ModMed's AI fax sorter and Blue on the same 100 referral faxes

ModMed ships an AI fax sorter. We wanted to know how it compares with Blue, so we took 100 inbound faxes from a 38-provider ENT group on ModMed and ran them through both.

How we scored it

We pulled 100 consecutive faxes from the practice's ModMed inbox. A person read every one and decided whether it was a true referral: a new patient being sent to the practice for care. Then we compared that answer with how ModMed's sorter and Blue classified each fax. 45 of the 100 faxes were true referrals.

The results

• ModMed's sorter marked 58. It found every true referral, plus 13 faxes that weren't referrals. That's 29% more referrals than really arrived.
• Blue marked 53. It also found every true referral, plus 8 that weren't, or 18% over the true count.

Neither AI missed a referral. The difference is in false flags: against ModMed's sorter, Blue cut them by 38%.

The two AIs got different faxes wrong

ModMed's false flags were mostly paperwork with nothing referral-like in it: lab results, a sleep study report, requests for medical records, a letter from a palliative care doctor, a surgery scheduling form, and a request to re-send a fax that came through cut off.

Blue's false flags were closer calls. They were documents that read like referrals but weren't new ones: pre-op clearance notes from a primary care doctor, a referral for physical therapy, appointment confirmations, a prescription request, and a thank-you-for-the-referral note sent to the wrong office.

That difference matters for the front desk. Every false flag is a fax someone has to open and set aside, and if outreach runs automatically, it's a patient who gets a message they didn't need.

What happened after tuning

We used the misses to tighten Blue's instructions, then ran the same 100 faxes again. On the recheck, Blue sorted all but one of its false flags correctly. The one it still gets wrong is a fax asking whether a previously referred patient was ever scheduled. It mentions a referral, but it isn't a new one.

Because the recheck used the same faxes we tuned on, it shows the fix works on this sample. The next batch of new faxes is the real test, and we'll publish that number when we have it.

More than a yes or no

Sorting is the first step. For each referral, Blue also pulled the structured data the front desk keys into the EHR by hand: the reason for referral (often with the ICD-10 code), the referring provider, and the insurance plan when the fax listed one. Blue can then text the patient a link to schedule.

What this means if you're on ModMed

You don't have to turn off ModMed's sorter to use Blue. ModMed delivers faxes as PDFs, which Blue reads directly; staff drag a batch into Blue once or twice a day and run it.

If you want to test this on your own faxes, run a sample through Blue and compare it with what your sorter and your staff did. Start free at useblue.ai/register.