HCP Research: How to Recruit and Interview Physicians, Nurses and Other Healthcare Professionals

Research with healthcare professionals is ordinary qualitative research wrapped in four constraints that do not apply anywhere else: access is brutal, payment is regulated, transparency reporting may be triggered, and anything a participant mentions about a product may create a pharmacovigilance obligation. Get those four right and the research itself is familiar. Get them wrong and you have a compliance incident rather than a study.

The headline access problem is real. Median physician response rates in research sit around 18%, with reported ranges of roughly 10–60% depending on specialty, incentive and follow-up. A specialist oncologist has perhaps twenty minutes of uncommitted time in a working day, and it is not at 2pm on a Tuesday when your moderator is free. This is the single strongest argument for asynchronous, AI-moderated interviewing in healthcare: a study a clinician can complete by voice at 22:40 after clinic, in six minutes of speaking, converts at rates a calendar invite never will.

Who counts as an HCP, and why the label matters

"HCP" covers physicians, nurses and nurse practitioners, pharmacists, physician assistants, dentists, allied health professionals and, in most compliance frameworks, anyone in a position to prescribe, dispense, purchase or recommend a medical product. Two adjacent groups are often confused with them and carry different rules:

• Payers and formulary decision-makers — commercially sensitive, rarely subject to transparency reporting, usually recruited through specialist networks.
• Practice and hospital administrators — a purchasing audience, not a clinical one, and frequently the actual buyer for medtech and health IT.

Scope the audience before you scope the method. A study that mixes prescribers and administrators without separating the samples produces findings that neither group would recognise.

The four constraints, in order of how often they bite

Recruiting: five channels, ranked by what they actually deliver

1. Your own customer or user list. Highest response, lowest cost, most bias. Perfect for product research; useless for market landscape work. Import the list and send personalised interview links so each participant arrives pre-identified and you never ask a question you already know the answer to.
2. Professional associations and society lists. Credible and specialty-accurate; slow, and often requires sponsorship framing that colours responses.
3. Specialist HCP panels. Fast and verified, and the default for pharma and medtech market research. Expensive, and heavy panel users are not representative of the specialty.
4. Referral and snowball recruiting. The only reliable route to genuinely rare specialties and to key opinion leaders. Slow to start, then compounds.
5. Congress and conference intercepts. Excellent for breadth in a short window, terrible for depth — nobody gives you thirty focused minutes in an exhibition hall. Capture short async studies there instead and let people complete them on the train home.

Whichever channel you use, verify. Ask for a licence or registration number where lawful, cross-check specialty against claimed practice setting, and build screener questions that a non-clinician cannot pass. A well-designed clinical screener asks about workflow specifics rather than credentials: which formulation you reach for first in a specific presenting scenario, what your unit's protocol says about a particular threshold, who signs off on a given order. Professional survey-takers fail those; real clinicians answer them without thinking.

Honoraria and fair market value

HCP payments are not incentives in the consumer sense. They are compensation for professional time, and they must be defensible.

The mechanics most agencies use:

• Hourly-rate model. A per-hour figure set by specialty, seniority and geography, then multiplied by the true participation time — including any pre-work. Life sciences clients typically maintain internal FMV bands and will not approve anything above them.
• Record-to-recruit ratio. Divide the size of the qualified universe by the required sample size. A large universe means a lower honorarium clears; a rare specialty with a universe of a few hundred needs a much higher one.
• Published market ranges. Short polls of around five minutes commonly sit in the $5–$15 range, while 60-minute in-depth interviews commonly run $200–$500 and above. Oncology, immunology and rare disease specialists command the highest effective rates, often $100–$300+ per hour.

Three rules keep this clean. Pay for time, not for outcome or opinion. Document the FMV basis before fielding, not after. And never let honoraria vary by what a participant says.

Shorter interviews save real money here. If an AI interviewer can extract the same depth in eighteen minutes that a human moderator needs forty-five minutes to reach — because it never spends time on rapport-building preamble, never re-asks something already answered, and probes only where the answer was thin — the honorarium falls proportionally and the response rate rises. That is the strongest commercial argument for AI moderation in HCP work.

Transparency reporting and the blinding exclusion

In the US, the Physician Payments Sunshine Act requires applicable manufacturers to report transfers of value to covered recipients through Open Payments. Market research is one of the few areas with a practical carve-out.

Payments made in genuinely double-blinded market research are not reportable, because the manufacturer does not know the identity of the participant and the participant does not know the sponsor. The exclusion also applies where only the manufacturer is blinded — single-blind research where the sponsor never learns who participated. If a study is only partially blinded, only the blinded portion falls outside reporting.

The operational consequence: decide the blinding design before recruitment, not after. If your sponsor needs to know exactly which KOLs participated — which is often legitimate for advisory work — accept that it is reportable and budget the compliance process. If the research genuinely does not need identities, keep the fieldwork with an agency or platform that holds them and deliver de-identified results. Retro-fitting blinding to a study that already exchanged names does not work.

This is one reason to keep participant identity separated from analysis output. Koji lets you run studies where the sponsor sees aggregated findings and anonymised quotes while identity stays with whoever recruited — and de-identification technique is covered in depth in the anonymisation guide linked below.

Adverse event reporting: the obligation nobody plans for

If your research touches a marketed medicinal product, a participant may mention a suspected adverse reaction. That mention can create a pharmacovigilance obligation for the marketing authorisation holder, and the obligation exists whether or not you were looking for it.

EphMRA maintains the standard framework here — a Code of Conduct alongside dedicated Adverse Event Reporting Guidelines, most recently revised in September 2025 — and it complements the ICC/ESOMAR code. Practically, a compliant HCP study needs five things in place before it fields:

1. Trained interviewers or a trained review process. Everyone touching the data must be able to recognise an adverse event mention.
2. A defined reporting route and timeline to the client's pharmacovigilance function. Reporting windows are tight and are set by the client's SOP — confirm the exact window contractually before fielding.
3. The four case elements. A reportable case classically requires an identifiable reporter, an identifiable patient, a suspect product and an event. Your process must be able to capture all four, or explain why it cannot.
4. Re-contact consent. Pharmacovigilance follow-up may need to go back to the reporter. If your consent form forbids re-contact, you have built a dead end.
5. A reminder script. The interviewer must remind a respondent who mentions an adverse event to report it through normal channels.

Asynchronous AI interviews do not remove this obligation — they change where it sits. Because every session is transcribed in full, adverse event review moves from "did the moderator notice?" to a systematic pass over complete transcripts, which is a genuinely stronger control than relying on live recall. Build the review step into your process and give it a named owner.

Designing an interview a clinician will actually finish

• Cap it at 20 minutes of participant time. Publish the cap and honour it. Clinicians abandon studies that overrun more readily than any other audience.
• Lead with the clinical scenario, not the product. "Walk me through the last patient where you considered switching therapy" outperforms any attitudinal opener.
• Use structured questions where you need to count. Koji supports six types — open_ended, scale, single_choice, multiple_choice, ranking and yes_no. Rank treatment attributes with ranking, capture confidence on a scale, take the free-text reasoning with open_ended, and let the AI probe every low score automatically.
• Offer voice. Speaking is faster than typing for a clinician between patients, and voice answers run considerably longer and richer than typed ones.
• Stay non-promotional. Research must not drift into product promotion, and it must never discuss unapproved uses. Have medical or legal review the discussion guide when the sponsor is a manufacturer.
• Localise properly. Multi-market HCP studies need genuine language coverage, not a translated survey — clinical terminology carries local convention.

Mistakes that cost studies

• Recruiting "physicians" without verifying specialty, then discovering half the sample is out of scope at analysis.
• Setting honoraria by negotiation rather than by documented FMV basis.
• Deciding blinding after names have already reached the sponsor.
• Fielding a study touching a marketed product without an adverse event process or trained reviewers.
• Booking 60-minute moderated calls when a 20-minute async study answers the same questions with three times the sample.
• Treating patients, caregivers and clinicians as one study. They need separate designs, separate consent and — where patients are involved — a much closer look at whether ethical review applies.

How Koji fits HCP research

Koji runs AI-moderated interviews by voice or text that clinicians complete on their own schedule, in their own language, with automatic follow-up probing that reaches the depth a static survey never gets to. Personalised interview links let you field to a verified HCP list without re-screening people you already know, CSV import handles the list itself, and every session produces a full transcript plus structured answers that aggregate into a report immediately — so a study that would have taken six weeks of scheduling closes in days.

Used properly, that is not just faster. Shorter sessions lower honoraria, higher completion improves sample quality, and full transcription makes adverse event review systematic rather than dependent on a moderator's memory.

Frequently asked questions

Do market research payments to physicians have to be reported under the Sunshine Act? Not when the research is genuinely double-blinded — the manufacturer does not know who participated and the participant does not know the sponsor. The exclusion also covers single-blind designs where only the manufacturer is blinded. If a study is partially blinded, only the blinded portion falls outside reporting. Decide the blinding design before recruiting, because it cannot be retro-fitted once identities have been shared.

What honorarium should we offer a physician for a 60-minute interview? Published market ranges commonly put 60-minute in-depth interviews at $200–$500 and above, with oncology, immunology and rare-disease specialists at the top. The right figure comes from your fair market value basis: specialty, seniority, geography and true participation time, sized against how rare the qualified universe is. Document that basis before fielding, and pay for time rather than for opinions.

Do we need ethics or IRB review for HCP market research? Market research that produces business decisions rather than generalisable knowledge usually falls outside the human-subjects research definition, but the answer depends on your jurisdiction, on whether patients or patient data are involved, and on whether you intend to publish. Studies touching patients, clinical outcomes or identifiable health information should be assessed properly rather than assumed exempt.

How do we handle an adverse event mentioned during an interview? Capture it, remind the respondent to report it through normal channels, and pass it to the client's pharmacovigilance function within the window their SOP specifies. A reportable case classically needs an identifiable reporter, an identifiable patient, a suspect product and an event. Make sure your consent permits the re-contact that follow-up may require, and give the review step a named owner rather than leaving it to whoever reads the transcripts.

How do you verify that a research participant is really the specialist they claim to be? Combine credential checks where lawful with clinical screener questions a non-clinician cannot answer — the protocol threshold their unit uses, who signs off a particular order, which formulation they reach for first in a specific presenting scenario. Credential checks catch impostors; workflow questions catch people who hold the credential but do not do the work in question.

Is asynchronous AI interviewing acceptable for HCP research? Yes, and it usually outperforms scheduled calls for this audience. Median physician response rates around 18% reflect calendar friction more than unwillingness. An async voice study removes the scheduling problem, shortens participation time, lowers the honorarium accordingly, and produces complete transcripts that make compliance review systematic.

Related resources

• Structured Questions Guide — the six question types and when to use each with time-poor experts
• Expert Interviews Guide — the general playbook for interviewing specialists
• AI Customer Research for Pharma & Life Sciences — the wider industry view
• Research Participant Incentives — incentive design outside regulated audiences
• Anonymizing Customer Interview Data — how to deliver findings without identities
• IRB Approval for User Research — when ethical review actually applies
• Multi-Language User Research — running multi-market clinical studies properly