Atlas by MatchMed

About the scores

Every number on Atlas is derived from publicly available government data. No surveys, no self-reporting, no recruiter claims. Here's exactly what we measure and how.

What Atlas measures

Atlas measures physician movement, not practice quality. For any practice, we show a single verifiable thing: of the physicians who have appeared on its Medicare roster since 2019, how long each one stayed.

That is arithmetic on public record. Atlas does not know, and does not claim to know, why any physician joined or left. The scores are designed to help a physician decide what to ask before signing a contract — not to reach a conclusion on their behalf.

Data source

Medicare Part B Provider Data, Centers for Medicare & Medicaid Services (CMS). This dataset captures physician-practice affiliations across annual snapshots. Atlas scores are calculated from 2019 onwards, using the complete longitudinal record of every physician who has appeared on a practice's Medicare roster. No proprietary, self-reported, or third-party data is used. Practices cannot edit, remove, or influence what appears in this data.

Atlas is not affiliated with, endorsed by, or sponsored by the Centers for Medicare & Medicaid Services or any other federal agency.

What the score is built to protect against

A raw departure count would be misleading. Before any score is calculated, Atlas accounts for the ordinary, benign reasons physicians move on, so that a practice is not marked down for normal career events:

The core idea

Every physician who joins a practice either stays or leaves. A long stay is, on balance, a signal of a working relationship. A departure is the end of one. How long a physician stayed before leaving carries information worth weighing — a very short stay and a stay that ends right around the partnership stage are different signals, and Atlas weights them differently.

Atlas does not assign a reason to any departure. It records how long physicians stayed and surfaces the patterns, so a physician evaluating an opportunity knows where to ask harder questions.

The scores

Each score is scaled 0–100. Higher reflects stronger physician retention.

Attrition Resistance
Primary component

Measures the weighted departure load of physicians who have left the practice. Rather than a simple exit count, each departure is weighted by how long the physician stayed. Shorter stays carry more weight, because they are more likely to reflect something worth investigating.

Tenure at departureWeightWhat it suggests to ask
0 – 2 yearsHighestEarly exits are worth asking about directly
2 – 4 yearsHighAsk what the path looked like on the way to partnership
4 – 6 yearsModerateAsk how partnership and growth decisions were made
6 – 10 yearsLowAmbiguous. Many ordinary explanations
10+ yearsMinimalNear-career-stage. Treated as low signal
Any individual departure may have a personal explanation, and Atlas assigns none. Across thousands of practices and eight years of data, idiosyncratic reasons average out. A practice with a repeated pattern of departures at a specific career stage is worth a closer look — whatever the explanation turns out to be.
Tenure Strength
Secondary component

Measures how long physicians currently at the practice have stayed. Calculated from the active roster only. Physicians who have already left no longer contribute to this metric.

Departed physicians should not continue to prop up a practice's score after they leave. Tenure Strength reflects the current workforce: the physicians who would actually be your colleagues if you join.
Cluster Signals
Score modifier

Two pattern-level signals that apply a downward adjustment when triggered. Both respond to patterns across multiple physicians, not to any single departure.

Temporal cluster

Fires when a concentration of departures occurs within a defined rolling window. This often coincides with a discrete event such as a leadership change or acquisition, and is worth asking about. Applies a downward adjustment.

Tenure similarity cluster

Fires when multiple physicians departed after a similar length of stay. A repeated exit point at the same career stage is a pattern worth understanding. Applies a downward adjustment.

Experience Level
Context only

Measures the collective seniority of the current roster. Derived from median years since medical school graduation. Displayed as context alongside the Retention Score but not included in the composite.

Experience level reflects practice maturity. Senior-heavy rosters can carry succession or transition considerations that would be incorrectly rewarded if folded into a retention composite, so this is shown as context only.
Retention Score
Composite

A weighted composite of Attrition Resistance and Tenure Strength, adjusted by cluster signals.

Retention Score =
  [(Attrition Resistance × primary weight)
  + (Tenure Strength × secondary weight)]
  × Cluster Modifier (if applicable)

Specific component weights are proprietary.

Insights & Observations

Practice detail pages include a short insight derived from the same CMS-derived fields shown in the metrics. Insights are deterministic observations, not AI-generated advice.

Insights are
  • Factual observations from CMS data
  • Descriptions of tenure and retention patterns
  • Transparent about their data sources
Insights are not
  • Predictions of future performance
  • Risk assessments or warnings
  • Prescriptive recommendations
  • Clinical or business advice

Each insight follows this logic:

  1. Extract facts from CMS data (tenure distribution, roster size, churn rate)
  2. Identify patterns (for example top-heavy tenure or significant roster reduction)
  3. Describe the observation factually
  4. List assumptions that could affect interpretation
Concentrated long-tenure workforce

Observation: Longer-tenure cohorts exceed newer cohorts with a high retention score.

What it describes: Concentrated experience and historical stability.

Assumption: Long tenure may correlate with stability.

Alternative interpretation: Could reflect limited growth, geographic constraints, or market conditions.

Significant roster reduction

Observation: Current roster size is 1 while all-time historical size exceeds 3.

What it describes: Observed reduction in active physician count.

Assumption: Reduction is factual and measurable.

Alternative interpretation: Could be natural transition, acquisition, consolidation, or retirement.

High churn rate

Observation: More than 40% of all-time physicians exited within 4 years.

What it describes: Elevated short-tenure exit rate.

Assumption: High exit rate may reflect a challenging environment.

Alternative interpretation: Could be early-career rotation, competitive market, or voluntary transitions.

Insufficient data

Observation: Fewer than 2 all-time physicians.

What it describes: Lack of historical data for pattern analysis.

Assumption: None.

Alternative interpretation: Absence of data is itself factual.

Key principle: Atlas shows what the data reveals about physician tenure patterns. It does not predict outcomes or prescribe actions.

What the score is not

The Retention Score is not a measure of clinical quality, patient outcomes, compensation, or workplace culture, and it is not a judgment of any practice or any physician. It answers one question: based on how physicians have historically moved through this practice, how well has it retained the people who joined it?

A lower score is not an accusation. It is a prompt to ask better questions. Use Atlas as one input in your due diligence, alongside site visits, peer conversations, and contract review with an attorney.

For practices: data and corrections

Atlas scores are derived from CMS Medicare Part B data, which may occasionally contain inaccuracies. If you represent a practice and believe your data or score reflects an error, we want to correct it. Flagged practices are reviewed against the underlying CMS record, and confirmed data errors are corrected in the next update cycle.

Email admin@matchmed.app with subject line Data Inquiry – [Practice Name]. We aim to respond within five business days.

Legal & Methodological Disclaimers

All scores are derived from CMS Medicare Part B public datasets. MatchMed makes no representations regarding the completeness, accuracy, or timeliness of underlying CMS data.

Scores are statistical estimates derived from observed physician movement, not factual declarations about any practice. They do not establish causation or assign a reason for any physician's departure.

MatchMed, LLC is not liable for any employment, contracting, or other decisions made in reliance on Atlas scores. Use is subject to our Terms of Service and Privacy Policy.