Where the leverage is

Not every cost lever is worth the same argument. This sheet ranks the ones that are, so a benefits committee can spend its time on the handful that move total plan spend rather than the ones that merely feel productive.

What does this add to the Iceberg Review?

The Iceberg Review tells you where your money already sits. It does not tell you what to do about it, and a reader who finishes it knowing their cost is concentrated is entitled to ask what comes next. This is what comes next.

The two run together. One finds the shape of your spend, the other ranks what can change it. Both sit under the Benefits Accountability Matrix, which is the four box canvas the series is built on.

Why start here rather than with an RFP?

Because replacing a network, an administrator or an advisor carries administrative load and member friction, and none of it is guaranteed to touch what is actually driving your cost. Working the levers is usually cheaper to try and faster to measure. Not always, and the sheet says so: a pharmacy contract review or a centers of excellence arrangement is its own project, and two of the top rows need a contract change before they move at all. The difference is that you will know which one you are buying, and if the arrangement really is the problem you will have the evidence to say so rather than the suspicion.

How is impact defined here?

As effect on total plan spend, not the saving on the claims a lever touches. That distinction is the whole ranking. A lever can cut 30% off everything it reaches and still be small, because what it reaches is small, and most of the sales material in this market quotes the first number while a plan sponsor is trying to budget the second.

Large means over 5% of total plan spend, moderate 1 to 5%, small under 1%. Those are starting estimates for a typical employer. Plan size and mix move them, which is why page two exists.

Why is there a separate column for evidence?

Because size and proof are different questions, and most lever lists collapse them into one. A lever can be large and badly measured. It can also be small and measured precisely. Payment integrity is the clearest case: every published recovery rate comes from a firm selling the service, so the honest mark is that the size is unknown rather than that it is small.

Reading the plan document is the other one. Nobody has measured what a full read turns up, and one unintended inclusion on a specialty drug can outrun several rows above it.

Which lever turns on the others?

Steerage. It is not a lever beside the rest so much as the thing that makes them work. Every site of care difference is worth zero until a member is routed to it, and members do not route themselves, because the saving belongs almost entirely to the plan. On the closest measured case, an ACL reconstruction, the member’s own saving from choosing the cheaper site was a median of eleven dollars.

Download the worksheet

Can you do this yourself?

Yes, and that is the point of publishing it. The Benefits Accountability Matrix is built so a plan sponsor can self source the answer rather than wait for somebody to present it. We would still encourage working through it with your advisor, since the judgment calls are where the value is, but nothing here is withheld to create a reason to call.

The prompt below turns a de-identified claims extract into every row of page two. Two cautions first. Use aggregate, de-identified data only, never a raw claims file in a public assistant, and run it inside whatever environment your organization has already approved for that data. If in doubt, hand the prompt to your administrator or analytics vendor and ask them to return the table.

I am working with a de-identified claims extract for one plan year of a US employer-sponsored group health plan. The file has, at minimum: a hashed member key, relationship code, place of service, procedure code (CPT or HCPCS), diagnosis code (ICD-10), DRG where present, national drug code or J code where present, billing provider or facility identifier, date of service, allowed amount and plan paid amount. Some fields may be missing. Tell me which ones before you start, and which rows below you cannot produce without them.

FIRST, STATE THE BASIS

Tell me whether the dollar totals you are about to use are ALLOWED amounts or PLAN PAID amounts, and use one consistently. State total members, total claimants and the total on that basis. If the file carries both, use allowed and say so.

THEN PRODUCE THESE ROWS

For each row give me the dollars, the share of total, and the measure named. Show the code sets or rules you used.

1. Infusion. Drug administration codes 96365 to 96379 plus the J codes billed on the same claim. Split by place of service into hospital outpatient (POS 19, 22) against office, home and independent clinic (POS 11, 12, 49). Measure: share of infusion dollars at hospital outpatient.

2. Imaging and outpatient surgery. Imaging CPT 70010 to 79999 and outpatient surgical procedures in CPT 10000 to 69999 with an outpatient place of service. Split hospital outpatient against ambulatory surgery center (POS 24) and office. Measure: share of dollars at hospital outpatient.

3. Specialty drugs. Specialty and injectable drugs on both benefits. Medical side: J codes and Q codes for drugs. Pharmacy side: whatever the file flags as specialty. Measure: share of specialty dollars billed on the medical benefit.

4. Pharmacy, non specialty. Pharmacy claims split brand, generic, and brand where a generic or biosimilar exists. Measure: brand share where a substitute exists.

5. High cost claimants. Members above $50,000 and above $100,000 for the year, counts and dollars. Measure: share of total above each line. These should match the Iceberg Review inputs.

6. Musculoskeletal. Primary diagnosis M00 to M99, split surgical episodes (a surgical CPT inside the episode) against conservative care. Measure: surgical share of MSK dollars.

7. COE eligible procedures. Joint replacement, spine fusion, cardiac surgery and catheterization, bariatric surgery and transplant, by DRG or CPT, listing the facilities used and dollars at each. Measure: share at any single facility.

8. Dominant system. Facility dollars grouped by hospital system where the facility identifier allows it. Measure: share going to the largest system.

9. Possible duplicates. Same member key, same date of service, same procedure code, same billing provider, paid more than once. Measure: dollars flagged. Call these possible, not confirmed.

10. Dependent spend. Dollars by relationship code. Flag any dependent still enrolled past the plan's age limit where age can be derived. If the file has no relationship code, say the row belongs to the enrollment file rather than claims.

CONSTRAINTS

- Do not estimate savings for any row. Report what was spent and where.
- Do not name or recommend any vendor, carrier, TPA, PBM or point solution.
- Where a code set is ambiguous, say what you assumed and how much it moves the row.
- If any row comes to less than 1% of total, say so plainly.
- End with the three rows carrying the most dollars, and one sentence on what a sponsor would want to know next about each.

Write the dollars and the share onto page two, then pick two rows from the top. Two, not eight. Each one gets an owner and a date, and the measure column is the number you check in ninety days.

What can a claims file not tell you?

Two rows on the front sheet cannot be answered from a claims file at all, and both are on the addendum for that reason. Reading the plan document end to end, and whether a near site clinic is being used to route toward anything. Neither is a spend category. Both are worth an afternoon.

Fringe Theory is independent and unaffiliated. Views expressed are my own and do not represent those of my employer. Nothing here is legal, tax, medical, or investment advice. Figures described as modeled or derived are my own arithmetic from the public sources listed, not disclosed data.