How do specialty practices measure referral leakage and cut it?

Referrals can come in all week while the front desk is answering calls, checking patients in, and working through everything else. Some patients book. Some never do. If you want to know whether referral follow-up is working, you need to keep both groups in the same report.

Health-system referral leakage concerns referrals going outside the network. Here, practice-side referral leakage means received referrals without a booking by the date your practice chooses to review them.

Hello Patient builds AI agents for healthcare. The agents call and text referred patients, answer practice questions from an approved knowledge base, and book their visits in your scheduling system. They can work through more of the referral list while your front desk handles the work in front of them. This page explains which referrals to count, which dates to compare, and what the Ortho1 results show.

Which incoming referrals belong in your count?

Count the referrals your practice or the locations you are reviewing actually received. A referral sent somewhere else does not belong in that list, even if your group could have treated the patient.

For a multi-location ENT group, decide first whether you are measuring the whole group or each location. If one office receives a referral and another office books the patient, the group report counts that as booked. A location report needs to show that transfer so you can match the location totals to the group total.

Keep the original received list as the outcomes change. Do not remove a referral just because it was hard to book. If you know why it stayed unbooked, keep that reason with the record. If you remove unresolved referrals, the report can make outreach look better than it was.

For this report, one referral episode is one request for a visit. Count a repeat fax for the same request once. Count a separate requested visit for the same patient as another episode. Use that episode count wherever the report or formula says “referrals.”

Which dates should you use to compare received, reached, and booked referrals?

First, pick the period when the referrals arrived. Then give every referral the same number of days to book. Count received, reached, and booked separately. A reached patient may not have replied. A booking means the patient chose a time, not that they came in.

For example, your practice might review referrals received during September and allow 30 days from each referral's receipt date. That is a measurement choice, not a recommended target. You would need to wait until the last referral had its full 30 days before treating the September result as complete. Comparing it earlier with an August list whose referrals have all had their full 30 days would give September less time to produce bookings.

Your team also needs one definition of “reached.” A delivered text, an answered call, and a patient reply are not the same thing. Put the definition beside the count. Keep referrals that are still unresolved in the report on the date you review the list, and add the reason when you know it.

Keep one row for each episode using the fields below. Your practice decides when a referral is ready for scheduling and handles any clinical questions.

Booked episodes divided by received episodes shows the share of your received list that booked. Booked episodes among those reached divided by reached episodes shows the booking share within that smaller group. Use the same list and review window for both. A count of patients reached cannot stand in for episodes reached without checking which episodes those patients belong to. For every rate in this article, report “not calculable” when the denominator is zero, rather than 0%.

Booking is also separate from closing the referral loop in CMS50v14 for the 2026 performance period. That measure concerns the referring clinician receiving the specialist's report.

What can you conclude from Ortho1's referral figures?

You can conclude that the outreach booked referral visits.

In Ortho1's public study, the orthopedic group's outbound SMS program reached 718 patients and booked 190 referral visits from May 1 through June 30, 2026, with no hires added. The booking share among patients who replied grew from more than a third in May to more than half in June.

Hear how it sounds here or request a free demo agent for your practice today by booking a call.

Why do some referrals remain unbooked?

Hello Patient's agents answer practice questions from your approved information, including which plans you accept, office hours, and locations. Clinical questions stay with your clinical team.

A referral may have been meant for another practice. The patient may have a plan your practice does not accept. The patient may not be able to afford the visit.

Keep the reasons you know in the report, along with any referrals you left out of the count. Keep unknown outcomes marked as unknown. An unbooked referral does not tell you whether the patient went somewhere else, decided not to get care, or simply has not booked yet.

What should you check in the first month?

Compare the time from receipt to readiness for scheduling with the time from readiness to the first contact attempt. Your practice chooses the review timing; this is not a clinical timing target.

For reached episodes without a booking, check appointment availability, accepted plans, and the patient's stated choice. Record what you know and give each unresolved episode a next action and an owner. Compare booking results with an earlier received list only after every episode has had the same time to book.

What should you ask a vendor claiming to reduce referral leakage?

Ask the vendor to show you the full received-referral list behind the result. You need to know when those referrals arrived, how long each one had to book, what counted as booked, what the baseline was, and what the vendor left out. Ask for the booking records that confirm each booking. If the vendor answers with how many patients it contacted, go back to the received list.

Then ask the vendor to demonstrate the referral booking workflow your practice will use. Your practice supplies the referral list and decides which visits Hello Patient's agents may book, with which providers and at which locations. Ask to see the agent work through a test referral by call or text, answer an approved practice question, and book a visit using your practice's configured scheduling rules.

Hello Patient works with many commonly used EHRs and practice management systems. Review HIPAA and security questions before you decide the scope.

Find out whether Hello Patient fits your practice's referral follow-up. Book a call and we'll show you how the agent calls or texts a referred patient, answers practice questions, and books a visit. Ask for a free demo agent to hear how it would answer your patients.

Frequently asked questions

Who at a specialty practice should receive referral conversations from an AI agent?

Before outreach starts, name a staff owner for each kind of conversation. Hello Patient's agent answers practice questions from the information your practice gives it, including which plans you accept, office hours, and locations. Text conversations can reach your team as tickets or messages, and calls can transfer live on the route and hours your practice chooses, so your team starts from what the agent collected. Clinical questions go to your clinical team. Set those destinations in advance so the agent knows where each conversation goes and your staff knows who owns the next step.

How does a specialty practice calculate its referral leakage rate?

For an operational unbooked rate, use: (received referrals minus booked referrals) divided by received referrals, multiplied by 100. When you measure a Hello Patient referral outreach pilot, use your own referral and booking records for both counts. Give every referral the same number of days to book and state the date you review the list. Keep exclusions visible and use the same rules in each period. The rate shows what share of the referrals in your defined list had not booked by that date. It does not tell you that every unbooked patient was lost or that more outreach could have booked them.

How much referral leakage is normal for a specialty practice?

The sources used here do not establish a comparable practice-side benchmark. For a Hello Patient pilot, your own referral records give you the most useful baseline because you can measure them with the same rules you will use later. Those rules cover which referrals you count, what counts as booked, which referrals you leave out, and how many days each referral has to book. A health-system network-leakage percentage measures a different problem and does not give your practice a default target. For an ENT group, a clean comparison against its own earlier referral list is more useful than an outside percentage when you cannot check what it counted.

How should a specialty practice reconcile two referral reports whose totals disagree?

Check the individual records before choosing which total is right. If the reports cover a Hello Patient pilot, compare both lists with the practice's referral and booking records. Mark referrals that appear in only one report, appear twice, or show different booking statuses. Then ask the report owners why each marked record was counted differently. Resolve each difference against the original record and note the correction. After that, recalculate both reports. The corrected entries need to explain the full gap between the two totals, not just enough of the difference to make the numbers look close.

What can referral tracking software tell a specialty practice about leakage?

Referral tracking software can show where a referral stands. For a Hello Patient outreach pilot, follow one received referral through the practice's contact and booking history. Check whether the software separates referrals waiting for contact, waiting for a reply, and booked. Look at unresolved referrals as well as completed ones. Then confirm that each reported total can be traced back to the records behind it. Better tracking can show you where follow-up is getting stuck. It does not contact the patient or book the visit by itself.

Can a practice attribute lower referral leakage to an AI agent if other things changed during the pilot?

Yes, when you separate the agent's effect from the other changes. Appointment availability, referral mix, staffing, or internal processes may also change during a Hello Patient pilot. More open appointments can make booking easier. A smaller share of referrals from plans the practice does not accept can also improve the result. Record those changes and look at where the improvement happened. If the locations that added appointments produced most of the gain, the extra appointments may explain part of it. The part of the gain those changes do not explain is your best estimate of the agent's effect.