RADV Audit Preparation: Pre-Submission Scrubbing Strategies

What Is Pre-Submission Scrubbing

RADV audit preparation through pre-submission scrubbing catches documentation issues before they become audit findings. Pre-submission RADV scrubbing is the systematic validation of HCC-mapped diagnosis codes and their supporting clinical documentation before encounter data is submitted to CMS for risk adjustment processing. The objective is straightforward: identify and address unsupported or problematic diagnoses before they generate risk-adjusted payments that would later need to be returned in a RADV audit.

Scrubbing operates on a simple principle: it is far less expensive to prevent an overpayment than to repay one. Every unsupported diagnosis that reaches CMS creates a liability — a payment the plan received that it may be required to return with interest, potentially extrapolated across its entire membership under current RADV rules.

• Scope: Scrubbing reviews all HCC-mapped diagnoses in encounter data files before submission, focusing on codes that carry the highest risk coefficients and the highest historical error rates
• Methodology: The process applies coding guideline validation, documentation sufficiency checks, provider credential verification, and statistical pattern analysis to every qualifying diagnosis
• Output: Diagnoses that pass scrubbing proceed to submission. Diagnoses that fail are either removed, corrected with proper documentation, or flagged for additional clinical review before resubmission
• Timing: Scrubbing occurs after coding is complete but before the encounter data submission deadline, creating a quality control checkpoint between the coding workflow and CMS submission

Plans that implement pre-submission scrubbing consistently report 40-60% reductions in RADV-eligible audit findings compared to their pre-scrubbing baselines. The investment pays for itself by avoiding recoveries that would otherwise be multiples of the scrubbing program cost.

Why Scrubbing Beats Remediation

The economics of pre-submission scrubbing versus post-audit remediation are not even close. Every dollar spent on prevention eliminates ten to fifty dollars in potential remediation cost.

• No Repayment Required: Diagnoses removed during scrubbing never generate an overpayment, so there is nothing to repay. Post-audit findings require full repayment of the risk-adjusted amount plus interest from the date of original payment
• No Extrapolation Exposure: Under current RADV rules, CMS can extrapolate audit findings across a plan's full membership. A finding of $500,000 in sampled overpayments can become tens of millions in extrapolated liability. Scrubbing eliminates the findings that trigger extrapolation
• No Compliance Penalties: Submitting unsupported diagnoses can trigger False Claims Act investigations and civil monetary penalties. Removing unsupported diagnoses before submission demonstrates good faith compliance effort
• Documentation Improvement: The scrubbing process surfaces documentation gaps that can be corrected before submission. A diagnosis flagged as insufficiently documented can often be supported through a provider addendum or supplemental documentation request, converting a would-be failure into a properly supported HCC
• Predictable Costs: Scrubbing costs are predictable operational expenses. Audit remediation costs are unpredictable, potentially catastrophic, and often accompanied by legal fees, consultant costs, and operational disruption

The RADV audit checklist provides a framework for audit preparedness, but scrubbing represents the proactive complement — ensuring that the data submitted to CMS is defensible before an audit is ever initiated.

The Scrubbing Process

Effective scrubbing follows a structured workflow that balances thoroughness with operational efficiency. The process must run quickly enough to meet submission deadlines while catching the issues that matter most.

• Data Extraction: Pull all HCC-mapped diagnoses from the encounter data file scheduled for submission. Include member demographics, provider identifiers, encounter dates, and all associated diagnosis codes per encounter
• Rule-Based Validation: Apply automated validation rules that check ICD-10 coding guidelines via an ICD-10 guideline auditor API, validating code-encounter type combinations, provider credential requirements, and date-of-service logic. This layer catches technical coding errors
• Statistical Pattern Analysis: Flag diagnoses that represent statistical outliers — providers coding certain conditions at rates significantly above peer benchmarks, sudden increases in specific HCC codes, or coding patterns inconsistent with the member's utilization history
• Documentation Sampling: For flagged diagnoses, pull supporting clinical documentation and validate that the record contains the elements required for RADV defensibility: face-to-face encounter, qualifying provider, definitive diagnosis language, clinical indicators, and treatment plan
• Disposition Decision: Each flagged diagnosis receives a disposition: pass (submit as-is), correct (fix coding error and submit), supplement (request additional documentation and resubmit), or remove (delete from submission file)
• Provider Notification: Notify providers whose diagnoses were flagged or removed, including specific documentation guidance for future encounters. This feedback loop drives the documentation improvement that reduces future scrubbing volumes

The scrubbing process should be integrated into the regular submission workflow rather than treated as an ad hoc exercise. Plans with the lowest RADV exposure scrub every submission cycle, not just annual sweeps.

Common Issues Caught by Scrubbing

Pre-submission scrubbing reveals the same categories of issues that OIG and CMS find during actual RADV audits — but catches them before they become findings. The lessons from OIG audit findings directly inform what scrubbing should prioritize.

• Specificity Mismatches: Diagnosis codes requiring greater specificity than the documentation supports. Example: coding Type 2 diabetes with diabetic chronic kidney disease (E11.22) when documentation states "diabetes" and "CKD" as separate unrelated conditions without establishing the causal relationship
• Non-Qualifying Encounter Types: HCC diagnoses coded from encounters that do not meet RADV's face-to-face requirement — lab-only visits, telephone consultations, or encounters billed under non-qualifying provider types
• Stale Problem List Entries: Conditions appearing on claims that trace back to problem list entries without current-year clinical evaluation. Automated scrubbing can detect when a diagnosis code appears on every encounter for a provider who uses problem list auto-population
• Duplicate Submissions: The same HCC condition submitted through multiple encounter records for the same member in the same payment year, creating unnecessary audit surface area without additional RAF benefit
• Coding Guideline Violations: Codes that violate ICD-10 Official Guidelines — such as coding acute conditions as chronic, assigning codes that require additional characters, or using combination codes when separate codes are required
• High-Risk Provider Patterns: Providers whose coding patterns for specific HCCs exceed peer averages by more than two standard deviations, suggesting potential systematic coding errors that warrant individual chart review

Technology for Automated Scrubbing

Manual scrubbing of every HCC diagnosis is not feasible at scale. Technology-enabled scrubbing applies automated validation to the full submission volume and routes only flagged items for human review.

• Rule Engine: A comprehensive library of validation rules covering ICD-10 guidelines, CMS encounter requirements, and RADV documentation standards. Rules should be updated with each CMS guidance release and OIG audit report
• NLP Documentation Review: Natural language processing systems that scan clinical documentation to verify that submitted diagnoses have corresponding narrative support, clinical indicators, and treatment documentation
• Statistical Modeling: Machine learning models trained on historical RADV audit outcomes that score each diagnosis by its probability of failing validation, enabling risk-based prioritization of human review resources
• Workflow Management: Case management functionality that routes flagged diagnoses to appropriate reviewers, tracks disposition decisions, manages documentation requests, and enforces review completion before submission deadlines
• Provider Reporting: Automated generation of provider-specific scrubbing reports showing flagged diagnoses, removal reasons, and documentation improvement recommendations
• Audit Trail: Complete documentation of every scrubbing decision — what was reviewed, what was flagged, what action was taken, and who made the decision. This audit trail itself serves as compliance evidence if the plan is selected for RADV

Technology should handle 85-90% of the scrubbing volume through automated validation, reserving human review for the 10-15% of diagnoses that require clinical judgment. This ratio makes comprehensive scrubbing economically viable even for large plans with millions of annual HCC submissions.

Building a Scrubbing Program

A sustainable scrubbing program requires organizational commitment, clear governance, and continuous improvement processes that drive down error rates over time.

• Executive Sponsorship: Scrubbing removes diagnoses from submissions, which temporarily reduces reported RAF scores. Leadership must understand that this short-term revenue reduction prevents far larger long-term audit liabilities and support the program accordingly
• Dedicated Team: Assign certified coders and clinical reviewers specifically to the scrubbing function. Mixing scrubbing responsibilities with chart review or coding production creates competing incentives that undermine scrubbing effectiveness
• Phased Implementation: Start by scrubbing the highest-risk categories — conditions with the highest RADV failure rates in OIG reports and the highest RAF coefficients. Expand coverage as the program matures and technology capabilities increase
• Provider Education Integration: Connect scrubbing findings directly to provider education programs. When scrubbing reveals that a specific provider's diabetes documentation consistently fails validation, targeted training addresses the root cause rather than the symptom
• Continuous Calibration: Regularly validate scrubbing accuracy by comparing scrubbed submissions against actual RADV outcomes. If diagnoses that passed scrubbing still fail audits, the validation rules need refinement. If diagnoses that were removed would have passed, the rules are overly conservative
• ROI Measurement: Track the financial impact of scrubbing by calculating avoided RADV exposure (diagnoses removed multiplied by their coefficient value and extrapolation multiplier) against program costs. Mature programs show 10:1 to 50:1 return on RADV prevention investment

The ultimate measure of a scrubbing program is not how many diagnoses it removes — it is how the plan's RADV error rate declines over time. A successful program drives its own obsolescence as documentation quality improves and fewer diagnoses require scrubbing intervention.