Choose a collaborative care vendor by testing six things: fidelity to the evidence-based Collaborative Care Model, the qualifications of the care managers and psychiatric consultants, a real measurement-based registry, hands-on billing support, transparent outcomes reporting, and the language and social-needs capabilities your population actually requires. Score each vendor against those criteria, not the sales deck.

Collaborative care (the "Collaborative Care Model," or CoCM) has more than 90 randomized controlled trials behind it, but the trials worked because of specific operational details — a tracked caseload, a psychiatric consultant reviewing it, and treatment that changes when symptoms don't improve. A vendor can call a service "collaborative care" while dropping the parts that make it effective. The job of a buyer — a health plan, an ACO, or a primary care practice — is to tell the difference.

What is a collaborative care vendor?

A collaborative care vendor supplies the people, technology, and workflow that let a primary care practice treat common behavioral health conditions in-house instead of referring patients out. In a typical arrangement, the vendor provides behavioral health care managers, access to a consulting psychiatrist, a patient registry, and the billing infrastructure to capture CoCM codes — while the patient's own primary care clinician stays responsible for prescribing and overall care.

The market ranges from staffing shops that place a care manager and little else, to full-stack partners that run the model end to end. The evaluation below is designed to surface which one you are actually talking to.

How do you evaluate model fidelity?

Fidelity is the first filter because everything else depends on it. Ask each vendor to describe, concretely, how they deliver the core elements of CoCM:

• A defined caseload registry every patient is tracked in — not a spreadsheet that gets updated when someone remembers.
• Measurement-based treatment using validated tools (PHQ-9 for depression, GAD-7 for anxiety), with scores recorded at defined intervals.
• Systematic psychiatric caseload review, where the consulting psychiatrist reviews patients who aren't improving and advises on changes.
• Treatment-to-target — a stated protocol for what happens when a patient doesn't respond, rather than a single check-in and a note.

If a vendor can't explain how they do all four, they are running behavioral health integration in name, not the model the evidence supports.

What should you ask about staffing?

The care manager is the engine of the model, so staffing questions matter more than most buyers expect. Cover:

• Credentials and licensure of the care managers, and whether they are behavioral-health-trained.
• Caseload size per care manager — an overloaded panel is the most common way fidelity quietly breaks.
• Psychiatric consultant access — how many hours, how quickly a caseload review turns around, and whether the same consultant stays with the practice.
• Turnover and coverage — who covers a panel when a care manager is out, and what continuity the patient experiences.
• Onboarding and supervision — how new care managers are trained into the model and who supervises their clinical work.

Ask for the ratios in writing. Vague answers here usually predict thin service later.

What technology and registry capabilities matter?

CoCM does not work without a registry, so the technology question is really a fidelity question. Probe:

• Registry function — can the team see, at a glance, who is overdue for follow-up, whose scores are worsening, and who is ready to step down?
• EHR integration — does the vendor's system write back to your electronic health record, or does it create a parallel workflow staff have to double-enter?
• Measurement capture — how PHQ-9 and GAD-7 scores are collected and trended over time.
• Reporting access — whether your practice or plan can pull its own data, or has to request it.
• Security and compliance — HIPAA safeguards, data ownership, and what happens to the data if the contract ends.

A registry you can't see into is a registry you can't trust.

How much billing support does the vendor provide?

CoCM reimbursement runs through specific monthly codes (99492, 99493, 99494, and G2214), and capturing them correctly is where many programs stall. Ask whether the vendor:

• Tracks billable minutes per patient per month against the code thresholds.
• Handles or supports claim submission, or leaves it entirely to your billing team.
• Documents time and consent in a way that survives an audit.
• Understands your payer mix — Medicare and, in states including New York, Medicaid cover collaborative care, but rules and rates vary by plan and state.

A vendor that improves clinical care but leaves the billing broken will not be sustainable. Confirm coverage specifics with each payer.

How should outcomes be reported?

Educational point, but the sharpest one: a real CoCM partner measures whether patients get better, not just whether they were enrolled. Look for reporting that shows:

• Symptom improvement — the share of patients with a clinically meaningful drop in PHQ-9 or GAD-7 scores.
• Engagement and reach — how many enrolled patients are actually being contacted and followed, not just counted.
• Time to follow-up and treatment adjustment, which shows the model is running as designed.
• Cadence — reporting on a regular schedule, in a format you can act on.

Be cautious of any vendor that reports enrollment volume and leaves out whether those patients improved. Ask to see a sample report before you sign.

What about language and social-needs capabilities?

The best-designed program fails the patients it can't communicate with. Match the vendor's capabilities to your population:

• In-language care — care managers and behavioral health providers who speak the languages your members actually speak, not just an interpreter line.
• Cultural competence — experience with the communities in your service area.
• Social-needs screening — screening for social drivers of health (SDOH), often captured with Z-codes, and a real pathway to connect patients with community resources.

If a meaningful share of your population faces a language or social barrier, these move from "nice to have" to disqualifying if absent.

Frequently asked questions

What's the single most important thing to check in a CoCM vendor?

Model fidelity — specifically whether every patient is tracked in a registry, treated with measurement-based tools, reviewed by a psychiatric consultant, and adjusted when they don't improve. Without those four elements, it isn't the model the evidence supports.

How is collaborative care reimbursed?

Through monthly CPT/HCPCS codes (99492, 99493, 99494, and G2214) tied to time spent. Medicare covers it, and Medicaid covers it in states including New York, though rules and rates vary. A good vendor helps you capture these codes correctly; confirm specifics with each payer.

How is a collaborative care vendor different from a therapy referral network?

A referral network sends patients out to separate providers, and more than half of patients never complete that referral. A CoCM vendor embeds the care in primary care with a tracked caseload and psychiatric oversight, so treatment happens inside the practice the patient already visits.

What outcomes should a vendor be able to show?

Symptom improvement measured by validated tools (PHQ-9, GAD-7), the share of enrolled patients actually being reached, and time to follow-up — reported on a regular cadence. Enrollment counts alone are not outcomes.

Should we pilot before committing?

Yes. A time-boxed pilot in one or two practices lets you verify caseload ratios, registry visibility, billing capture, and reporting quality against the vendor's claims before you scale.