How AI handles the diabetes technology authorization document loop end to end.

How does fax triage software handle CGM and insulin pump prior authorization paperwork?

TL;DR — Fax triage software handles CGM and insulin pump prior authorization paperwork by recognizing inbound DME supplier forms, payer determination letters, and documentation requests the moment they arrive, matching each to the open authorization and the right patient, extracting the decision or the missing-information request, and routing it to the work queue that owns the next step. The alternative is what most practices do now: a shared fax folder where a determination letter can sit for three days while a patient waits on a pump.

The paperwork loop nobody designed

Getting a patient onto a CGM or an insulin pump generates a document cycle that runs through your fax line four or five separate times. Nobody designed it this way. It accumulated.

The typical sequence:

  1. The supplier requests documentation. A DME supplier faxes asking for chart notes, the most recent A1c, and documentation of insulin regimen or testing frequency.
  2. Your practice sends clinical documentation. Someone pulls the notes, confirms they meet the payer's coverage criteria, and faxes them back.
  3. The payer responds — eventually. An approval, a denial, or a request for more information arrives by fax. Which of the three it is may not be obvious from page one.
  4. If denied or incomplete, the loop repeats. Additional documentation, an appeal, sometimes a peer-to-peer review.
  5. The approval arrives and has to reach the supplier and the chart. If it's filed but nobody tells the supplier, the patient still doesn't have the device.

Every arrow in that sequence is a fax landing in a shared inbox. The clinical work is not the bottleneck. The document handling is.

Then the whole thing runs again. CGM supplies reorder on a recurring cycle, pump supplies have their own schedule, and many payers require reauthorization at intervals or when a plan year turns over. A practice managing a few hundred diabetes technology patients isn't working through a queue that empties — it's maintaining a standing population of open authorizations, each generating documents at unpredictable intervals. That's the part that makes manual tracking break down. Any individual authorization is manageable. Two hundred of them, at different stages, with documents arriving in no particular order, is not.

Why delay here is a clinical problem, not a filing problem

It's tempting to treat this as back-office housekeeping. The data argues otherwise.

Research on prior authorization and diabetes technology access found that under PA requirements, 46% of CGM prescriptions were delayed by an average of 82 days and 21% were denied. When PA requirements were temporarily suspended in that study population, CGM use rose from 36.2% to 81.0%.

Those numbers describe payer behavior, not your fax inbox. But your fax inbox adds to them. If a determination letter arrives Monday and gets indexed Thursday, you've contributed three days to an 82-day average — and if it was a request for additional information rather than a denial, you've burned three days of a response window that may have a deadline attached.

The physician-side data points the same direction. The AMA's 2025 prior authorization survey found 95% of physicians report PA delays access to necessary care, 79% report patients abandoning treatment because of authorization problems, and 26% report a PA-related serious adverse event. Practices complete 39 PA requests per physician per week on average.

Document turnaround is the part of that chain you control.

What the automation actually does at each step

Fax triage software works this cycle by recognizing where in the sequence each inbound document sits.

On the supplier documentation request. The system identifies it as a DME request, matches it to the patient, and — this is the part that matters — recognizes whether an authorization is already open for that patient and device. If one is, the request attaches to it. If not, it opens a new work item rather than becoming an orphan document in a folder.

On the payer determination. Classification here has to be finer than "payer letter." An approval, a denial, and a request for additional information all require different next actions and different urgency. Good systems extract the decision, the authorization number, and — on denials — the stated reason, then route accordingly. An approval goes one way; a denial with a 30-day appeal window goes somewhere with a clock on it.

On the documentation request. These are the ones that quietly kill authorizations. A payer asks for one more piece of information, the letter is filed to the chart as correspondence, and nobody acts. Weeks later the authorization expires and the process restarts. Extraction that identifies "this is a request, here's what's being asked, here's the deadline" turns a filed document into a task.

On duplicates. DME suppliers frequently re-fax the same request two or three times. Deduplication prevents your staff from working the same item repeatedly and prevents your queue from looking worse than it is.

Honey Health's Fax Triage and Prior Authorization agents are designed to hand off to each other at exactly this seam: inbound document recognized and matched on one side, open authorization updated and next action queued on the other.

What automation catches that people reliably miss

Not because your staff are careless. Because these are the failure modes of volume, and no amount of diligence scales past a certain document count.

  • Determination letters filed to the wrong chart. Two patients with similar names, a rushed afternoon. The authorization looks unanswered and someone follows up on an approval that already arrived.
  • Requests for information read as denials. Page one says the request can't be approved as submitted. Read quickly, that's a denial, and the appeal path gets started instead of the much simpler send-the-missing-document path.
  • Expiring documentation windows. Chart notes and A1c results have recency requirements under many payer policies. A packet assembled from notes that have aged past the window gets denied for a reason that had nothing to do with medical necessity.
  • The supplier who never got the approval. The authorization is approved and filed. The supplier is still waiting. The patient calls in week three to ask where their pump is.
  • Pattern blindness. Fifteen denials in a quarter from the same payer for the same missing element is a fixable process problem. Seen one at a time, it's fifteen individual annoyances.

Where humans still own the work

Automation handles the document flow. It does not handle the parts that require clinical judgment or a conversation.

Peer-to-peer reviews are a physician on a phone call. No part of that automates, and any vendor implying otherwise is overselling.

The clinical narrative on an appeal — the paragraph explaining why this patient needs this device despite the payer's stated criteria — requires someone who understands the case. AI can assemble the supporting documentation and surface what the denial cited. The argument is human work.

Judgment calls on borderline coverage criteria. Whether a patient's documented testing frequency meets a payer's threshold is sometimes a genuine interpretation question, and getting it wrong wastes a submission cycle.

Escalation when a patient is stuck. Knowing when to call the payer directly, involve the supplier's rep, or go around the process is experience, not workflow.

The honest framing: automation removes the document handling so your coordinator's time goes to the appeal narrative and the escalation instead of to reading faxes and searching for charts.

The measurement that makes the case

If you want to justify this internally, track two numbers before and after.

Document turnaround — hours from fax arrival to the document being matched, filed, and actioned. This is the number automation moves most dramatically, often from days to under an hour for high-confidence documents.

Authorization cycle time — days from initial supplier request to approved-and-communicated. This is the number your physicians and patients feel. It won't drop as sharply, because payer response time dominates it and you don't control that. But the portion you own becomes visible, and visible is what lets you defend the spend.

A third number worth watching: denials attributable to process rather than medical necessity — missed deadlines, stale documentation, incomplete submissions. Practices are often surprised how much of their denial volume sits in that bucket, and it's the category most directly addressable by handling documents faster.

Collect the before numbers for at least two weeks, and collect them honestly. The temptation is to measure a clean week. Measure a bad one too — the week someone was out, or the January stretch when plan-year resets pile up — because that's the week the current process actually fails, and it's the strongest part of your case.

Frequently Asked Questions

Can fax triage software submit the prior authorization itself?

Not on its own. Triage handles the inbound side — recognizing, matching, extracting, routing. Building and submitting the authorization package is a separate capability, often a separate agent or module. Many practices deploy triage first because it's lower-risk and produces measurable results quickly, then add submission automation once the document flow is stable.

How does it know which open authorization a document belongs to?

Through a combination of extracted identifiers: patient demographics, the authorization or reference number if the payer included one, the device or HCPCS code, and the requesting supplier. Match confidence varies with how much of that the document actually contains. Payer letters that include a reference number match reliably; supplier faxes with only a patient name are harder and more often land in review.

What happens to documents the system can't confidently match?

They go to an exception queue with the AI's best guess pre-populated, so a coordinator confirms rather than starting from scratch. This queue needs a named owner. An unstaffed exception queue is how practices end up believing the automation failed when it was working as designed.

Does this help with denials and appeals?

It helps with the timing and the visibility, not the argument. Extracting denial reasons consistently means you can see patterns across payers — which is how you fix the upstream submission problem rather than appealing the same denial repeatedly. The appeal letter itself still needs a person who knows the case.

Is patient data safe with a third-party fax vendor?

Any vendor handling protected health information must sign a BAA and maintain safeguards under HIPAA's Security Rule. Ask for the BAA, a current SOC 2 Type II report, and specifics on where documents are stored and for how long. Vague answers to those questions are themselves an answer.

Will this work for other DME beyond CGM and pumps?

Yes — the document pattern is the same for nebulizers, CPAP, orthotics, and most other durable medical equipment: supplier request, clinical documentation, payer determination, possible appeal. Endocrinology practices tend to feel it most acutely because diabetes technology volume is high and renewal cycles are frequent, but the workflow generalizes.

More of our Article
CLINIC TYPE
LOCATION
INTEGRATIONS
More of our Article and Stories