Quick answer: A nephrology practice automates prior authorization by connecting an AI agent to the EHR so it reads orders as they're placed, checks payer rules to decide whether authorization is required, and assembles the clinical packet — labs, diagnosis codes, prior therapy history — that each payer demands. The work splits into three buckets that automate differently: recurring dialysis authorizations need expiration tracking and pre-emptive renewal, ESA and IV iron requests need current lab values pulled automatically, and CKD drug requests need documented step-therapy history attached before submission.
Start by sorting your prior authorization volume into three buckets
Before you evaluate a single vendor, spend a week counting. Most nephrology groups have never broken their PA volume down by type, and the breakdown determines where automation pays first.
Sort every authorization your practice submitted last month into three groups:
- Recurring authorizations — dialysis, ongoing ESA therapy, maintenance immunosuppression. These repeat on a schedule and expire.
- Lab-gated drug authorizations — ESAs, IV iron, and anything where the payer's criteria reference a specific value with a recency requirement.
- Step-therapy drug authorizations — SGLT2 inhibitors, GLP-1s, newer agents on non-preferred tiers, where approval turns on what the patient tried first.
Most nephrology practices find the split lands somewhere near 40/35/25, though it moves with payer mix and how much dialysis you manage in-house. The number that matters more is which bucket generates your denials. That's usually not the biggest bucket.
For scale context, the 2025 AMA Prior Authorization Physician Survey found practices average 40 prior authorizations per physician per week and spend 13 hours of physician and staff time on them. In a six-nephrologist group, that's a full-time job plus change, usually carried by one or two people.
How do you automate recurring dialysis authorizations?
This bucket has the clearest automation story because the work is genuinely mechanical.
A recurring authorization has a start date, an end date, and a renewal requirement. Your staff currently tracks these in a spreadsheet, a shared calendar, or somebody's memory. When one slips, the claim denies weeks later and your billing team works a denial that never needed to exist.
Automating it means three things:
- Registering every active authorization as a tracked object with its expiration date, payer, patient, and modality.
- Triggering renewal on a lead time — typically 30 to 45 days before expiration, so a slow payer doesn't create a gap.
- Re-submitting with current documentation rather than copying last cycle's packet forward, since payers increasingly reject stale attachments.
The measurable outcome here isn't hours saved. It's the count of lapsed authorizations, and the target is zero. Track that number before you start; most practices discover it's higher than they assumed because lapses surface as denials months later and get miscategorized.
Honey Health's prior authorization agent handles this bucket as a standing queue rather than a request-by-request workflow, which is the structural difference between catching expirations and chasing denials.
ESAs and IV iron: making lab-gated criteria work on first submission
Anemia management authorizations fail for a boring reason: the labs attached don't meet the payer's recency or threshold requirements.
Payer criteria for erythropoiesis-stimulating agents and IV iron typically reference hemoglobin, ferritin, and transferrin saturation, with a window on how recent the draw must be. A packet built on Tuesday from labs drawn seven weeks ago gets denied — not because the therapy is wrong, but because the documentation is stale.
Automation fixes this by inverting the sequence. Instead of a coordinator gathering labs when they build the request, the agent watches for the order, checks whether qualifying labs exist inside the payer's window, and either attaches them or flags that a draw is needed before the request goes out.
What to require during evaluation:
- Can it read discrete lab values with draw dates, not just a scanned results document?
- Does it know payer-specific thresholds and recency windows, or does it apply one generic rule?
- Can it hold a request and alert staff when labs are out of window, rather than submitting a packet that will predictably fail?
That third capability is the one vendors skip in demos and the one that moves your first-pass approval rate.
CKD medications and the step-therapy documentation problem
SGLT2 inhibitors are where nephrology PA gets genuinely fiddly, and where automation earns trust or loses it.
Jardiance and Farxiga sit on non-preferred formulary tiers for many plans, with step therapy attached. The denial pattern is consistent: the payer wants documented evidence the patient tried and failed a preferred agent, and the chart contains that history in narrative form rather than as a structured record.
A well-configured agent builds the step-therapy record explicitly — naming the prior drug, the dose, the start and stop dates, and the documented reason for discontinuation — and attaches it as part of the packet rather than leaving the reviewer to infer it from a note.
The subtlety that separates a nephrology-aware tool from a generic one is indication logic. Step therapy through metformin generally applies to the type 2 diabetes indication. It typically does not apply when the drug is prescribed for chronic kidney disease or heart failure, which are separate approved indications with separate criteria. An agent that submits every SGLT2 request with a metformin trial attached is doing unnecessary work; one that submits the CKD indication without distinguishing it invites an avoidable denial.
Ask any vendor to walk through this specific case. The answer tells you whether they've configured for nephrology or are selling you a generic PA engine with a specialty label on the box.
A 90-day rollout that doesn't blow up your queue
The practices that get this wrong try to automate everything at once, lose visibility into what's working, and end up running two parallel processes for six months.
A sequence that holds up:
Days 1–30: baseline and narrow scope. Measure first-pass approval rate, median time from order to decision, and lapsed-authorization count. Pick one payer — your highest-volume Medicare Advantage plan — and one drug class, usually ESAs. Configure only that.
Days 31–60: run parallel and compare. The agent builds packets; a coordinator reviews before submission. You're checking whether the packets are actually better, not whether the software runs. Expect to correct payer rules during this window; that's the point of it.
Days 61–90: expand and release the review gate. Once first-pass approval on the pilot slice beats your baseline, drop the human pre-review for that slice and add the next payer or drug class. Keep the review gate on anything newly configured.
Two things to hold firm on. Don't let a vendor talk you into skipping the parallel window — it's where you find out whether their payer rules match reality. And don't expand faster than you can measure; if you can't attribute a change in approval rate to a specific configuration, you've lost the feedback loop.
Most implementations run 6 to 12 weeks end to end, and the constraint is almost always EHR integration and payer configuration rather than the software itself.
What your team still handles after automation
Automation moves the work; it doesn't eliminate the role. Set expectations with your PA coordinator early, because the ones who think they're being replaced tend to quietly withhold the payer knowledge that makes configuration work.
What stays human:
- Peer-to-peer reviews. A platform can schedule the call and brief the physician. It can't take the call.
- Appeals that need a new clinical argument. Automation proves documentation exists. It doesn't construct a novel medical-necessity case.
- Exceptions and edge cases. Off-label use, unusual dosing, patients whose history doesn't fit a decision tree.
- Payer relationships. Someone still needs to know which reviewer at which plan actually moves things.
The 2025 CAQH Index found electronic adoption for medical prior authorization at 40%, up from 31% two years earlier. That leaves a majority of medical PAs still moving through channels that need a human somewhere in the loop, and nephrology's payer mix skews toward the harder end of that distribution.
The honest framing for your board is capacity redeployed, not headcount removed. A coordinator who stops assembling packets and starts working denials and exceptions is worth more, and that's a defensible claim. "We'll cut a position" usually isn't.
Frequently Asked Questions
What should a nephrology practice automate first?
Recurring dialysis authorizations, in most cases. The work is mechanical, the failure mode is expensive, and success is easy to measure — lapsed authorizations should go to zero. ESAs are a reasonable alternative first target if your anemia-management denial rate is high.
How much does prior authorization automation cost?
Pricing usually runs per-authorization or per-provider per-month, and vendors rarely publish it. Build your comparison against current cost rather than against zero: the 2024 CAQH Index put the provider cost of a manual prior authorization near $10.97 versus roughly $5.79 for a fully electronic one, before counting denial rework.
Do we need to switch EHRs to automate prior authorization?
No, and be skeptical of anyone who suggests it. PA automation platforms sit alongside the EHR, reading orders and writing authorization numbers back. What matters is integration depth — whether the platform can read discrete lab values with draw dates, not whether you're on a particular system.
How does automation handle Medicare Advantage plans specifically?
The same way it handles any payer, but the payoff is larger because MA plans apply utilization management that traditional Medicare doesn't, and kidney-disease MA enrollment is high. Ask vendors for their configured payer list against your top ten by volume.
Will automation reduce our denial rate?
It reduces administrative denials — the ones caused by a missing lab, an undocumented step-therapy trial, or a lapsed authorization. It won't change the outcome of a genuine medical-necessity dispute. In most nephrology groups the administrative share is the larger one.
What if a payer doesn't support electronic submission?
Good platforms still handle it, submitting through the portal or by fax the way staff would. Ask specifically how the vendor handles your fax-only payers, because "we support electronic PA" sometimes means "we don't touch the payers that don't."
