Durable medical equipment prior authorization is structurally different from infusion center prior authorization in ways that are not obvious from the outside. The surface-level process looks the same: a referral arrives, fields get extracted, an authorization form gets filed, the payer responds. But the specific challenges that slow down DME authorization are distinct enough that tools and workflows designed for infusion centers often do not translate cleanly to DME operations.
We built Coral with infusion centers as the primary use case. As we spoke with DME providers who were dealing with related but distinct problems, we had to be honest with ourselves about what transferred and what did not. This article describes the DME prior authorization problem as we understand it, including where our approach helps and where the differences matter.
Higher Fax Volume, More Document Types
A small infusion center might process 15 to 25 new referrals per week. A comparable-size DME provider might process 50 to 100. The higher volume comes partly from the broader range of products DME providers supply, partly from the fact that many DME items have shorter authorization windows than infusion therapies and therefore require more frequent re-authorization, and partly from the nature of the referral pathway: DME orders come from a wider variety of physician specialties and therefore from a wider variety of document styles.
The document type variety compounds the volume problem. DME referrals arrive as written orders on letterhead, on standardized CMS form 484 (for home health oxygen), on custom physician order pads, through electronic prescription routing, and as pages pulled from discharge planners' binders. The format variation across all of these is substantially higher than the variation seen in infusion referral documents, which typically follow a more limited set of referring practice templates.
For a document extraction tool, higher volume combined with higher format variation is a harder problem than higher volume alone. It requires the extraction logic to handle more layout permutations without sacrificing accuracy on any individual document type.
The Coverage Determination Layer
DME prior authorization has an additional complexity layer that infusion authorization typically does not: Local Coverage Determinations (LCDs) and National Coverage Determinations (NCDs). CMS publishes these policies to define the clinical criteria under which specific equipment categories are covered under Medicare. A power wheelchair authorization is governed by a different LCD than a CPAP authorization, which is governed by a different LCD than a hospital bed authorization.
Prior authorization requests for these items need to demonstrate not just that the item was ordered, but that the patient's clinical condition meets the coverage criteria defined in the applicable LCD. This means the referral documentation needs to include the right clinical information, and the authorization form needs to be structured to address the coverage criteria specifically, not just to identify the requested item.
For an intake coordinator at a DME provider, this creates a dual challenge. First, they need to know which LCD applies to the requested item. Second, they need to assess whether the documentation in the referral actually satisfies the coverage criteria for that LCD. If the documentation is incomplete, they need to request additional information from the referring physician before they can file the authorization, not just after the payer denies it.
This coverage determination step is a judgment task that document extraction cannot fully automate. What extraction can do is identify which fields are present and which are absent relative to the LCD requirements, so the coordinator can make an informed decision about whether to file, request supplemental documentation, or contact the referring physician proactively.
Denial Patterns That Differ by Product Category
In infusion center prior authorization, denial patterns cluster around a few common reasons: insufficient documentation of medical necessity, step therapy requirements not met, incorrect diagnosis code for the requested therapy. These patterns are consistent enough that experienced coordinators can anticipate likely deficiencies before submission.
DME denial patterns are more fragmented by product category. Power mobility denials often involve functional assessment documentation requirements. CPAP denials often involve sleep study interpretation format requirements. Wound care supply denials often involve wound measurement frequency and documentation standard requirements. The pattern knowledge that helps a coordinator avoid denials for one product category does not necessarily transfer to another category even within the same DME operation.
This fragmentation makes it harder to develop and apply standard prior authorization expertise across a DME provider's full product range. It also makes automation that focuses only on field extraction less directly useful than in the infusion context, because the value in DME authorization is often upstream of field extraction, in the documentation completeness assessment.
Where Document Extraction Helps in DME
Despite the additional complexity, the core intake bottleneck in DME authorization is still the same as in infusion: coordinators spend too much time manually reading referral documents and transcribing information that is already present in the document into authorization forms. That problem is real, it is measurable, and document extraction addresses it.
For DME, the fields that matter most and that extraction handles well are the structured data fields: patient demographics, payer information, ordering physician NPI, HCPCS code for the requested item, and diagnosis codes. These fields appear in DME referral documents and can be extracted with comparable reliability to infusion referrals, assuming document quality is sufficient.
What extraction does not address directly is the coverage documentation assessment: whether the clinical notes in the referral satisfy the LCD criteria for the requested item. That assessment requires knowing which LCD applies and what it requires, then evaluating the documentation against those criteria. That is a rules engine problem layered on top of the extraction problem.
The Honest Scope Boundary
We want to be direct about where Coral's current capability fits the DME problem and where it does not.
Coral extracts the structured clinical fields from referral documents and presents them in a pre-filled authorization form format. This reduces the transcription burden for DME coordinators handling straightforward referrals with complete documentation. The coverage documentation assessment for LCD-governed items remains a coordinator judgment task. We flag fields as present or absent but do not currently evaluate coverage criteria compliance.
For DME providers whose intake bottleneck is primarily transcription burden on high volume, Coral addresses a real portion of the problem. For providers whose primary challenge is coverage determination complexity, that is a different tool requirement that we are working toward but have not yet built.
The DME prior authorization problem is harder than infusion authorization in specific ways. Solving the transcription piece helps. Pretending that solving the transcription piece solves the whole problem would not be accurate, and accuracy about scope is more useful than optimistic coverage claims.
What We Are Learning From DME Conversations
The conversations we have had with DME intake teams have been useful in shaping how we think about document extraction beyond infusion. The coverage determination problem points toward a documentation completeness check as the next useful feature layer: before a coordinator reviews extracted fields, showing them which LCD-required documentation fields are present versus absent in the referral would reduce the back-and-forth with referring physicians that currently adds days to DME authorization timelines.
That feature requires building and maintaining a structured representation of LCD documentation requirements by HCPCS code category, then matching the clinical content of the referral document against those requirements. It is a more complex problem than field extraction, but it is well-defined and approachable.
DME prior authorization is a different kind of hard from infusion authorization. Treating it as the same problem with higher volume would produce a tool that helps with part of the problem and leaves the harder part unchanged. The right approach is to be precise about where extraction helps and honest about where additional capability is needed.