Methodology: Managed-Care Operating Metrics
What these figures are
The Operating Metrics card on a managed-care insurer’s stock page shows the operating statistics a health insurer reports in its own quarterly earnings release — the SEC Form 8-K exhibit filed alongside the results. These are the numbers payers use to describe how much of each premium dollar went to care, how many members they cover, how fast medical costs are rising, and how efficiently they run. Oxford Ledge reads them directly from the as-filed exhibit and normalizes the labels so you can compare an insurer to itself over time.
They are an Oxford Ledge Signal: our reading of a public filing, not a re-audit of the company’s books and not a licensed data-vendor feed. Every value on the card carries its filing as-of date and links back to the SEC source.
The metrics, in plain English
| Metric | What it measures |
|---|---|
| Medical loss ratio | The share of premium dollars paid out as medical claims. An 85% MLR means 85 cents of every premium dollar went to care. The single most important payer metric — a rising MLR means medical costs are outrunning the premiums set a year earlier. |
| Medical membership | The number of people enrolled at period end — the base that generates premium. Reported by line of business (commercial, Medicare, Medicaid) and by base: total medical vs risk-based (members whose costs the insurer bears) vs fee-based. |
| Medical cost trend | The year-over-year rate at which per-member medical costs are rising — healthcare inflation as the payer experiences it. The number that most directly threatens earnings, because premiums were priced a year ahead against a forecast of it. |
| Operating cost ratio | The share of revenue spent on administration and selling — overhead, technology, commissions — distinct from claims. Lower signals scale and efficiency. |
| Days in claims payable | Outstanding unpaid claims divided by average daily medical cost, in days (often 45–55). A gauge of reserve adequacy: a sudden drop can flatter earnings if the insurer is under-reserving. |
Why every value shows a base
Two insurers can print the same-named metric on different terms, and the same insurer prints more than one version of it. A bare number would hide that, so the card labels it:
- MLR definition. A medical loss ratio calculated the GAAP-reported way (claims over premium revenue as booked) differs from the ACA-regulatory ratio, which uses a statutory numerator and denominator with quality-improvement adjustments. They are different numbers; we treat GAAP-reported as the headline and badge the regulatory one where an insurer leads with it.
- Membership base. Total medical membership counts everyone; risk-based counts only members whose medical costs the insurer bears; fee-based (administrative-services-only) accounts are ones where the employer bears the cost and the insurer just processes claims. Risk-based is where the premium and the risk concentrate, so we badge which base a membership figure is on and compare only like with like.
- Scope. The card shows the consolidated (whole-company) figure; a single line-of-business or reporting-segment total never stands in for the consolidated number.
These badges are the honesty contract of the card: they exist precisely because the same word can mean different things, and a reader deserves to know which one they are looking at.
Why the peer table leaves cells blank
On the industry peer table, each column pins one definition — the medical loss ratio on the GAAP-reported basis, membership on the total-medical base. An insurer is shown in a column only if it reports the metric on exactly that basis. An insurer that discloses, say, only an ACA-regulatory MLR or only a risk-based membership leaves an honest blank in that column — never a value pooled in from a different definition, and never a fabricated zero. A blank is a coverage gap, and the column footnotes how many insurers it covers.
Why membership is four rows, not one number
Every one of these companies publishes a single headline membership figure, and those figures are not comparable with each other. They count different populations: some fold in members the company only administers on an employer’s behalf, some fold in drug-only plans, and each company draws the line where its own reporting draws it. Ranking those totals would produce a table in which every individual number is true and the ordering is false, so we do not publish that ranking. Instead we compare membership at the level where the definitions actually line up — the line of business.
Four rows, each a different kind of customer and a different kind of economics:
- Members it administers only. A large employer pays its own employees’ medical bills and hires the insurer to run the plan — process claims, build the provider network, handle service. The insurer earns an administrative fee per member and bears no medical risk. This row is the one that most distorts a headline total, because these members can be counted alongside insured members while generating a fraction of the revenue.
- Medicaid members. State coverage for people on low incomes, run under contract by a private insurer at a rate the state sets.
- Medicare Advantage members. A private plan replacing Original Medicare, paid for by a federal rate set annually.
- Medicare drug-only plans (Part D). Prescription coverage bought separately from medical coverage. A Part D member is not a medical member, which is why this row is never added to the others.
Two rows carry a base that is not the plain member count, and where that happens the cell says so. A figure marked at-risk only counts the members whose medical costs the company actually bears, rather than everyone enrolled. We show both bases in the same row rather than dropping one, because a blank teaches less than a labeled figure — but the label travels with the number so the two are never read as the same thing.
A blank here means the company does not report that line in the exhibit we read, or reports it on a basis that would not be comparable. It is never a zero. Where a company’s own line covers more than one of these categories at once — a Medicare figure that also includes Medicare Supplement, for instance — we leave the cell blank and say why above the table rather than place a number in a column it does not belong to.
Some rows we were asked for and deliberately did not build. A dual-eligible row (people who qualify for both Medicare and Medicaid) is not published: those members are, by definition, already counted inside both the Medicare Advantage and Medicaid rows, so a third row would double-count them. A row subtracting drug-only members from a risk-bearing total is not published either, because the components the companies disclose are not known to be exhaustive, and a subtraction over an incomplete set produces a number with no defined meaning. Where a government count of the same population exists, it is compared on the Managed Care page rather than here; the two are counted differently and neither is corrected against the other.
How we read and normalize it
For each insurer we fetch the most recent quarterly earnings 8-K exhibit from SEC EDGAR, locate the operating-metrics table, and read each value verbatim from the reported column — never a prior-year or year-to-date column standing in for the quarter. Membership is read as a period-END count, not a quarterly flow. We map the insurer’s label (“Benefit ratio”, “Medical care ratio”, “Total medical membership”) to a common metric name and record its base (GAAP vs ACA, total vs risk-based) alongside every figure.
The reading is checked against a hand-verified golden set for each insurer before any figure is allowed to publish; a value whose base the extractor cannot confidently determine is withheld rather than shown on a guessed default.
What we deliberately do not do
- We do not re-audit. The figures are as reported by the issuer under regulatory convention. We normalize labels; we do not restate the company’s numbers.
- We do not pool different definitions. A GAAP MLR and an ACA MLR, a total-medical and a risk-based membership, are different numbers; the peer table compares only like with like and leaves the rest blank.
- We do not show a value we are unsure of. Any reading our extractor flags — an out-of-band ratio, an ambiguous base, a mismatched scale — is dropped from the display, never shown greyed or asterisked. Honest or absent.
- We do not treat these as forecasts, ratings, or advice. They are historical operating results a company disclosed, presented for study. Medical cost trend in particular is an issuer estimate, carried as directional context.
Freshness
The stock page is cached at the edge, so a brand-new filing’s figures can lag up to a day on the cached page. Each value is stamped with the filing it came from, so the as-of date always tells you exactly how current the number is.