How Carevoices runs AI-moderated HCP research

How are Carevoices interviews designed?

Interviews follow the participant’s own answers with laddering: from what happened, to why it mattered clinically, to how it shaped a treatment or purchase decision. The approach is informed by Reynolds and Gutman’s laddering research (1988). Every interview is built from your brief, so the guide reflects the decision you need to make.

Source: Reynolds & Gutman, Journal of Advertising Research (1988).

• Natural conversation with 5–7 levels of laddering on clinical reasoning. Source: Laddering method after Reynolds & Gutman (1988).

How does sponsor blinding work?

Studies are sponsor-blinded by default. Invitations, screeners, consent and the interview itself never name the commissioning company, and stimuli are reviewed so they do not reveal it. Unblinded work, such as an advisory board the sponsor runs directly, is agreed in the study plan.

• Sponsor-blinded by default: participants never learn who commissioned the study. Source: Carevoices study protocol.

How are adverse events handled?

Every transcript is screened for adverse events, product complaints and other safety information, in every interview language. Reports go to your drug-safety or complaint-handling team within 24 hours of the interview, and a full reconciliation is delivered at study end so your safety records and ours agree. Whether a report meets a regulatory reporting obligation remains the sponsor’s determination.

Consent includes the adverse-event disclosure required by the EphMRA Code of Conduct and the BHBIA Legal and Ethical Guidelines.

• Every transcript is screened for adverse events, with reports to your drug-safety team within 24 hours and a full reconciliation at study end, in every interview language. Source: Carevoices adverse-event procedure; screening applies in every interview language.
• For medical devices, product complaints and potential adverse events raised in an interview are reported to your complaint-handling team within 24 hours, with a reconciliation at study end. Source: Carevoices adverse-event and product-complaint procedure. Regulatory reporting obligations remain the sponsor’s determination.
• Participant consent includes the adverse-event disclosure the market-research codes require. Source: Carevoices consent template, written to the EphMRA and BHBIA adverse-event guidance.

What does Safe Harbor de-identification remove?

Deliverables are de-identified to the HIPAA Safe Harbor standard: all 18 identifiers listed in the regulation are removed, from names and geographic subdivisions smaller than a state to dates, contact details and record numbers. Insights deliverables never include participant identities, and no audio or video is shared by default.

Sources: 45 CFR §164.514(b)(2); HHS de-identification guidance.

• Every deliverable is de-identified to the HIPAA Safe Harbor standard, with all 18 identifiers removed. Insights deliverables never include participant identities, and no audio or video is shared by default. Contact details reach the sponsor’s drug-safety team only if the clinician agrees to adverse-event follow-up; any unblinded work is agreed in the study plan. Source: 45 CFR §164.514(b)(2), HHS guidance on the Safe Harbor method.

Which market-research codes does Carevoices follow?

Studies are run in line with the three codes healthcare market research is held to.

• EphMRA Code of Conduct
• BHBIA Legal and Ethical Guidelines
• Insights Association Code of Standards and Ethics

• Run in line with the EphMRA, BHBIA and Insights Association codes of conduct. Source: EphMRA Code of Conduct, BHBIA Legal and Ethical Guidelines, Insights Association Code of Standards and Ethics.

Who are the participants, and how are they verified?

Physicians, specialists and KOLs, NPs, PAs and pharmacists are recruited through our physician network partner, whose public figures describe a verified network covering over 85% of US physicians and an increasing number of NPs, PAs and pharmacists. These are the partner’s figures, not Carevoices figures. Nurses, pharmacists and allied health professionals are also recruited through the healthcare-specific panels Carevoices integrates with. Sponsors and agencies can bring their own lists with the same sponsor blinding, adverse-event screening and de-identification (a sponsor-supplied list means the sponsor knows who was invited, which can affect Sunshine Act reporting).

Specialties covered include: Allergy & Immunology, Anesthesiology, Cardiology, Dermatology, Emergency Medicine, Endocrinology, Family Medicine, Gastroenterology, General Surgery, Hematology / Oncology, Infectious Disease, Internal Medicine, Neonatology / Perinatology, Nephrology, Neurology, Neurosurgery, Obstetrics & Gynecology, Ophthalmology, Orthopaedic Surgery, Otolaryngology (ENT), Pathology, Pediatrics, Physical Medicine & Rehabilitation, Plastic Surgery, Psychiatry, Pulmonology, Radiation Oncology, Radiology, Rheumatology, Thoracic Surgery, Urology, Vascular Surgery.

• Recruit from a verified network that includes 85%+ of US physicians, plus NPs, PAs and pharmacists, through our physician network partner. Source: Our physician network partner’s public company page (“over 85% of physicians and an increasing number of NPs, PAs, and pharmacists”).
• Nurses, pharmacists and allied health professionals through the healthcare-specific panels Carevoices integrates with. Source: Carevoices recruitment integrations.
• Every clinician is identity-, license- and NPI-verified before an invitation; roles without an NPI are identity- and license-verified. Source: Our physician network partner’s verification process; Carevoices screening.
• Roles without an NPI are identity- and license-verified before an invitation. Source: Carevoices recruitment integrations and screening.

How are honoraria paid?

Participants receive fair-market-value honoraria, calibrated to specialty and interview length and paid in compliance with the Sunshine Act. Honorarium amounts and timing are set out in each study invitation.

Reference: CMS Open Payments.

• Fair-market-value honoraria, paid in compliance with the Sunshine Act. Source: Carevoices honoraria policy; partner payment process.

How is research data protected?

Security and access controls apply to every study and to the research repository. The SOC 2 Type II report and the penetration-test summary are available under NDA, alongside the BAA template and the DPA.

• HIPAA-enabled, with a BAA available. Source: Carevoices BAA template, available on request.
• SOC 2 Type II compliant. Source: SOC 2 Type II report, available under NDA.
• Independent third-party penetration test; penetration-test summary available under NDA. Source: Penetration-test summary, available under NDA.
• Encrypted in transit (TLS 1.2+) and at rest (AES-256). Source: Carevoices security controls; detailed in the SOC 2 Type II report.
• Customer and participant data is never sold and never used to train AI models. Source: Carevoices data-use policy and DPA.
• US data residency. Source: Carevoices hosting configuration (US regions).
• Data is deleted within 30 days of a written request, unless the law requires it to be kept. Source: Carevoices Privacy Policy (retention) and DPA Section 11.
• Single sign-on and audit logs. Source: Carevoices authentication and access controls.

What does "80+ languages" mean?

The number of languages in which interviews run directly in the participant’s language, with the same laddering depth, rather than through a translated script. Adverse-event screening and de-identification apply in every interview language. Original-language transcripts are kept alongside English versions for source verification.

• 80+ languages. Source: Languages in which interviews run directly, with the same laddering depth (not translated scripts).

How are "24 hours" and "1–2 weeks" measured?

"24 hours" runs from receiving your brief to sending a proposal. "1–2 weeks" runs from approval of the study plan to delivered results. Rare and low-incidence specialties can take longer, and the proposal states the expected timeline when they do.

• A proposal within 24 hours of your brief, and results in 1–2 weeks from plan approval. Source: 24 hours runs from brief received to proposal sent; 1–2 weeks runs from study-plan approval to delivered results. Rare and low-incidence specialties can take longer.

What is the quality bar?

An interview counts toward your plan only when it is complete and meets the quality bar: the participant is the verified clinician invited, answers are on topic and substantive, and the interview passes the quality controls below. Interviews that fall short do not count.

• Only completed interviews that meet our quality bar count toward your plan. Source: Carevoices quality bar, applied to every interview before it counts.
• Every voice interview is screened for AI-generated answers through answer-similarity scoring and response-latency analysis, engagement is monitored as the interview happens, every interview is scored on length, depth and coverage of the research objectives, and interview links block VPNs and data-center addresses. Interviews that fall short do not count toward your plan. Source: Carevoices interview quality controls, applied to every interview; ties to the quality bar (quality_billing).

What goes into the research repository?

Every Carevoices study, plus the transcripts, reports and survey exports you upload. Questions asked across the repository return synthesized answers that cite the studies behind them, under the same access controls as the rest of your account. Uploads are stored as provided: Carevoices de-identifies only the studies it runs, so de-identify any upload that contains participant identifiers before adding it.

• Every study lands in one searchable research repository, alongside the past qual and quant you upload. Ask a question across all of it and get a synthesized answer that cites its sources. Source: Carevoices research repository. Uploads are stored as provided; Carevoices de-identifies only the studies it runs.

Which industry figures do we cite, and where do they come from?

Third-party figures are listed with their original source and the scope they support.

• An AI agent passed 99.8% of survey attention checks. Westwood, PNAS (2025).
• Researchers discard an average of 38% of online survey data. Kantar, The State of Online Research Panels (2022).

Updates to this page

Questions about how a study runs: proposals@carevoices.health.