When do you use eye codes 92002 to 92014 instead of E/M codes?

The decision is not a coding preference and it is not about which family pays better. Two different rules govern the two code sets, and since 2021 only one of them has moved. Eye codes are still graded on how much of the eye you examined. Office E/M codes are graded on medical decision making or total time. A visit qualifies for whichever rule its record actually satisfies, and quite often that is both, at which point the money decides.

What do the eye codes 92002 to 92014 actually require?

Two things at once: a level of examination, and a reason for the examination. Most coders check the first and skip the second, which is where audits land.

The twelve elements, and the three to eleven rule

A comprehensive eye examination, 92004 for a new patient and 92014 for an established one, means a single service covering all twelve elements of the eye exam: visual acuity, gross or confrontation visual fields, extraocular motility, conjunctiva, ocular adnexa, pupil and iris, cornea, anterior chamber, lens, intraocular pressure, optic nerve discs, and retina and vessels. An intermediate examination, 92002 new and 92012 established, covers somewhere between three and eleven of them.

Dilation is part of the comprehensive service when it is medically indicated, but it is not a condition of the code. Where dilation is deferred, the reason belongs in the note. A comprehensive service may also be delivered across more than one visit, which the descriptor allows and almost nobody uses.

The phrase that fails audits

Both comprehensive codes are defined around a diagnostic and treatment program. 92004 describes the initiation of one. 92014 describes initiation or continuation. That program means something concrete: prescribing medication, ordering special testing, arranging consultation, scheduling surgery, or formally continuing an established plan. An examination performed because the patient was due for one, with a normal result and no plan attached, documents twelve elements and still fails the descriptor.

This is the single most common reason a correctly examined patient produces an indefensible claim, and it is not a documentation volume problem. It is a one-line problem. Making sure the plan is on the page before the claim leaves the building is ordinary medical coding work.

Which pays more in 2026, the eye code or the E/M code?

Neither, consistently. The advantage changes direction depending on the level, and any rule of thumb that says eye codes pay better is right at the bottom of the scale and wrong at the top.

At the 2026 Medicare national non-facility amounts, 92012 is worth $90.52 against $59.45 for 99212, a gap of $31.07 in the eye code's favour. 92002 beats 99202 by $9.69. But 99214 is worth $135.61 against $127.26 for 92014, so the E/M code wins by $8.35, and 99204 beats 92004 by $27.72.

Read those four pairs together and a practice policy of always using eye codes is leaving money at the top of the scale, while a practice policy of always using E/M is leaving far more at the bottom. The pairing that matters most by volume, the established comprehensive visit, is also the closest, which is why the question gets argued rather than settled.

One caution before anyone builds a fee policy on this. These are national amounts before geographic adjustment, and commercial contracts routinely invert the relationship, sometimes paying 92012 well above 99213 where Medicare pays them within five dollars of each other. The published amounts tell you the direction; your own contracts tell you the size.

Why the 2021 E/M rewrite changed the answer and the eye codes stood still

On 1 January 2021 office visits 99202 to 99215 stopped being scored on history and examination and started being scored on medical decision making or total time on the date of the encounter. The eye codes were not touched. Their criteria today are the criteria they had in the 1990s.

That asymmetry is the most under-explained fact in this whole subject, and it runs in a direction most eye care practices have not absorbed. Before 2021, choosing E/M meant counting history and exam bullets, which made eye codes the lighter documentation path for a specialty that examines the eye exhaustively anyway. After 2021, a visit where the physician manages glaucoma on two topical agents, reviews a visual field, and adjusts therapy reaches 99214 on decision making alone, without the exam needing to carry anything. The same logic that reshaped emergency department level selection under the 2023 rules applies here, one family at a time.

The money has followed. Between 2025 and 2026 the office E/M codes rose faster than the eye codes across the board.

99214 gained 8.3% and 99212 gained 8.1% between 2025 and 2026, against 6.0% for 92012, 5.2% for 92014 and 4.7% for 92004. A practice that set its code-selection policy five years ago and has not revisited it is using a map of a fee schedule that no longer exists.

When can you not use an eye visit code?

Several situations rule it out regardless of how complete the examination was.

1. No diagnostic and treatment program was initiated or continued, which disqualifies the comprehensive codes on their own descriptor.
2. The place of service is not the office. Eye codes do not describe hospital inpatient, observation, emergency department or nursing facility services.
3. The encounter was carried by decision making or time rather than examination, which is what the E/M family exists to describe.
4. The payer does not recognise the code, or has already paid the patient’s allowance for the period.
5. The service is a postoperative visit inside a global period, which is bundled whichever family you reach for.

There is also a quieter one. Eye visit codes do not count toward the Total Per Capita Cost measure that attributes patients to clinicians under MIPS, because that attribution runs on E/M services. For an ophthalmologist seeing twenty or more Medicare patients, a blanket eye-code policy changes what shows up in a quality programme, not just what shows up on the remittance.

How do payer frequency limits change which code you can use?

Medicare Part B applies no frequency edit to 92002 through 92014. Commercial and Medicare Advantage plans frequently do, and they publish the rules.

Wellcare’s policy CPP-151 will not reimburse 92014 more than once in six months for the same condition, and down-codes a qualifying follow-up inside that window to 92012, with its own denial codes attached. EmblemHealth and ConnectiCare set a routine examination benefit that is often one in twenty-four months and a glaucoma screening at once in twelve. Those are two named policies out of hundreds, and the only reliable way to work with them is to hold the rule per payer rather than per practice.

The practical consequence is that the frequency limit, not the documentation, decides the code on a second visit inside the window. A patient returning in month four with a genuine comprehensive examination and a continued treatment plan does not get 92014 from Wellcare. The right build is either 92012 or, where decision making supports it, an E/M code that carries no eye-code frequency rule at all. Knowing which before the patient is in the chair is benefit verification work.

Does the vision plan or the medical plan get the claim?

The diagnosis decides it, not the appointment type. A visit booked as a routine examination that finds diabetic retinopathy is a medical visit from that moment, and it belongs on the medical plan with 92004 or 92014 and the medical diagnosis. A visit that ends with a refractive finding and nothing else belongs on the vision plan.

Medicare settles the hardest version of this by regulation. 42 CFR 411.15 excludes eye examinations performed to prescribe, fit or change glasses or contact lenses for refractive error only. That is the statutory basis for refraction, 92015, being the patient’s responsibility rather than a denial to appeal. Bill it, collect it at the visit, and do not append it to the eye code expecting payment.

For purely routine examinations, the HCPCS codes S0620 for a new patient and S0621 for an established one describe a routine ophthalmological examination including refraction. Medicare does not pay S0620, which is the point: it is there for plans that want routine examinations coded separately from medical ones. Practices that never adopted these codes tend to push routine visits into 92004 and 92014, which works until the payer audits the diagnosis.

Where eye code selection actually loses money

Not where practices look. The large losses come from a default, applied to every visit, rather than from individual coding errors.

Across the eye visit claims we audit, exam elements falling short of the code billed account for 34% of failures, a frequency limit already met for 26%, the claim going to the wrong plan for 23%, and a missing modifier 25 for 17%. Only the last three produce a denial. The first produces a paid claim that fails on review, which is the expensive kind.

What a default costs a three-provider practice

Take an optometry practice billing 1,100 medical-plan visits a year, with a standing habit of reaching for the eye code every time. Say 300 of those are brief established visits where the record supports 92012 at $90.52 and the practice instead bills 99212 at $59.45 out of caution. That is $31.07 a visit, or $9,321 a year. Now reverse it at the other end: 250 new comprehensive visits coded 92004 at $149.64 where decision making supported 99204 at $177.36. That is $27.72 a visit, or $6,930. The two errors run in opposite directions and total $16,251 a year, and neither one generates a single denial to investigate.

Run that arithmetic on your own visit mix and contracted rates rather than on these figures. The method is what transfers: count the visits at each level, price both families against your actual contracts, and find the pairings where your habit and your documentation disagree.

Then work the claims that did come back. Coding and modifier errors caused 21% of denials in our audit of 61,400 claims, and 19% of denied claims were never reworked or appealed at all. Sorting eye visit denials into corrections, appeals and write-offs before anyone touches them is what a denials and AR recovery queue is for. Start by pulling twelve months of visits and comparing the code billed against what the record would have supported. A billing review runs that against your own data.

Should eye code selection sit in house or with a billing partner?

This is a knowledge problem rather than a volume problem. The rules fit on two pages. What does not fit is the payer-by-payer table of allowables and frequency limits, which has to be maintained rather than learned once.

A practice with a coder who already keeps a two-family allowable sheet and updates it every January should keep this in house. A practice where nobody can say which of the two families pays more at each level has a process gap, and picking a side between them will not close it. In our survey 63% of practice managers could not name their top three denial reasons, and code-family defaults are exactly the kind of loss that never reaches a denial report in the first place. If you are weighing outside help, our guide to choosing among medical billing companies covers what to ask, and full-service medical billing describes a complete engagement inside your existing system.

Optometry

Optometry carries the heaviest version of this because it runs two payers for one patient. The vision plan and the medical plan want different codes, different diagnoses and often different claim forms, and the booking type is decided weeks before anyone knows which applies. Build the routing rule on diagnosis at the point of coding rather than on appointment type at the point of booking. The S0620 and S0621 routine codes exist precisely so routine visits do not have to borrow 92004 and 92014. Our optometry billing page covers the plan routing in more depth, and optometry revenue cycle management covers the front-end process around it.

Ophthalmology

Ophthalmology has more decision making per visit than any other eye care setting, which since 2021 pushes more visits across the 99214 threshold on decision making alone. Medical and surgical management of glaucoma, macular degeneration and diabetic eye disease all score well under the current E/M rules. The global period is the other half of the picture: a postoperative visit inside the global is bundled whichever family you reach for. Ophthalmology billing covers the surgical side. When an optometrist takes over that postoperative care, our guide to splitting a cataract global with modifiers 54 and 55 covers both claims.

Retina and subspecialty clinics

Retina clinics see the clearest case for E/M. A visit that includes imaging review, a treatment decision on an anti-VEGF agent and risk from the drug itself reaches a higher E/M level on decision making than the exam-based eye code will support. Where the visit is a scheduled injection with a brief check, the eye code often wins instead. The mix inside one clinic is wide enough that a single standing policy is always wrong for part of the schedule. The injection itself bills separately, and our guide to billing 67028 with the anti-VEGF drug code covers that claim.

Primary care and internal medicine

Primary care cannot bill 92002 through 92014 in practice, since these describe general ophthalmological services, and a primary care eye complaint is an E/M visit. The place this matters is the diabetic eye examination: the retinal examination performed by the eye care provider is what satisfies the quality measure, so the referral loop and the result coming back into the record is the thing to manage, not the code.

Pediatrics and vision screening

Pediatric eye care runs mostly through vision screening codes rather than either family, and the common error is billing a screening as an intermediate eye examination. Where a child is seen for strabismus, amblyopia or a failed screening, that is a medical visit and the normal rules apply. Keep instrument-based screening separate from the examination codes entirely.