How to get more clinician referrals for your practice

Most private practices grow on clinician referrals, and most have no method for getting them. The referrals arrive from whoever already knows you — a former colleague, a supervisor from training, the one physician who sent someone in 2021 and kept going. That is a real asset and also a fragile one, because you did not choose it and cannot replace it on demand.

This is how to do it deliberately. No part of it is secret, and you can do all of it yourself.

Who actually refers to a practice like mine?

Wider than most practices assume, and the list is public.

For a behavioral-health practice, the realistic referrers are primary care physicians and paediatricians, OB-GYNs and midwives, psychiatrists and psychiatric nurse practitioners who are full, other therapists whose caseloads are closed or whose specialty differs from yours, school counsellors and university health services, and — often the highest-value and most overlooked — neighbouring practices that share your population but not your niche.

The federal provider registry, NPPES, lists every clinician in the United States who bills insurance under an NPI: name, practice address, taxonomy code, and whether the record is active. It is free and downloadable in bulk. For any county you name, you can produce the complete list of licensed clinicians who could plausibly send you a patient. Most practices have never seen that list for their own territory, and it is usually several hundred people long where they assumed it was a dozen.

How many referral sources do I actually need?

Fewer than you would think, but more than you have.

A practice with three referral sources has a concentration problem: one retirement, one hire, one change in a health system’s internal policy, and a third of your inbound disappears with no warning. Practice-marketing convention puts a healthy list somewhere between twenty and ninety active relationships. The number matters less than the shape — you want enough that no single departure is a crisis, and you want to know which ones are actually producing.

Who should I approach first?

Not alphabetically, and not whoever you happen to remember.

Rank the list before you touch it. The signals that matter are specialty adjacency (does this clinician see people who need what you do), practice type (a solo clinician decides alone; a large group has a referral pathway and a gatekeeper), proximity, and whether the practice is demonstrably still active at the address on file. Registry data goes stale — clinicians retire, practices move, records linger — so verifying that a practice is still there before you spend postage on it is not optional.

Working the top of a ranked list beats working the whole list badly. Twelve well-chosen introductions a month is a program; two hundred generic letters is a mailing.

What am I allowed to say to them?

Less than a marketing agency will tell you, and more than your caution assumes.

The safe territory is education and access: who you are, what you treat, what your credentials are, how someone refers to you and how quickly you can see them. That is genuinely useful information to a physician deciding where to send a patient at 4pm on a Friday.

The dangerous territory has two parts. The first is your own claims. Any credential, specialty, outcome, or affiliation you state about yourself needs to be verifiable against something better than your website — a certificate, a board lookup, a licence record. The discipline that works is default-deny: nothing prints unless it has been checked, and a fact that is true but unverified does not print until it is. Practices get into trouble here through drift, not dishonesty. A phrase written once in 2019 and copied forward across every piece of collateral since is the classic pattern.

The second is anything that looks like paying for referrals. Federal anti-kickback rules are the headline, but for behavioral health the sharper constraint is usually your own professional code and your state’s fee-splitting statute. The ACA code prohibits fee splitting and referral remuneration outright. The APA requires that payments between professionals reflect services actually provided rather than the referral itself. Several states prohibit referral fees by statute. Practically: gifts of nominal value and genuine education are ordinary practice; revenue shares, payments that scale with how many patients someone sends, and free services that function as compensation are not. When it is close, it is a question for your attorney and not for your marketing vendor.

How do the small details go wrong?

Through automation nobody checked.

Here is a real example from building this. The rule that decides whether a clinician is addressed as “Dr.” looked simple: check whether their credential string contains “MD”. Run that against a corpus of 9,756 real clinician records from the public registry and it addresses 24 chaplains as doctors — because “MDiv” contains “MD” — while missing 522 people who had actually earned the title, because their credential was recorded in a form the rule did not anticipate.

Nobody would make that mistake writing one letter by hand. Everybody makes it at scale. The lesson generalizes: the failure mode of referral outreach is not saying something outrageous, it is a hundred small errors of address, credential, and specialty that each tell a clinician you did not actually look at them. The fix is that a system should abstain when it is not certain rather than guess.

What should I measure?

Referrals. Not opens, not replies, not “impressions”.

The unit that matters is an activated referrer: a clinician who has sent you at least one patient. Everything else is a leading indicator, useful for deciding who to follow up with and useless for deciding whether the program works. Read activation at around ninety days, because that is roughly how long the behavior takes to show up — a physician has to see the right patient before they can refer them.

Two practical requirements. You need to know which referrals came from the program rather than from your existing network, which means asking at intake and recording the answer. And you need the leading indicators anyway — who engaged, who called — because they tell you where to spend the next month’s attention.

Should I do this myself or hire it out?

Do it yourself if you have someone who owns it.

The work is not hard. Pull the registry for your counties, rank it, verify your own claims against source documents, have counsel look at the package once, send it, follow up, record what comes back, and repeat next month with a list that has partly rotted since. Practices rarely stop because any step is difficult. They stop at month four, because it is the thing that gets dropped when the schedule fills.

Hire a person — a liaison who walks into offices — if you can afford the salary and keep them busy. A human being builds relationships no system can, and the ceiling is higher. Hire infrastructure if the constraint is that nobody has the hours and the compliance question is what has been stopping you.

The honest version of all three: the method above is the same in each case. What changes is who carries it.