Take me through what a payer does with a claim after you submit it, and how you keep denials from piling up.

Two different failures, and they are not the same thing

Before the payer's logic runs at all, the claim has to be accepted. A submitted claim passes through the clearing house and the payer's front door, where it is checked for structural validity and for whether it can be identified: is the transaction well formed, does the member identifier resolve, is the provider recognised, are the required fields present. Failing here produces a rejection. The claim was never adjudicated, no decision has been made about coverage, and no appeal rights arise because there is nothing to appeal. You fix and resubmit.

A denial is different. The claim was accepted, adjudicated, and a decision was reached that the payer will pay nothing or less than billed. That decision has reasons attached, and it carries appeal rights and appeal deadlines.

Candidates who use the two words interchangeably then give wrong answers about remedy, timing and metrics, because a rejection is a data problem you own entirely and a denial is a disagreement with a counterparty. Keeping them separate is the first signal.

What adjudication actually does

Adjudication is a rules pipeline, and it runs roughly in this order because each stage depends on the last.

It identifies the member and confirms coverage was active on the date of service, which is the date that matters rather than the date of submission. It identifies the provider and determines whether they are in network for this member's plan, because that changes the fee schedule and the member's cost share. It checks whether the plan covers this benefit at all, and whether any limit has been consumed: a visit cap, an annual maximum, a frequency limit on a screening test.

It then applies clinical edits. Is the procedure consistent with the patient's age and sex? Is the diagnosis one for which this service is considered medically necessary? Are two procedure codes on the claim ones that should not be billed together, or is one already included in the other, so that billing both amounts to unbundling? Was authorisation required and obtained? Is this a duplicate of a claim already paid?

Only then does it price. Pricing applies the contracted rate or fee schedule to arrive at the allowed amount, applies any grouping methodology for an inpatient stay, and splits the allowed amount between the payer's liability and the member's deductible, copayment and coinsurance. If another plan is primary, coordination of benefits determines the order and this payer pays only its secondary share.

The output is a remittance advice, returned electronically, that explains the outcome line by line rather than in aggregate. Each line carries adjustment reason codes saying why the paid amount differs from the billed amount, with supplementary remark codes adding detail. A claim can be partly paid, so reading the remittance at claim level and not at line level is a common and costly mistake: the money you are missing is often on one line of an otherwise paid claim.

Eligibility checking is a snapshot, and its expiry is the problem

An eligibility enquiry asks the payer whether this member is covered, under which plan, with what benefits and what accumulated cost share. It is real-time and it is cheap, which is why the correct place for it is at scheduling and again at the point of service.

The subtlety is that it answers as at the moment you ask. Coverage terminates when employment ends, plans change at renewal, dependants age out, and a member can have coverage that is active today and retroactively cancelled for non-payment of premium. So an eligibility response obtained three weeks before a procedure is evidence of diligence, not a guarantee of payment, and the practical discipline is to re-verify close to the date of service and to record the response you relied on with its timestamp. When a payer later denies for eligibility, the stored response is your appeal.

It is also the cheapest denial prevention available. Eligibility and registration errors are consistently among the largest denial categories and are almost entirely avoidable at the front desk, where correcting them costs a minute rather than a month.

Prior authorisation, and the thing it does not promise

For some services the payer requires approval before the service is delivered. The provider submits the clinical justification, the payer reviews it against its medical policy, and returns an approval with an authorisation reference, usually scoped to a specific service, a specific quantity or number of visits, a specific provider or facility, and a validity window.

The scoping is where authorisations fail. The procedure performed turns out to be a different one from the one approved because the surgeon found something different once they were in. The authorisation covered six sessions and the twelfth was delivered. It named one facility and the patient was treated at another in the same group. It expired before the patient could be scheduled. In each case there was an authorisation and the claim is still denied.

And the point that separates a strong answer: an authorisation is a determination of medical necessity, not a promise of payment. It does not override eligibility, so if coverage has lapsed by the date of service the claim is denied regardless. It does not override benefit limits or coordination of benefits. It does not survive the service being coded differently from what was requested. Anyone who tells a patient that an authorisation means the plan will pay has made a commitment the plan has not made.

Emergency care sits outside this, since authorisation cannot be a precondition for treatment that cannot wait; those cases are notified after the fact instead.

Denial management as an improvement loop, not a queue

Working denials one at a time is how organisations spend permanently. The discipline that works treats the denial population as data.

Categorise every denial by root cause using the reason codes plus your own investigation, and separate the categories into preventable and unpreventable. Eligibility, missing authorisation, non-covered service under a plan you could have checked, untimely filing, and duplicate submissions are preventable, and their fix lives upstream in registration, scheduling or the claim scrubber. Medical necessity disputes and downcoded services are contested, and their fix is documentation and appeal. Contract underpayments are neither: they are a payer holding to a rate other than the one you modelled, and the fix is contract management.

Then work the two ends differently. Preventable categories go to process owners with a measured rate, and success is the rate falling. Contested denials go to appeals, with the clinical documentation attached, and the useful metric is overturn rate by payer and by category, which tells you which fights are worth having.

Two deadlines govern all of this and neither is negotiable. Timely filing limits mean a claim not submitted within the payer's window is not payable at all, however valid it was. Appeal windows run from the determination date, so a denial sitting unworked in a queue quietly becomes a write-off. This is why denial inventory is measured by age rather than by count, and why the oldest item in the queue is the number a manager should be watching.

The metric that ties it together is first-pass yield: the proportion of claims paid correctly without human touch after submission. It is the honest measure because every point of it is work that did not have to happen, and it moves almost entirely through front-end data quality rather than through hiring more people to appeal.