned on appeal, broken down by service category. This new reporting requirement will generate a public dataset that researchers and policymakers can use to assess the scope and patterns of prior authorization denials across the MA market for the first time.

Python: State-Level MA Market Share by Parent Organization

The following script downloads the CMS monthly MA enrollment file, aggregates the most recent 12 months of enrollment by contract, plan, and state, computes market share by parent organization for each state, and identifies states where a single MA insurer controls more than 50% of total MA enrollment.

import requests
import pandas as pd
from io import BytesIO
import zipfile

# CMS publishes monthly MA enrollment by contract/plan/state/county at data.cms.gov
# The flat file endpoint returns a ZIP containing the CSV

BASE_URL = "https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports"
ENROLLMENT_URL = (
    BASE_URL
    + "/MCRAdvPartDEnrolData/Downloads/MA-State-County-Enrollment-by-Month.zip"
)

resp = requests.get(ENROLLMENT_URL, timeout=180)
resp.raise_for_status()

# The ZIP contains one CSV with all months; load it
with zipfile.ZipFile(BytesIO(resp.content)) as zf:
    csv_name = [n for n in zf.namelist() if n.endswith(".csv")][0]
    with zf.open(csv_name) as f:
        df = pd.read_csv(f, dtype=str, low_memory=False)

# Standardize column names: strip whitespace, lowercase, replace spaces with underscores
df.columns = [c.strip().lower().replace(" ", "_") for c in df.columns]

# Enrollment column is typically 'enrollment'; coerce to numeric
df["enrollment"] = pd.to_numeric(df.get("enrollment", df.get("total_enrollment", "0")), errors="coerce").fillna(0)

# Keep only the 12 most recent months available in the file
if "month" in df.columns:
    df["month"] = df["month"].str.strip()
    recent_months = sorted(df["month"].unique())[-12:]
    df = df[df["month"].isin(recent_months)]

# Aggregate enrollment by contract ID, plan ID, state, and parent organization
group_cols = [c for c in ["contract_id", "plan_id", "state", "parent_organization"] if c in df.columns]
agg = (
    df.groupby(group_cols, dropna=False)["enrollment"]
    .sum()
    .reset_index()
    .rename(columns={"enrollment": "total_enrollment_12mo"})
)

# State-level total enrollment per parent organization
if "parent_organization" in agg.columns and "state" in agg.columns:
    state_parent = (
        agg.groupby(["state", "parent_organization"])["total_enrollment_12mo"]
        .sum()
        .reset_index()
    )
    # Total MA enrollment per state
    state_total = (
        state_parent.groupby("state")["total_enrollment_12mo"]
        .sum()
        .rename("state_total")
        .reset_index()
    )
    state_parent = state_parent.merge(state_total, on="state")
    state_parent["market_share_pct"] = (
        (state_parent["total_enrollment_12mo"] / state_parent["state_total"] * 100)
        .round(1)
    )

    # Identify states where a single MA insurer controls more than 50% of enrollment
    dominant = state_parent[state_parent["market_share_pct"] > 50].sort_values(
        "market_share_pct", ascending=False
    )
    print("States with a single MA insurer controlling >50% of enrollment:")
    print(
        dominant[["state", "parent_organization", "total_enrollment_12mo", "market_share_pct"]]
        .to_string(index=False)
    )
else:
    print("Parent organization column not found; printing top contracts by state:")
    top = agg.sort_values("total_enrollment_12mo", ascending=False).head(25)
    print(top.to_string(index=False))

A few implementation notes. CMS enrollment files use inconsistent column naming across vintages, so the script normalizes column names before processing. The “parent organization” field groups individual contracts under their corporate parent, which is critical for computing insurer-level market share — a large insurer like UnitedHealthcare operates dozens of distinct contracts under different plan names, and without the parent organization field those contracts would appear as separate entities. The 50% concentration threshold identifies markets where a single insurer can exercise significant leverage over providers and beneficiaries alike, which is a standard threshold used in antitrust market definition analysis.

Connecting MA Data to Other Federal Healthcare Datasets

The MA data ecosystem connects naturally to several adjacent federal datasets:

Limitations and Research Cautions

Several limitations affect MA data research. The most significant is the partial disclosure of bid data: because individual plan bids are confidential, researchers cannot directly observe the profit margin embedded in plan bids or compare bids across competing plans in the same market. The publicly available benchmark and premium data allow inference about bid-benchmark relationships but not direct bid comparison.

The star rating system has been criticized for measurement issues that complicate quality comparisons. The CAHPS survey component, which measures member experience, is self-reported and may reflect demographic and socioeconomic differences in survey response patterns rather than genuine plan quality differences. CMS has introduced statistical adjustments for socioeconomic factors in recent star rating cycles, but the methodology remains contested.

Enrollment data has a 2–3 month publication lag, and monthly enrollment files are subject to revision as CMS processes late-arriving enrollment transactions. Point- in-time enrollment counts from early in the AEP may differ significantly from the final enrollment figures published after all transactions are processed.

Finally, the rapid growth of MA has created a policy environment in which the data itself is changing. New transparency reporting requirements for prior authorization, expanded RADV audit authority, and ongoing HCC model recalibrations mean that the policy and measurement context surrounding MA data is evolving faster than in most other federal data programs. Analysis of MA data should account for these regulatory changes when making year-over-year comparisons.

Medicare Part D prescription drug coverage is bundled into most MA plans as MA-PD. The CMS Part D prescriber public use files document drug utilization at the prescriber level and are analyzed separately. See Medicare Part D: Prescriber Data, Drug Spending, and Formulary Analysis.

Financial relationships between drug and device manufacturers and clinicians — relevant to understanding formulary and prescribing incentives in MA plans — are disclosed through CMS Open Payments. See CMS Open Payments: The Sunshine Act Database of Industry-Physician Financial Relationships.

Hospital quality measures, which bear on MA network adequacy assessments and care outcomes for plan members, are published through CMS Hospital Compare. See CMS Hospital Quality: Readmissions, Complications, and Patient Experience Data.