EDGE Server Business Rules | Health Data Max
ACA / Marketplace Risk Adjustment

EDGE Server business rules, made operational.

Every enrollment, medical, pharmacy, and supplemental diagnosis file you send to the HHS-operated EDGE server is checked against hundreds of rules before a single record counts toward risk adjustment. Here's how that pipeline works — and how Health Data Max catches the rejects before CMS does.

Reference: CMS / CCIIO EDGE Server Business Rules (ESBR) and the EDGE Server Interface Control Documents (ICD)

The four inbound files
EnrollmentFull replacement
Medical claimsIncremental
Pharmacy claimsIncremental
Supplemental diagnosisIncremental
4
Inbound file types
4
Verification edit categories
9+
XML validation levels
100%
Of paid records must be accepted
Why the rules matter

Only accepted records get paid.

Issuers in states where HHS operates a risk adjustment program submit their data to an EDGE server, where it is verified level by level. Accepted records are the only records eligible for Risk Adjustment (RA) and High-Cost Risk Pool (HCRP) calculations — so a rejected claim isn't just an error to fix later, it's revenue and compliance risk sitting outside the model. CMS receives only aggregated, summarized results; the detailed accept/reject reporting comes back to you, and reconciling it is on you.

The four inbound files

Different files, different submission logic.

The single most important operational distinction: enrollment is a full replacement each time, while all three claim files are incremental. Sending a claim file as a "full" resubmission gets those claims rejected as duplicates.

FILE 01Full replacement

Enrollment

A complete replacement file inclusive of all enrollees — each submission fully supersedes the prior one. CMS recommends monthly submission.

  • Subscriber vs. non-subscriber / dependent requirements
  • Enrollment start/end dates and periods crossing calendar years
  • Dual and overlapping coverage handling
  • Premium amounts, partial-month premium, and disenrollment
  • Enrollment Period Activity Indicators (EPAI) and mandated coverage
FILE 02Incremental

Medical claims

The most rule-dense file. Submitted incrementally; header-level and line-level checks both apply, and full-file resubmission triggers duplicate rejections.

  • Claim header ↔ claim line dependencies
  • Header and line duplicate checks (with modifier exceptions)
  • Voiding and replacing previously submitted claims
  • FFS vs. capitated and mixed-claim submission
  • Overlapping inpatient stays, interim billing, cross-year claims
FILE 03Incremental

Pharmacy claims

Incremental submission with its own duplicate, void, and replace logic distinct from medical claims.

  • Header, issuer, and plan-level rules specific to pharmacy
  • Duplicate pharmacy claim detection
  • Claim Processed Date Time rules
  • Voiding and replacing pharmacy claims
  • FFS and capitated claim submission
FILE 04Incremental

Supplemental diagnosis

Incremental add/delete records that adjust the diagnoses tied to previously accepted claims — a common source of quiet rejects.

  • Acceptable sources, including diagnoses from health assessments and medical records
  • Duplicate supplemental detail record checks
  • Adding and deleting supplemental diagnosis codes
  • Voiding supplemental diagnosis detail records
  • Linkage back to the original accepted claim
How a file is processed

The verification pipeline, start to finish.

File processing is designed to evaluate as many data elements as possible before rejecting anything. It moves top-down through the XML — and a failure at the header level stops the file before it's ever archived.

STEP 01

File type verification

Structure is checked first. Pass, and a Job ID is assigned; fail, and no Job ID is created — you get a System Error (SE) report.

STEP 02

Header level

Pass and the file is archived and continues; fail any header check and the whole file is rejected and not archived.

STEP 03

Issuer & plan levels

Issuer- and plan-level verifications confirm the submitting entity and plan references before record detail is evaluated.

STEP 04

Detail levels

Enrollee, enrollment period, and claim header/line segments each get required, face-validity, referential, and logical edits.

STEP 05

Outbound reports

An ESFAR report is produced for every file, plus detail and summary accept/reject reports with specific reject codes.

Verification edits

Four edits, applied in two stages.

Every data element on every file runs the same gauntlet. Understanding the order is the key to fixing rejects fast — because an element that fails the first stage never even reaches the second.

01

Required

Confirms a non-null value is present in the data tag. Elements are Required, Situational, or Not Required per the ICD.

02

Face validity

Confirms the value conforms to the specified data type and restrictions (format, length, allowed characters).

03

Referential

Confirms the value matches an entry in the common reference data table set — codes, plans, and identifiers must exist.

04

Logical

Confirms the value satisfies the defined business logic — the cross-field, date, and dependency rules that make up most of this document.

Order of operations
1
Required + face validity firstEvery element is checked for presence and format. Anything that fails here is marked reject and goes no further.
2
Referential + logical nextOnly elements that passed stage one are checked against reference data and business logic.
Because stage two depends on stage one, a single missing or malformed field can mask deeper problems you won't see until the first is fixed and the record is resubmitted. Reject codes and descriptions are published in the REGTAP Library and returned on your outbound files.
Recurring rule themes

Where submissions actually break.

Across all four files, the same categories of business logic account for most rejections. These are the areas worth validating before you submit — not after the reports come back.

Duplicates

Header- and line-level duplicate checks, plus the specific modifier and two-claim exceptions that determine when otherwise-identical lines are allowed.

Voids & replacements

Each file has its own multi-step void and replace sequence. Doing them out of order — or as a full resubmission — creates duplicate and orphan rejects.

FFS vs. capitated

Fee-for-service, capitated, and mixed-claim submissions follow different rules — and misclassification changes how a claim is processed and selected.

Dates & cross-year

Claim Processed Date Time, overlapping inpatient stays, interim billing, and enrollment periods that cross calendar years all carry precise date logic.

Enrollment integrity

Subscriber and dependent requirements, dual and overlapping coverage, premium changes, disenrollment, and EPAI values.

Supplemental linkage

Supplemental diagnosis adds and deletes must tie to accepted claims from acceptable sources — health assessments and medical records included.

Where Health Data Max fits

Validate and correct before the file leaves your hands.

Health Data Max gives your team a working interface for all four inbound files — applying the same required, face-validity, referential, and logical checks the EDGE server will, so issues surface while they're still cheap to fix.

01 / VALIDATE

See rejects before CMS does

Every enrollment, medical, pharmacy, and supplemental record is checked against EDGE business rules in a clear UI — before submission, not after the accept/reject reports come back.

02 / CORRECT

Guided corrections in-line

When a record trips a rule — a duplicate, a broken header/line dependency, an out-of-order void — the platform explains what failed and how to fix it, mapped to the actual reject code.

03 / RECONCILE

One database, continuously in sync

Inbound submissions and outbound EDGE reports live in one place, so accepted vs. rejected records reconcile automatically and nothing paid-eligible slips outside the model.

See what your EDGE submissions are leaving behind.

We'll run a complimentary evaluation of your risk adjustment data and show you where records are rejecting — and what that's worth. One recent pilot surfaced close to 40% of unclaimed revenue.

Book a demo →