A behavioral-health appeal QA scorecard: review the evidence before the prose
A free appeal quality scorecard with must-pass gates, evidence mapping, source checks, a fictional draft review, and a named human approval checkpoint.
Direct answer
Behavioral health appeal quality assurance: what operators need to know
A free appeal quality scorecard with must-pass gates, evidence mapping, source checks, a fictional draft review, and a named human approval checkpoint. Review responsiveness to the actual denial reason. Require source support for material factual and clinical statements. Verify the applicable policy, version, service, and arrangement.
One denial · Plain read in 2 business days · A de-identified worksheet, not the letter
See the workflow: Prior-authorization workflow.
On this page: Direct answer

Review an appeal in this order: the decision being challenged, the available review route and clock, the evidence supporting each material statement, the applicable policy or criteria reference, the appropriate disclosure scope, and the actual version a named human approves. Style comes after those checks. A fluent letter with an invented event, wrong policy, or unsupported clinical assertion is not ready to send.
The free scorecard below uses must-pass gates and improvement checks. It is an original operating aid, not a validated clinical instrument, legal opinion, payer requirement, or prediction of appeal success. Use synthetic examples to train the process. Actual patient records, clinical judgments, coding, appeal rights, and submission decisions belong in your organization's approved workflow with appropriately qualified reviewers.
Key takeaways
The short version
- Review responsiveness to the actual denial reason.
- Require source support for material factual and clinical statements.
- Verify the applicable policy, version, service, and arrangement.
- An unresolved must-pass gate blocks approval regardless of writing quality.
- Approval names the reviewer, packet version, and next transmission owner.
Take the template with you
Free to copy · no email required
Appeal review scorecard
Mark each gate pass, revise, or unresolved; keep source evidence in approved systems. The score does not predict coverage or payment.
Check,Gate type,Result,Secure evidence reference,Reviewer,Required revision,Revision owner Correct decision and scope,Must pass,,,,, Applicable route and clock verified,Must pass,,,,, Material facts supported,Must pass,,,,, Clinical argument reviewed by authorized clinician,Must pass,,,,, Policy reference verified and applicable,Must pass,,,,, Disclosure and recipient approved,Must pass,,,,, Actual packet version approved,Must pass,,,,, Reason-to-evidence mapping clear,Improvement check,,,,, Requested outcome specific,Improvement check,,,,, Follow-up owner and receipt plan clear,Improvement check,,,,,
Saves as behavioral-health-appeal-qa.csv
Try the workflow · fictional demo case
Find the sentence that blocks approval
A fictional draft says the payer received a renewal on Monday and ignored a policy requirement. The record contains an upload attempt on Monday, no receipt, and an unverified policy citation.
Select a check to reveal the finding and next action. Nothing is sent to a payer.
0 of 3 checks reviewed. A named person still approves the next action.
This fictional example illustrates an offered Marsa workflow. We do not connect this public example to a payer or present it as a customer result. We measure demo starts and completions using fixed labels; no input text is collected.
Explore authorization, appeals, benefits, claims, audit, and coordination automations1. Name the decision and the requested change
The first review question is what the appeal asks the reviewer to change. Identify the actual adverse decision, the relevant service scope, the stated reason, and the requested outcome in the secure case record. A letter that argues generally for behavioral-health care may still leave the payer unable to identify which determination is disputed. The operating summary should make the issue clear without copying sensitive details into a public tool.
Read the complete notice and relevant response context. A code, portal label, or internal note can omit the specific reason or instructions. Check whether the issue concerns a clinical determination, an administrative condition, a claim reporting error, or another process. The selected route must answer the established problem. An appeal scorecard should not turn every unpaid item into a clinical argument merely because the team has a letter template available.
Write the desired change in one sentence before reviewing the draft. For a fictional case: review the retained amendment against the unpaid service because the response appears to rely on the earlier decision. That sentence is a working hypothesis to verify, not a conclusion that the payer made an error. It gives the reviewer a precise claim to test against the evidence and a clear standard for whether the draft is responsive.
2. Verify the review route and clock
Record the source for the selected review route, submission requirements, recipient, and deadline. The relevant process can differ by payer arrangement, product, service, denial type, and jurisdiction. Keep internal preparation targets separate from external requirements. An unexplained date copied from another case is not a dependable control, even if it gives the team a comfortable working margin.
Ask the authorized reviewer to resolve an unclear route before the packet is finalized. A correction, information response, reconsideration, appeal, or another process may require different content and evidence. Where a legal or contractual interpretation is involved, route it to the appropriate qualified person. The scorecard records the question and approval; it does not decide the rights or requirements for a real case.
Preserve the original timeline when a route changes. If the team first attempted a correction and later determines that a formal review is needed, record the reason for that change and the source supporting the next action. Do not assume that an earlier contact restarted or preserved every clock. The reviewer should be able to explain the sequence from retained events rather than rely on a final status that hides the earlier work.
3. Check every material statement against its source
A material statement is one that could change how the request is understood: a service date, a transmission event, an authorization condition, a clinical fact, a quoted policy, or a description of the payer's action. For each one, identify the approved source and whether the wording accurately reflects it. An evidence index makes this review easier by connecting the statement or argument to the supporting location.
Use three states: supported, needs revision, and unresolved. Supported means the designated reviewer can locate evidence appropriate for the statement. Needs revision means the source supports a narrower or different claim. Unresolved means the required support has not been established. Do not treat the absence of a contradiction as proof. A plausible statement can still be unsupported, especially when it was carried forward from an earlier draft.
When AI assists with drafting, review invented specificity carefully. A draft may convert attempted upload into received request, turn a general policy summary into an exact requirement, or add a confident explanation of the payer's motive. Those changes can make prose sound persuasive while making the record less accurate. The human reviewer should remove or correct unsupported detail and preserve the question that still needs an answer.
4. Keep clinical review with the authorized clinician
The operations reviewer can check whether a clinical argument has a source and designated owner, but cannot substitute a checklist for professional clinical judgment. The appropriate clinician reviews the relevant facts, reasoning, and support under the organization's process. The scorecard should record who accepted that responsibility and which version was reviewed. A generic clinical reviewed flag without a reviewer or version is difficult to rely on later.
The draft should address the actual stated reason using the applicable evidence. More clinical language does not necessarily create a stronger response. Ask whether the argument explains the point in dispute, whether the source supports it, and whether the requested outcome follows from the reviewed reasoning. If the available evidence does not support the proposed assertion, keep that gap visible instead of asking the writer to make it sound more convincing.
Do not reproduce licensed clinical criteria or imply that an AI model has authoritative access to them without verification. Use the organization's approved source and access arrangements, with the applicable version and scope. A public educational worksheet can name the need for a verified reference without publishing proprietary content or establishing medical necessity for an individual. The boundary protects both the quality of the argument and the integrity of the review.
5. Verify policy applicability, not just that a link opens
A policy reference needs more than a working URL. Check the issuing organization, product or arrangement, service scope, effective period, and relevant section. A document can be genuine and still not govern the case being reviewed. Record the basis for using it, and identify any conflict with the actual notice or other applicable instructions. The qualified reviewer determines the interpretation required for the response.
Distinguish a direct quotation from an editorial summary. If quoting is appropriate, ensure the words and context are accurate and used within the organization's permitted access and disclosure practices. If summarizing, avoid presenting a broad principle as an exact requirement. The review should make it possible to tell which statement comes from the source and which part is the organization's argument about its application.
Keep a checked date and version reference. A policy page may change after the draft is prepared, while the relevant question concerns an earlier period. Preserve the applicable source through approved methods and retention procedures. Do not silently update the citation in a historical packet to today's version. A dependable appeal record shows the reference the reviewer actually used and the reasoning supporting that choice.
6. Review a fictional draft with two defects
Harbor Example's fictional draft says a renewal was received on Monday and that a payer policy required a particular response. The source record shows an attempted portal upload, an error message, and no located acknowledgment. The policy citation came from an unsourced internal note. The prose is clear and the requested outcome is understandable, but two material assertions fail the evidence gate.
The submission owner verifies the acknowledgment question and revises the event description to match the established evidence. The policy owner locates the applicable source and asks the qualified reviewer to confirm its scope and interpretation. Until those actions are complete, the draft stays in review. The team does not average the two failures into an otherwise high quality score or let a deadline concern silently remove the checkpoint.
Once the facts and reference are established, the reviewer checks the revised argument, packet index, recipient, disclosure scope, and actual version. The next action identifies who will transmit it through the approved channel and who will verify receipt. The fictional case demonstrates a review method, not a payer error or a successful appeal. No customer recovery result is implied by completing the exercise.

7. Use must-pass gates before improvement scores
The downloadable scorecard has seven must-pass gates: correct decision and scope, verified route and clock, supported material facts, authorized clinical review where relevant, applicable policy reference, approved disclosure and recipient, and approval of the actual packet version. Mark each pass, revise, or unresolved. Your organization may adapt the gates to its process. The example is not an industry standard or a validated readiness measure.
Improvement checks can then address clarity, evidence mapping, a specific requested outcome, and a useful follow-up plan. Those checks help make a supported packet easier to review. They cannot compensate for a failed must-pass gate. Avoid a percentage score that lets several strong cosmetic checks outweigh one unsupported material statement. The approval decision should remain explicit, with the unresolved issue and responsible person visible.
Agree the meaning of each result before using the scorecard across reviewers. Two people should review the same synthetic draft and compare their decisions. If one marks supported and another marks unresolved, inspect the evidence definition and role boundary. The exercise should improve consistency, not pressure reviewers to agree with the most senior person or produce a reassuring green report.
8. Record the approved version and follow-through
Approval should identify the reviewer, role, packet version, scope of review, and approval event in the organization's system. If different people approve clinical content and transmission readiness, keep those responsibilities distinct. The final submission owner needs to know which checks are complete and which action is authorized. A comment saying looks good can be ambiguous when the packet contains several versions or unresolved questions.
Connect the approved version to the transmission and acknowledgment evidence. If an attachment changes after approval, follow the organization's process for reviewing the revision. If a channel fails, preserve the attempt and verify the applicable contingency route. The case remains owned through the next expected response. Sending a reviewed packet is meaningful progress, but it is not the same as a favorable decision or reconciled payment.
When the response arrives, evaluate it against the actual requested outcome and retained evidence. Record what changed, what remains unresolved, and the next approved action. Do not classify every acknowledged appeal as a success or every unfavorable result as proof that the review process failed. A quality process can improve factual integrity without guaranteeing an outcome controlled by the payer and the applicable requirements.
Calibrate reviewers with a sentence-level exercise
Give two reviewers the same synthetic source and draft sentence. The source says an upload was attempted at 10:15, an error appeared, and acknowledgment was not located. The draft says the payer received the packet at 10:15. Ask each reviewer to mark the sentence supported, revise, or unresolved and explain the source boundary. The operating finding is that the draft asserts receipt beyond the available evidence.
A useful revision describes the established attempt and preserves the acknowledgment question. The owner then identifies the next verification action. Do not reward a reviewer for replacing the sentence with an equally unsupported statement that the payer never received anything. Both receipt and non-receipt can remain unresolved after an error. This exercise tests factual precision rather than writing style.
Record disagreements in the review definition, then repeat the exercise with a policy-scope mismatch and an attachment-version change. The purpose is consistent evidence handling. It is not a certification, a clinical competency assessment, or a predicted appeal outcome.
9. Turn review findings into better source work
Classify recurring defects at a process level: unsupported event, incorrect scope, missing clinical review, unverified citation, unclear request, packet-version confusion, or receipt gap. Use counts with denominators and inspect a sample. A rise in detected defects after introducing the scorecard may mean the team can finally see problems that were previously hidden. Compare stable definitions before drawing a trend conclusion.
Assign the smallest upstream improvement. If drafts repeatedly confuse transmission and receipt, fix the submission-state definitions. If evidence is present but difficult to locate, improve the index. If policy scope is unclear, improve the maintained rule register and review ownership. Asking writers to be more careful is less useful than repairing the source control that feeds the same error into every draft.
Marsa Health offers appeal drafting and evidence-review automation. The free concierge review uses de-identified worksheets, and public demonstrations use fictional cases. A named human approves any real payer action and your team handles submission. The scorecard shows the review standard we want the workflow to make visible. You can use it freely, inspect the sample case, or request a demo without providing patient records. Implementation confirms the approved data handling and operating scope.
Common questions
Answers before you build.
Does a high score predict an appeal win?
No. This is an editorial review aid. Must-pass gates help establish a supported, reviewable packet; the payer outcome depends on the actual case and applicable requirements.
Can an AI draft pass without human review?
A named authorized human reviews the material facts, clinical argument where relevant, sources, recipient, and actual version before payer submission.
What happens when a source is missing?
Mark the item unresolved, assign an owner and next action, and keep the relevant approval gate open. Do not invent the missing support.
Practical closeout
Use this operator checklist.
- Review responsiveness to the actual denial reason.
- Require source support for material factual and clinical statements.
- Verify the applicable policy, version, service, and arrangement.
- An unresolved must-pass gate blocks approval regardless of writing quality.
- Approval names the reviewer, packet version, and next transmission owner.
Continue through the cluster
Verified customer case studies are added only with customer permission and supporting evidence; none is implied by these operational examples.
Sources & methodology
Trace the operational claims.
Marsa Health Editorial reviewed the primary and research sources below on November 4, 2026. We translate them into workflow controls, distinguish proposals from final rules, and flag where plan, program, state, contract, or clinical requirements vary.
- 01Health care payment and remittance advice CMSPrimary background on payment responses. The scorecard is an original operating aid, not a payer appeal rule.Accessed or rechecked November 4, 2026
- 02Minimum Necessary Requirement U.S. Department of Health and Human ServicesPrivacy background for appropriate information handling; actual disclosure questions require the organization's qualified review.Accessed or rechecked November 4, 2026
- 03How to appeal an insurance company decision Centers for Medicare & Medicaid ServicesFederal overview of internal appeals, external review, notices, and general appeal timing. Plan and state rules can differ.Accessed or rechecked November 4, 2026
- 04External Appeals Centers for Medicare & Medicaid ServicesFederal external-review process and consumer protections.Accessed or rechecked November 4, 2026
- 05CMS Interoperability and Prior Authorization Final Rule CMS-0057-F Centers for Medicare & Medicaid ServicesCurrent implementation dates, decision timeframes, denial-reason requirements, metrics, and API provisions for impacted payers.Accessed or rechecked November 4, 2026
Organizational author. Editorial review covers source accuracy, search intent, workflow boundaries, and human-oversight requirements. This material is educational and does not provide clinical, legal, coding, or coverage advice.
No named clinical or legal expert reviewer is attributed to this version. Marsa Health does not invent reviewer credentials.
Read our editorial methodRevision history
What changed and when
November 4, 2026
Initial publication, source review, and operational editing.