How to automate fax and referral intake in a medical practice

Where fax time hides

Referrals, outside records, prior authorization decisions, results, refill requests and payer letters still reach most practices by fax or e-fax. MGMA notes that in 2019, 89% of healthcare leaders said their organization used a fax machine. Every page that arrives needs a person to open it, work out what it is, match it to the right patient, file it under the right document type in the EHR and send it to the right work queue.

MGMA's list of the hidden steps reads like a referral coordinator's morning: triage, patient matching, routing, rework when pages are missing, splitting packets that cover several patients, and outbound requests asking another office to fax over its last note. In a sponsored March 2026 MGMA poll, respondents said fax burden clusters in prior authorizations, denials and appeals, and referrals, with records requests and signatures also named.

Fax work spills into the rest of the front office. In a sponsored July 2026 MGMA poll, 75% of 216 respondents said clinicians use workarounds for EHR or practice management problems frequently or sometimes, and referrals and prior authorization were among the workflows named. Records requests carry their own clock. HIPAA generally requires action on a patient's access request within 30 days, and one 30-day extension needs a written reason and a completion date given to the patient in time. Texas generally requires an electronic copy within 15 business days of a written request when the EHR can produce it. A records request sitting in the fax queue is a compliance problem as well as a service one, so have your privacy officer confirm which deadline applies. This is general information, not legal advice.

An inbox full of faxes is still a fax queue

In MGMA's sponsored March 2026 poll, with 292 responses, 73% said they have a digital fax solution fully integrated with their EHR or practice management system and 24% said they do not. MGMA draws a useful line between digital fax that swaps the machine for an inbox and fax that feeds routing and indexing. With the first kind, pages arrive as PDFs and a person still opens each one.

Automation starts after the page arrives. An agent reads each fax and labels the document type, splits a packet that covers more than one patient, pulls the identifiers, proposes a patient match with a confidence score and suggests where the page should be filed and which queue should get it. MGMA itself points to AI-assisted classification, extraction and routing as the next step for practices.

The outbound side counts too. When a referral arrives without the last office note, someone calls or faxes the referring office, waits, and then matches the reply to the original referral. MGMA lists those outbound requests among the hidden steps, and each reply restarts the matching work when it lands.

What to automate and what to keep with your staff

The split follows the risk. Filing a page under the wrong patient is a privacy incident, so a person checks every match the agent is unsure of. Clinical urgency and the decision to accept a referral stay with your clinical team, and the agent never reads a result to judge how serious it is.

Start with the document types your staff already file under in the EHR. The agent maps each fax to one of them, and anything that fits none goes to a person. Outside records for existing patients, payer letters and pharmacy refill requests are good first types because they arrive often and need little judgment to route. New referrals come next, once your specialty checklist is written down.

Patient matching rules and the confidence threshold

Matching is where a fax workflow earns trust or loses it. A good rule needs at least two identifiers to agree, such as name and date of birth, and treats an MRN or a payer member ID as strong confirmation when the sender includes one. A common surname with no date of birth goes to a person. So does a name that matches two patients, a date of birth that conflicts with the chart, or any page where the agent read the identifiers with low confidence, such as a poor scan.

Set the threshold high at first and lower it only with evidence. For the first weeks, have staff confirm every match the agent proposes, so you learn its error rate on your own faxes. Then let high-confidence matches go through with a one-click confirmation and keep the rest manual. Audit a sample of filed documents every week, and treat any wrong-chart filing as the privacy incident it is.

Referral completeness is the second rule set. Write down, for each specialty, what a complete referral includes. For a GI practice that might be the order, the reason for referral, recent office notes, demographics and a copy of the insurance card. The agent checks each referral against that list and drafts one specific request for whatever is missing, so the referring office gets a single clear ask.

The controls to insist on

Before go-live, ask for each of these controls in writing:

• Every filed page records the identifiers the agent matched on and its confidence score.
• Matches and packet splits below the threshold wait for a person before anything is filed.
• Faxes marked urgent or stat go to a person at once and never wait in a batch.
• The agent reads only what it needs to process each document.
• A log shows who confirmed each match and where each page was filed.
• Retention follows your existing rules for faxes and the documents filed from them.
• Any service that stores your faxes, including the e-fax provider, has the HIPAA agreements your privacy officer requires before patient data flows through it.

Worked example: what the hours look like

This is an illustrative example with assumed numbers, not a client result. Take a 12-provider GI practice that receives 180 inbound faxes a day, or 900 a week, and 30 of each day's faxes are new referrals. Assume each fax takes 4 minutes to identify, match, file and route. That is 3,600 minutes, or 60 staff hours a week.

Now assume 80% of faxes, 720 a week, are high-confidence matches that need a one-minute confirmation, which is 720 minutes. The other 180 stay manual at 4 minutes, another 720 minutes. Fax handling falls to 1,440 minutes, or 24 hours, which gives back 36 staff hours a week. The calculator below uses these fax numbers.

Referrals add a second layer. Assume each of the 150 weekly referrals takes 10 more minutes today to check completeness and ask for missing items, which is 25 hours. If the agent runs the checklist and drafts the request, and staff spend 3 minutes reviewing it, that falls to 7.5 hours. Across both, the practice goes from 85 hours to 31.5 hours, about 53.5 staff hours a week given back. Your mix of fax types decides how much of that is real, so count a week first.

Start with a one-week fax audit

MGMA recommends starting with a one-week audit, and it is the right first step. For five days, have staff tag every inbound fax by type and note how many touches it took and the minutes spent. Count the multi-patient packets, the pages filed and then moved, and the referrals that needed a follow-up request. By Friday you will know your own routine share.

Measure the same things after go-live: faxes per day by type, minutes and touches per fax, days from referral received to visit scheduled, misfiled documents found in the weekly audit and requests sent for missing items.

Our five-day AI Workflow Sprint does the next part for $2,500. It runs on sample faxes in which your team has replaced patient details with made-up values, with nothing filed into a chart and no system access, and ends with the staff hours each workflow costs you and a fixed quote for the build. Full refund if we do not find 10 staff hours a week. Whoever builds it, insist that any later automatic filing has an agreed scope and documented validation.

Common questions