Clinical documentation improvement is described as a hospital program, usually staffed by nurses who read charts before the claim drops. The same logic fits a ten-provider practice. The work is smaller: a monthly sample of notes, three query templates and one short report.
What this covers
- Measure the E/M level mix, the denial reason codes and the unspecified diagnosis rate before you change a single template.
- Most lost money sits in five note patterns: carried-forward history, unnamed data review, unspecified diagnoses, unstated time and unedited template text.
- A query asks, it does not direct. Name the clinical indicator, offer an open answer, and put the reply in the record in the clinician’s own words.
- Track a response rate by clinician. A query nobody answers is a note that stays incomplete.
- Cloned text and missing signatures lose claims on medical necessity, and those denials are harder to appeal than a coding edit.
If it is not documented, it has not been done.The Medicare documentation principle that decides most medical necessity denials.
Since January 1, 2021, an office or outpatient visit level comes from medical decision making or from the total time the practitioner spent on the date of the encounter. History and physical examination still belong in the note, when they are clinically appropriate, and they no longer select the level. That change moved the requirement from volume to reasoning, and reasoning is what a thin note fails to show.
What CDI Looks Like Without a CDI Team
A hospital documentation specialist reads the record before the bill goes out and queries the physician while the patient is still in the building. A practice cannot staff that and does not need to. It needs a reviewer who reads a sample of notes before billing, a written query process, and a monthly report.
The first half of the job is completeness. Medicare’s documentation principles ask for the reason for the encounter, relevant history and examination findings, the assessment and the plan, and warn against using the volume of documentation to decide the level billed.
Every claim gets two questions: does the record support the level billed, and was the service reasonable and necessary. Medicare expects the specific sign, symptom or complaint to appear for each service billed. A chart audit shows which clinicians answer both.
The second half is money. Each pattern carries a cost: a level billed below the work performed, a claim denied for lack of medical necessity, a rework cycle on a record request.
Find the Patterns That Cost You Before You Write Anything
Four reports answer most of the question, and all four come out of the practice management system you already run. Pull twelve months so one unusual month does not set the agenda.
| What to pull | Where it lives | What it usually shows |
|---|---|---|
| E/M level mix by clinician | Level distribution report | One clinician at level 4 or 5 for most visits while peers sit near half draws payer attention |
| Denial and adjustment reasons | Remittance advice CARC codes | Medical necessity denials cluster by clinician, payer and diagnosis family |
| Twenty most billed diagnoses | Diagnosis frequency report | Unspecified codes mean the specificity never reaches the claim |
| Modifier 25 frequency | Claim level modifier report | Heavy use without a separately identifiable service invites review |
Read the mix against the case load, not a target. A practice managing complex chronic disease should sit higher than a clinic treating sprains, and Medicare is explicit that a higher level is not appropriate when a lower one fits the patient.
A program that pushes every note up one level is a billing risk, not documentation improvement. If a level moves, the note should show what earned it: a medication started or stopped, a test interpreted, a risk recorded.
The Five Note Gaps Behind Most Lost Reimbursement
Chart review keeps returning to the same handful of sentences. They are honest notes that leave the work invisible, and a coder cannot code what the note does not show.
| The weak line | What it fails to show | A line that carries the level |
|---|---|---|
| Refilled current medications. | A decision, a dose change, any risk considered | Metformin 1,000 mg twice daily continued; A1c 7.8 in December, started empagliflozin, discussed risk |
| Reviewed labs, no changes. | Which test, which value, which decision followed | Reviewed CMP from March 2: creatinine 1.4, eGFR 42; held NSAIDs, repeat in three months |
| Discussed with cardiology. | Whether a management decision was made, and by whom | Called Dr. Alvarez in cardiology: agreed to hold the beta blocker, begin titration, call if weight rises |
| Diabetes and hypertension, stable. | The specificity the diagnosis code needs | Type 2 diabetes with stage 3 chronic kidney disease, and hypertension, both at goal |
| Same review of systems as the prior note. | Whether the information was obtained today | Denies chest pain or dyspnea today; two episodes of nocturnal dyspnea since May 1 |
Those gaps decide which of two adjacent codes fits. A 99214 calls for moderate medical decision making, or 30 to 39 minutes of total time on the date of the encounter, and the note text is the evidence for one or the other.
Time is the second route to a level. Only time spent on the date of the encounter counts, and time spent writing the note on a later day does not.
For office and outpatient visits the level comes from medical decision making or total time on the date of the encounter. Under medical decision making the note must meet or exceed the level in two of three elements: the number and complexity of problems addressed, the amount and complexity of data reviewed and analyzed, and the risk of complications, morbidity or mortality.
Medicare’s answer is direct: you would not order testing the patient does not need to meet a level of service. Every test counted as part of the visit must be medically necessary on its own.
How to Write a Query a Clinician Will Answer
A query is a question about the record, not an instruction about the code. ACDIS and AHIMA, the associations that publish the standards, require queries to be non-leading so the clinician answers from their own judgment.
Two diabetes medications are documented and nothing says whether the diabetes is at goal or a complication is present.
Quote the data: A1c 7.8 in December, eGFR 42 on March 2, two episodes of nocturnal dyspnea reported today. A question without indicators is a guess.
Ask what the condition is and how it is managed, with an option for the clinician to write their own answer, the construction ACDIS and AHIMA expect of a multiple choice query.
The answer belongs in the record, attributed and dated, so a coder codes from the clinician’s own statement.
Update the record, recheck the code, note the outcome. If no answer arrives inside your window, escalate once to the medical director.
A query can clarify, add specificity or resolve conflicting information in the record. It cannot invite a diagnosis the record does not support, and it cannot be worded so one answer is obviously wanted. Clean medical coding starts from the clinician’s own words.
Read the query with the clinical indicators removed. Could a reasonable clinician answer that the finding is not clinically significant, or that they cannot determine it from today’s visit? If not, the query is leading.
Make Answering Cheap for the Clinician
Query programs fail on friction, not on clinical disagreement. Most clinicians agree with the premise. They stop answering when the question costs four minutes and arrives a month late.
- One question per note. A ten-item query is a survey and gets ignored.
- Send it during the encounter or the next morning, not three weeks later.
- Route it to the clinician who wrote the note, never through the front desk.
- Track four numbers: notes reviewed, queries sent, queries answered, and answers that changed the code.
- Publish the wins, such as a denied claim reversed because the note named the functional limitation.
Those four numbers are the dashboard. Published query benchmarks come from inpatient work measured per discharge and do not transfer to an office practice, so your own baseline is the comparison.
Show the clinician the note text, the payer response and the amount at issue, one case at a time, and keep peer rankings out of it. A program that feels like a scoreboard gets the shortest sentence a clinician can write. Medical coding and billing support works best when both sides see the same case.
Cloned Notes, Missing Signatures and the Audits That Follow
Two record integrity problems cause the worst documentation denials, and both begin as convenience features. Palmetto GBA, a Medicare contractor, defines the first plainly: cloning occurs when documentation is the same from one patient to the next, and it fails the medical necessity requirement because it lacks specific, individual information.
Each note has to stand on its own for its own date of service, and the differences from the previous visit carry the weight. Auto-filled examination text, a carried-forward review of systems and a pasted assessment block all read as boilerplate.
The second problem is authentication. Medicare requires services provided or ordered to be authenticated by the author with a handwritten or electronic signature, and stamped signatures are not acceptable. When a signature is illegible, a signature log or an attestation can identify the author, and that attestation must be signed and dated by the author of the entry.
When a reviewer asks for a signature log or attestation, the review extends by 15 days and the practice has 20 calendar days to respond. A missing signature on an order means the reviewer proceeds as though the order was never received. Keep a current signature log for every clinician, locums included.
Enforcement interest follows the same themes. The HHS Office of Inspector General keeps live work plan projects on evaluation and management services, including one on visits billed the same day as minor surgery without modifier 25. Modifier 25 is not the problem by itself. The problem is appending it where the note does not document a significant, separately identifiable service.
A Thirty Day Start for a Small Practice
The first month is diagnostic. Do not rewrite templates or issue policy. Collect evidence and one narrow set of questions.
Twelve months of level distribution, denial reasons by CARC code, the twenty most billed diagnoses, and modifier 25 frequency.
Score each note against the level billed: is the condition specific, is the reviewed data named, is the risk stated, is time recorded if time selected the level.
Write one template for each of the three most common gaps, with clinical indicators built in, and have a clinician read the wording.
Record the four numbers, agree a response window, book a twenty-minute monthly review, and name one person who owns the queries.
Keep the sample small and the cadence steady. Ten notes per clinician each month finds a pattern and still gets finished. The same reviewer, the same three queries and the same four numbers, month after month, is the whole method.
Documentation questions practices ask most
It is the routine of reviewing a sample of notes against the level of service billed and the coverage rules, then asking clinicians to clarify anything the record does not show. Three pieces make it work: a monthly chart sample, a written query process, and a four-number dashboard.
No. The ACDIS and AHIMA guidelines require queries to be non-leading, so the clinician answers from their own clinical judgment. A compliant query points at a gap or a conflict in the record and cites the clinical indicators behind the question. A message naming the code you want is not a query.
Not on the strength of the list alone. A code needs support in the record for the encounter where it is reported, which usually means the condition was monitored, evaluated, assessed or treated that day. A problem list is an index kept for care continuity, and the encounter note is the evidence.
Four: notes reviewed, queries sent, queries answered, and the share of answers that changed the code or the level. Add the dollar value of documentation-related denials by reason code and the share of claims carrying an unspecified diagnosis. Your own baseline is the comparison that matters.
A certified coder or an experienced auditor can run the review, and most small practices start there. Clinical judgment stays with the clinician, which is the reason a compliant query asks rather than decides. With no coder in house, an outside chart audit sets the baseline.
Because the reviewer reads the note, not the patient. When a note fails to state the specific sign, symptom or complaint that made the service reasonable and necessary, the payer’s only evidence is a thin record. Medicare contractors treat cloned or boilerplate documentation as a failure of medical necessity.
It matters differently. Risk adjustment pays plans based on documented conditions, and those charts are audited, so a diagnosis needs support at a qualifying encounter in the payment year. Chronic conditions do not carry over from one year to the next, and the same habits support visit levels.
The bottom line
Clinical documentation improvement in a practice is a sample of notes, a handful of queries and a report that fits on one page. The patterns that cost money repeat: unspecified conditions, unnamed data, unstated time and text carried forward from a prior visit. Fix the three that show up most in your own charts, measure whether the answers arrive, and repeat the cycle next month.
How much of last quarter did your notes leave on the table?
Send us de-identified notes with the levels billed, your denial reasons by code, and your top diagnosis codes. We will show you the documentation patterns costing you money, write the first three query templates with your team, and hand back a one-page dashboard you can run every month.
Request a free documentation reviewEvaluation and management rules, documentation requirements and payer policies change with each payment year, and commercial policies differ from Medicare. Confirm the current guideline and the payer’s own medical policy before you change how a service is documented or billed.
