When Diagnosis Codes Outrun the Note: A Mental Health Audit

A diagnosis code can be valid, billable, and still wrong for the encounter. In mental health care, that mismatch often appears when a code describes recurrence, severity, or remission more precisely than the clinician’s assessment supports. The reverse happens too: a clinician documents a clear diagnosis, but the claim carries a vague code copied from an earlier visit. Both situations deserve attention. Coding audits should identify unsupported detail and missed supported detail—not simply push every diagnosis toward a more specific code. The goal is straightforward: report the most specific ICD-10-CM diagnosis supported by the record, applicable coding rules, and the circumstances of the encounter. ## Specificity Is an Evidence Standard, Not a Contest More characters do not automatically make a code better. Specificity means selecting the appropriate code within the relevant category using documented clinical distinctions. For mental health diagnoses, those distinctions may include: - Single versus recurrent depressive episodes. - Current episode severity. - Presence or absence of psychotic features. - Partial or full remission. - Substance involved and the documented substance-related condition. - Acute, chronic, or unspecified presentation where the classification provides those options. Not every diagnosis family uses every distinction. Review the current ICD-10-CM Alphabetic Index and Tabular List, including inclusion terms, exclusion notes, and sequencing instructions. Confirm that the code set applies to the date of service. An unspecified code is not automatically an error. It can be appropriate when the available documentation does not establish a narrower diagnosis. However, repeated unspecified coding despite clear assessments may reveal a workflow problem. ### Outpatient uncertainty matters In outpatient settings, do not code diagnoses documented as probable, suspected, questionable, rule out, or similar expressions of uncertainty as though established. Report the signs, symptoms, or other reasons for the encounter to the highest degree of certainty supported, following applicable guidelines. A working differential is clinically useful. It is not permission to turn every possibility into a confirmed claim diagnosis. ## Four Mental Health Scenarios Worth Auditing The following examples are hypothetical. Code selection always depends on the complete record and current coding guidance. ### 1. Recurrent depression without documented recurrence **Hypothetical example:** A claim reports F33.1, major depressive disorder, recurrent, moderate. The signed assessment says only “depression,” while an old problem-list entry carries the more detailed diagnosis. **Audit response:** Determine whether current documentation establishes the recurrent, moderate diagnosis. Do not infer recurrence from multiple appointments or severity from medication dosage. If the record remains unclear, obtain clinician clarification rather than choosing a different diagnosis independently. ### 2. A screening score becomes a diagnosis **Hypothetical example:** An intake questionnaire shows elevated anxiety symptoms. A staff member selects generalized anxiety disorder before the clinician completes the assessment. **Audit response:** Confirm whether the clinician established that diagnosis. Screening results can inform clinical judgment, but a score alone does not establish the disorder. Review who can add diagnoses and how preliminary entries reach the claim. ### 3. Improvement is treated as remission **Hypothetical example:** A follow-up note says, “Sleeping better; fewer depressive symptoms.” The claim changes to a remission code. **Audit response:** Look for the clinician’s documented remission assessment. Improvement does not independently establish partial or full remission. The same caution applies when symptoms improve but the record does not support a change in severity. ### 4. An old diagnosis follows every visit **Hypothetical example:** A patient previously received treatment for an adjustment disorder. Months later, the diagnosis automatically appears on claims for encounters focused on a different condition, without documentation showing that it remains relevant. **Audit response:** Review whether the diagnosis is current and reportable for that encounter. A longitudinal problem list and an encounter diagnosis list serve different purposes; neither should populate the other without review. ## Run a Small, Repeatable Audit An effective audit does not require reviewing every chart. It requires a defined question, a consistent method, and follow-through. ### Step 1: Define the scope before selecting records Choose a manageable question, such as: “Do depression codes match documented episode pattern, severity, and remission status?” Specify the review period, clinicians, locations, encounter types, and whether you are evaluating submitted claims or pre-bill records. Include both initial evaluations and follow-up visits when relevant. Avoid mixing diagnosis specificity, procedure coding, and medical necessity into one undifferentiated score. They are related but require separate findings. ### Step 2: Combine random and targeted sampling Use a small random sample to examine routine performance, then add targeted records with higher-risk patterns: - Diagnoses that remain unchanged across many encounters. - Changes between unspecified and highly specific codes. - Remission codes without an obvious assessment update. - Multiple potentially conflicting diagnoses. - Diagnosis-related denials or repeated corrections. Label targeted results separately. A sample deliberately enriched for problems cannot establish the practice’s overall error rate. ### Step 3: Review the documentation behind each code For each encounter, capture the submitted diagnosis, relevant signed assessment, and the reason the code does or does not fit. Use this checklist: - [ ] Was the diagnosis established by an appropriate clinician? - [ ] Does the record support the code’s required distinctions? - [ ] Are contradictory statements unresolved? - [ ] Is an uncertain outpatient diagnosis being treated as confirmed? - [ ] Is the condition relevant and reportable for this encounter? - [ ] Were applicable tabular instructions followed? - [ ] Is the code valid for the service date? - [ ] Does the claim match the finalized encounter diagnosis? Do not use medication choice, appointment frequency, or a screening score to independently fill diagnostic gaps. ### Step 4: Classify findings by cause Separate unsupported specificity from supported specificity that was missed. Also distinguish outdated carry-forward diagnoses, invalid codes, conflicting documentation, and claim-entry errors. These categories lead to different fixes. Clinician education will not resolve a claim interface that substitutes an old diagnosis. A software adjustment will not resolve an assessment that never states the diagnosis clearly. ## Clarify Without Steering the Clinician When documentation is incomplete or conflicting, use a neutral query that identifies the issue and asks for clinical clarification. For example: “The assessment documents depression, while the encounter diagnosis lists recurrent major depressive disorder, moderate. Please clarify the diagnosis for this encounter, including episode pattern and severity if clinically established, or indicate that these cannot be determined.” Avoid asking clinicians to select a more severe diagnosis to satisfy coverage requirements. Keep clarification grounded in the existing clinical evidence and preserve the query and response under practice policy. If a submitted claim needs correction, route it through the established billing and compliance process. Do not silently rewrite a signed note. Any amendment should preserve authorship, timing, and the original record; it must not manufacture support for previously unsupported coding. ## Put Technology Behind the Review, Not in Charge IKON EMR’s AI scribe, telemedicine, billing, and patient portal capabilities can support the surrounding workflow, but diagnosis selection still requires clinical judgment and coding review. An AI-generated draft should be checked for diagnostic qualifiers that the clinician did not establish. During telemedicine encounters, maintain the same documentation discipline used in person. Patient portal questionnaires provide useful history, but patient-entered labels should not automatically become confirmed diagnoses. Billing teams should reconcile claim diagnoses with the finalized assessment. HIPAA compliance also belongs in the audit workflow. Use IKON EMR’s HIPAA compliance capabilities alongside appropriate access controls, staff training, and organizational safeguards. Keep audit extracts limited to necessary information and avoid copying sensitive narrative into unsecured spreadsheets or email. Software alone does not make a practice compliant. ## Close the Loop With Focused Feedback Assign each finding an owner, corrective action, and follow-up date. Share a few de-identified examples with clinicians and coding staff, emphasizing the documented distinction rather than blaming individuals. Track separate measures: unsupported specificity, missed supported specificity, unresolved queries, and recurrence of previously corrected problems. Recheck comparable encounters after changes take effect. Do not set a blanket target to eliminate unspecified codes; that can reward unsupported diagnoses. ## Short FAQ ### Should every unspecified diagnosis trigger a query? No. Query when clarification is clinically supported and materially useful. Do not pressure clinicians to document certainty they do not have. ### Does an accurate diagnosis guarantee payment? No. Coverage, medical necessity, procedure selection, authorization, and other requirements still apply. Code accuracy is essential, but it is only one part of claim review. ### How often should a practice audit? Set a cadence based on risk and resources. Periodic small reviews, plus targeted checks after code updates or workflow changes, are often more actionable than an isolated large audit. The strongest coding audit asks one consistent question: does this encounter’s documentation support this diagnosis code? Keeping that question central protects clinical accuracy without turning specificity into an exercise in overcoding.