Cigna's Curious Modifier 25 Policy

Cigna’s Policy

In late May, Cigna announced that it would require all claims with CPT codes between 99212 – 99215 and a modifier 25 to be accompanied by the submission of a medical record prior to payment.?This seemingly innocuous policy was a bold move that sent shock waves through the medical coding departments of providers, and Cigna quickly delayed its enactment.?Cigna has since announced that it will not implement the policy, following backlash from provider groups. Unpacking the significance of this simple policy is a fascinating look into the complexity of medical billing and the actions payers are looking to take to ensure accurate claims payment.

?The E&M line will be denied if we do not receive documentation that supports that a significant and separately identifiable service was performed.

The CPT codes included in the Cigna policy are used for evaluation and management office visits for established patients. These are the codes used for nearly every visit to a patient’s primary care physician - they are extremely common.?The 4 codes in the range represent different levels of intensity/complexity based on the nature and duration of the visit, with higher payment for the codes higher in the range.?

The Modifier 25 included in the Cigna policy indicates a “separate and distinct E/M service”.??Unbundling” is a core concept in medical coding and describes the situation where a code’s included services fall entirely within a broader more comprehensive code.?We say the first code is “bundled into” the comprehensive code.?When both codes are billed together (they are “unbundled”), the duplicated services are reimbursed twice.

For example, CPT 45378 is for a diagnostic colonoscopy and CPT 45382 is for a diagnostic colonoscopy with control of bleeding.?45382 includes all of the services of 45378, as well as an additional service: control of bleeding.?If 45382 is billed, 45378 cannot generally be billed because it would duplicate the diagnostic colonoscopy payment.?

More generally, all surgical codes encompass the services essential to the performance of the procedure, such as evaluation on the day of surgery, anesthesia, usual supplies and materials (e.g. sterile trays/drugs), immediate postoperative care, and typical follow-up care (extending up to 90 days after a procedure for major surgeries).?These essential services are included in the surgery codes, and so codes dedicated to these individual services are not billed separately.?E/M codes are the most common service unbundled from surgical codes.

Modifier 25’s purpose is to say “I know it looks as if this code could be unbundled from a broader code, but actually it is entirely unrelated to the broader code”.?For example, if a patient has hip surgery (which as a 90-day follow-up period), any normal follow-up care visits in the 90 days after surgery would not be billed as a separate E/M code.?However, if a provider saw a patient for an unrelated reason, a separate E/M code would be billed.?Diagnosis codes can provide some information to distinguish these scenarios, but modifier 25’s presence makes it clear.

Claims Auditing

In 1996 CMS implemented the “National Correct Coding Initiative” (NCCI) which included several edits designed to ensure appropriate coding.?One of these was the “Procedure To Procedure” (PTP) table, which lists all of the 2-code combinations where unbundling can occur – there are over 600,000 combinations on the table currently.?Commercial payers all have variants of this table.?This table was a significant technological advancement to claim auditing because by documenting all code combinations where unbundling could occur, payers could use computers to audit these claims.?The computer is good at applying the black and white rule of “code A does not go with code B”.

Except modifier 25 introduces a massive grey area.?The modifier essentially excepts a claim from the application of the PTP edit, aside from a very limited number of circumstances where the PTP table says the modifier is not allowed.?These are generally situations where the two codes would be clinically impossible, such as billing for a closed and open surgery.?But 97.3% of the code combinations on the PTP table are like the hip replacement example, where the clinical situation is not fully described by the two CPT codes alone, and an evaluation of the appropriateness of modifier 25 requires reviewing the medical record.

As a result, modifier 25 misuse is one of the most common forms of medical billing errors.?It is not uncommon for payers to see providers who attach a 25 modifier to every single E/M code they bill (whether it is needed or not, much less appropriate or not).?Modifier 25 is simply an easy way to reduce denials.?Further, these claims are relatively low-cost in general (the added E/M codes pay between $75 and $500) and so do not warrant significant attention from payer’s payment integrity activities due to the relative abundance of more expensive audit findings elsewhere and the cost/benefit of investing auditing resources in these claims.

Cigna's Curious Decision

So why was Cigna asking for medical records prior to payment of these claims if they are not reviewing these medical records today??One possible explanation is that Cigna has figured out how to have a computer review a medical record for modifier 25 appropriateness.?Natural language processing (NLP) is a promising future and many payers are working aggressively towards harnessing it’s power for payment integrity uses (among many other uses, including risk adjustment and clinical management).?At the same time, nowhere else in the industry have we seen NLP able to do most types of medical coding audits effectively.?As the hip replacement example shows, modifier 25 appropriateness is subtle, and tasks like determining the order of two events and their primary purpose is a struggle for current NLP technology.???

It is not impossible that Cigna has this technology, but if they did I still do not think that explains this policy.?The prevalence of modifier 25 abuse would mean the technology would deny a lot of claims.?This would lead to many provider appeals, and those appeals cannot be handled by a computer, posing a significant operational problem for Cigna (for relatively scant savings).?Further, providers may sue Cigna over the NLP model arguing it unfairly denied claims, and an adverse ruling could potentially hinder any NLP-powered auditing activities in the future.?Cigna would be taking a huge gamble in the court system with first-generation technology while angering its provider network in the process.

I propose that Cigna wanted these medical records without ever intending to issue denials.?The growth of NLP and other machine learning/AI technologies in the last decade was driven by two earlier developments: big data and stronger computer power.?Cigna has the computer power, but may lack the data.?Historically, medical records have been haphazardly managed by payers – they may be stored in PDFs rather than as machine learning-ready data files, different areas of the company (i.e. risk adjustment, payment integrity, clinical management, etc.) may store medical records in different repositories, and the medical records may not be categorized in a way that lets them easily be integrated with other data sources (such as the claim that was submitted as a result of the medical record).?Payers have made significant progress in addressing these problems in recent years, and Cigna may now be operationally ready to manage a large influx of medical records.?Building a large and machine learning-ready dataset of medical records would be a huge competitive advantage in the race toward strong NLP/AI tools to power significant advancements in risk adjustment, payment integrity, and clinical management.

The easiest way for Cigna to get that large influx of medical records may be through an expansive reimbursement policy.?Targeting individually selected claims, specific providers, or haphazardly requesting records would all lend themselves to operational and potential legal difficulties.?Attempting to integrate providers’ medical record systems into Cigna’s is a more viable path, though is likely not something most providers are eager to participate in.?Modifier 25 abuse is an ideal scenario – it leverages Cigna’s current audit activities, has the backing of CMS and universal commercial payer policy, and is extremely common.

I think Cigna was mostly interested in using the medical records the reimbursement policy would send them for purposes entirely unrelated to modifier 25 abuse, but used the cloak of modifier 25 abuse to justify the request.?Medical record in hand, Cigna had a power dataset to power its advancements in NLP and AI.?Providers would shoulder the burden of sending medical records to Cigna, but would not see an increase in denials.?If providers sued, Cigna could say the administrative burden on providers is a small step towards ensuring payment integrity for the system and that nobody is really harmed by the policy – perhaps not a convincing argument to the providers impacted, but a much better legal gamble than the increase in denials scenario discussed above.?

Conclusion

We may never know if my hypothesis is right.?Payers are extremely protective of their data assets because of the significant competitive advantages they can unlock.?Nevertheless, I anticipate that we will see slow but steady advancement in medical record NLP/AI technology over the next decade, and not until the end of that decade will we see computers that can audit a claim for modifier 25 abuse with reasonable accuracy.?Regardless of if Cigna’s modifier 25 policy is related to their work towards that goal or not, payers are working hard to use medical records as the catalyst to significant advancements in risk adjustment, payment integrity, clinical management, and more, and we are in for an exciting decade as we see the industry changed by these tools.