Medical coding sits between what you documented and what you get paid. Code too conservatively and you leave money on the table. Code past your documentation and you have created an audit exposure that outlives the payment. Our medical coding services put a certified coder between the signed note and the claim, so the codes match the chart in both directions.
Coded from your documentation, not your superbill
Certified coders assign CPT, ICD-10-CM and HCPCS Level II codes from what is actually in the note. A superbill tells us what the provider meant to bill. The note tells us what can be defended. Where the two disagree, the note wins and the provider hears about it before the claim leaves, not after a payer asks. Diagnoses are coded to the highest specificity the record supports, because an unspecified code where the note names laterality or type is one of the quieter reasons a claim pays late or not at all.
E/M levels under the 2021 rules
Office visit levels are chosen on medical decision making or total time on the date of the encounter, and nothing else. A 99213 needs low-complexity MDM or 20 to 29 minutes. A 99214 needs moderate MDM or 30 to 39 minutes, which usually means a chronic condition that is worsening, a prescription drug being managed or data from outside records being reviewed. A 99215 needs high MDM or 40 to 54 minutes. We read for those elements, not for the length of the note, and we flag the two habits that cost the most: providers who bill every follow-up as a 99213 out of caution, and templates that generate a 99214-sized note for a 99212 problem. Medicare visits also get checked for the G2211 add-on where a longitudinal relationship is documented.
Modifiers and edit pairs
Modifier 25 and modifier 59 attract the most payer scrutiny and the most CO-97 bundling denials, so we use them only where the record shows a separately identifiable E/M or a distinct procedural service. Every code pair on the claim is checked against the NCCI procedure-to-procedure table and the MUE unit limit for that quarter, and where the pair is allowed with a modifier we confirm the documentation supports it rather than adding the modifier to get paid. Imaging and diagnostic services get the 26 or TC split that matches who owned the equipment and who read the study. Telehealth visits carry modifier 95 or the place of service the payer wants, POS 10 or 02 for Medicare and often POS 11 with 95 for commercial plans. A CO-4 denial means we got one of these wrong, and we track them by coder.
Written queries, tracked to an answer
When a note does not support a code, or two readings are possible, the encounter is held and a query goes to the provider in writing. It states what is missing, for example a time statement or the status of a chronic condition, and never suggests the answer. Held encounters sit on a query list with a date and the provider's name, and that list is part of your weekly report so nothing sits unanswered. When the provider amends the note, the encounter is coded from the amended record and the query and the answer stay attached to it in case a payer or an auditor asks later. Queries that repeat become a training topic, which is how the documentation improves instead of the query list growing.
Who does the coding, and how it is checked
Coders are credentialed through AAPC or AHIMA and assigned to your practice by specialty, so the person reading a cardiology note codes cardiology every day rather than once a month. A second coder reviews a sample of coded work each week, and disagreements are settled against the documentation and the guideline in force on the date of service, not by seniority. Findings from that review feed a quarterly feedback report per provider: which documentation elements are routinely missing, which service lines are consistently under-coded relative to the note, which modifiers are drawing denials. That is the part that changes next quarter rather than only cleaning up last quarter. Where you want an independent baseline, a coding audit of a defined chart sample can be scoped separately, and it is usually more useful before we start than after.
What most medical coding services do not tell you
Two things. First, coding quality shows up in the remittance, not in an accuracy percentage. A claim that clears the scrubber and pays is still a coding failure if a 99214 visit went out as a 99213, so we read the 835 for that pattern and report it as revenue, not as a score. Second, code sets move. CPT changes every January, ICD-10-CM every October 1 and NCCI every quarter, and a coder working from last year's book will produce clean-looking claims that deny. We load each update before its effective date, and when a claim from March is corrected in September it is coded by the rule that applied in March. That is also why our claim submission team and our coders sit in the same weekly review: a rejection that traces back to a code goes back to the coder, with the claim, and the fix is made where the error started.
What Is Included
- CPT, ICD-10-CM and HCPCS Level II assigned from the signed note
- E/M level selection under the 2021 MDM and time rules
- Modifier 25, 59 and 26/TC review against NCCI edits
- Written provider queries, tracked to an answer
- Second-coder review and quarterly feedback by provider
| Edit standard applied | NCCI procedure-to-procedure pairs and medically unlikely edits, updated quarterlySource: CMS, National Correct Coding Initiative |
|---|---|
| Modifier-related denial code | CARC 4: the procedure code is inconsistent with the modifier usedSource: X12, Claim Adjustment Reason Codes |
| Place of service codes | POS 11 is office, POS 10 is telehealth in the patient home, POS 02 is telehealth elsewhereSource: CMS, Place of Service Code Set |
Medical Coding Services: Common Questions
Yes. Every coder holds a current AAPC or AHIMA credential, most commonly CPC or CCS, and is assigned to your practice by specialty. Coding is not done by billers assigning codes as a side task during charge entry.
It comes back to the provider as a written query before billing, stating exactly which element is missing. We will not code past what the note supports, and we will not silently down-code work you actually did to make the claim safer.
Coding rules are specialty-specific, so tell us your specialty and payer mix and we will tell you honestly whether it is one we handle well. Family practice, internal medicine, behavioral health, physical therapy, pain management, cardiology and dermatology are the ones we code most.
From the signed note in the EHR wherever we have access to it, because a superbill records what the provider intended to bill rather than what the documentation supports. Where only superbills are available, we say so and code accordingly.
In writing, through whatever channel your providers already check, and never as a leading question that suggests the answer. Queries are tracked on a list with dates so an unanswered query does not quietly hold up a claim.
Read access to the EHR, your specialty and payer mix, your current fee schedule and a named provider contact for queries. The first two to four weeks run in parallel with your current process so you can compare the codes before you switch.
CPT changes every January, ICD-10-CM every October and NCCI every quarter. We load each update before its effective date and code by the rule in force on the date of service, which matters when a claim is corrected months later.