Billing for Treatment Centers: In-House, Outsourced, or Automated?

If you run billing for a behavioral health treatment center, the question "should we keep this in-house or outsource it?" is older than most of the codes you bill. What's changed is the other side of the table.

Payers are no longer adjudicating your claims by hand. They run them through automated systems that can deny in seconds, at a scale no billing team can match by appealing one claim at a time. That shift doesn't just make denials more annoying. It changes the math on how you staff and tool your revenue cycle. A billing model that was "good enough" against a slow, human payer can quietly bleed six figures a year against an automated one.

This guide is written for the operator making that call: an owner, COO, or RCM lead at a multi-site IOP, PHP, residential, or SUD program with real claim volume and a messy payer mix. It isn't for the solo therapist wondering whether to learn modifier rules. We'll lay out the three real models, the honest cost of each, where each genuinely wins, and how the automated-payer shift should tilt the decision. No pitch. Just the trade-offs as they actually are.

What billing actually costs a treatment center at volume

Before comparing models, get honest about what you're paying today. Most of the cost is invisible on the billing line item.

A treatment center's revenue leaks in five places at once: denials that never get reworked, days in A/R that tie up cash, soft denials (silent downcoding) nobody tracks, write-offs past timely filing, and the staff time spent fighting all of it. Behavioral health runs hot on every one of these. Practices land at 68–85% clean claims against a 90–95% medical baseline. That 15–30% gap is revenue sitting in limbo at any given moment.

Sources: Global AHS - BH Denial Rate Benchmarks · ICANotes - BH Billing Metrics & KPIs 2025 · Supahealth aggregate data from 200+ behavioral health practices.

The math compounds fast at scale. Every 1% reduction in denial rate recovers roughly $4,000–$8,000 in annual revenue per full-time clinician. A 20-clinician center moving from 18% to 10% denials isn't recovering a rounding error. It's recovering a hire. The level-of-care services that define treatment centers (IOP, PHP, residential) deny at the highest rates in behavioral health, 20–30%, because they're the most prior-auth- and concurrent-review-intensive. So your billing model isn't a back-office choice. It's one of the larger levers you have on margin.

The payer side is industrializing denials

Here's the part that's genuinely new, and the reason a 2019-era billing setup is riskier than it looks.

Payers have spent the last few years automating adjudication. Consider the most-documented example. An investigation found Cigna used an automated system, "PxDx," to reject over 300,000 claims in two months, an average of about 1.2 seconds per claim, with physicians signing off on batches without opening patient files. One former reviewer described it this way: "we literally click and submit. It takes all of 10 seconds to do 50 at a time."

It's not just one payer. UnitedHealth faced a class-action lawsuit over an algorithm called nH Predict used to deny post-acute care. The suit alleged a 90% error rate, meaning roughly nine in ten of its denials were overturned when patients pushed back. A 2024 U.S. Senate Permanent Subcommittee on Investigations report found major Medicare Advantage insurers leaned on automation to deny prior authorization for post-acute care.

Sources: ProPublica - How Cigna Saves Millions by Rejecting Claims Without Reading Them (2023) · STAT News - Denied by AI: UnitedHealth algorithm investigation (2023)

The provider-side burden is measurable. In the AMA's annual prior-authorization survey of practicing physicians, 95% report that prior auth causes care delays, 79% say it leads patients to abandon treatment, and 26%, more than one in four, report that a prior-auth delay led to a serious adverse event. The volume is staggering. Medicare Advantage insurers alone made nearly 53 million prior-authorization determinations in 2024. Denial rates are climbing year over year across the industry, not falling.

Sources: AMA - Prior Authorization Physician Survey · KFF - MA Insurers Made Nearly 53 Million Prior-Auth Determinations in 2024

To be fair, automation isn't inherently the villain. The same systems can approve legitimate claims faster, and not every denial is wrong. The problem is the asymmetry. Payers adjudicate at machine scale while most providers still appeal by hand, one fax at a time. The only durable answer is to bring comparable leverage to your side of the cycle. That's the lens to carry into the three models below: which one lets you fight an automated payer without burning out a billing team?

The three models for treatment-center billing

Strip away the marketing and there are three ways to run a treatment center's revenue cycle:

1. In-house RCM team: you hire, train, and tool billers, coders, and AR specialists.
2. Outsourced RCM / billing company: a third party runs billing for a percentage of collections.
3. Agentic billing platform: software with behavioral-health-trained agents that run the rules-based work, with your team supervising the edges.

These aren't mutually exclusive. Many centers blend them (in-house front-end eligibility, outsourced back-end appeals, or a platform augmenting a lean internal team). But the dominant model sets your cost structure and your ceiling on denial performance. Let's take them one at a time, honestly.

Model 1: In-house RCM team

When it wins: large, stable claim volume; a concentrated payer mix you know cold; leadership that wants full control and visibility; and enough scale to afford redundancy when someone quits.

An in-house team gives you the tightest feedback loop. Your billers sit down the hall from your clinicians, so a recurring documentation gap can get fixed in a team meeting, not a support ticket. You own the data, the relationships with your top payers, and the institutional memory of which Medicaid MCO downgrades which code.

The true loaded cost is higher than the salaries. A treatment-center RCM function isn't one biller. It's eligibility/auth, charge entry, coding, claims follow-up, denials/appeals, and patient billing, often 3–8 people depending on volume. Add software (clearinghouse, scrubbing, reporting), training to keep up with payer changes, and management overhead. Then add the cost nobody budgets: turnover. Billing roles churn. When your denials specialist leaves, your appeal backlog and A/R days spike for the two months it takes to rehire and ramp. For an auth-heavy IOP/PHP center, that gap is expensive.

In-house also struggles with the new payer reality on its own. A human team simply cannot scrub every claim against every payer's current rules across 50 states at the speed an automated payer denies them. In-house works best with tooling, not instead of it.

Model 2: Outsourced RCM / billing company

When it wins: you're scaling fast and can't hire RCM talent quickly enough; you want to convert a fixed cost into a variable one; or your in-house function is underwater and you need a competent operator now.

A good billing company brings a trained team, existing payer relationships, and processes on day one. You typically pay 4–9% of collections (sometimes higher for complex behavioral health), which scales with revenue rather than sitting as fixed payroll. For a center bleeding revenue through an overwhelmed internal team, that can be an immediate upgrade.

The honest downsides are real and worth naming:

• Incentive misalignment. A billing company paid on collections optimizes for its margin. It will happily collect the easy 80% and may not chase the hard, low-dollar denials that are expensive to appeal, exactly the ones that add up at a treatment center.
• Visibility loss. Your denial data lives in their system. You learn about a systemic problem later, and your clinicians don't get the tight feedback loop that prevents denials upstream.
• Behavioral-health specificity. Many general medical billing companies treat BH as an afterthought and miss the level-of-care, concurrent-review, and parity nuances that drive your denials.

Outsourcing solves a staffing problem. It does not, by itself, solve the automated-payer problem. You've just hired humans to fight machines on your behalf, and you're paying a percentage for it.

Model 3: Agentic billing platform

When it wins: you want to keep control and visibility and match the payers' automation; you have a complex, multi-state payer mix where rules change constantly; or you want to make a lean in-house team dramatically more productive instead of growing headcount linearly with volume.

An agentic platform runs the high-volume, rules-based work continuously: eligibility checks, benefit and auth verification, claim scrubbing against payer-and-state rules, denial routing, and appeal drafting. It learns your specific payer patterns over time. Your team moves up the value chain to supervision, exceptions, and the judgment calls software shouldn't make.

The honest limits matter, and any vendor who won't name them is selling you something. A platform can't override a payer's policy. When a payer moves a code to non-covered, good software catches the change and stops submitting, but it can't make the payer pay. It can't fix a wrong diagnosis or make a clinically thin session billable. And it won't rescue a broken underlying clinical workflow. What it does fix is the cross-team, cross-payer, cross-state work humans get wrong simply because there's too much of it to hold in working memory. That's precisely the work the automated payers are exploiting.

The hidden costs nobody quotes you

When you compare a billing company's "7% of collections" to a biller's salary, you're comparing the wrong numbers. The real cost of a billing model is the revenue it fails to capture:

• Denial rework at $25–$70 of staff time per claim, even when the claim eventually pays.
• Days in A/R: cash you've earned but can't use. Same-day note completion and clean first-pass claims pull this down 5–10 days; a slow model pushes it up.
• Soft denials / silent downcoding: a 90837 paid at the 90834 rate. At 8–10% of claims quietly downgraded by $30–$50, that's $2,400–$5,000 lost per 1,000 claims, and most centers never track it.
• Credentialing lag: clinicians seeing patients before they're in-network with your payers, generating unbillable sessions.
• Timely-filing write-offs: claims that age out past 90–180-day windows and die with no appeal.

Sources: HFMA - Navigating the Rising Tide of Denials · CareCloud - How Much Are Denials Really Costing You?

A denial that reaches the appeal stage is 5–10× more expensive than one prevented before submission. The model that wins is the one that prevents the most denials upstream, not the one with the lowest sticker price.

A decision framework

There's no universal answer. There's a fit. Use your own numbers:

Three questions decide it. What's your claim volume? (Low volume rarely justifies a full in-house team.) How complex and changeable is your payer mix? (More complexity favors automation over manual processes.) How mature is your in-house RCM expertise? (Strong leadership makes in-house plus tooling powerful; its absence makes outsourcing or a platform safer.)

Cost and capability, side by side

Sources: HFMA - Standardizing Denial Metrics · Experian Health - State of Claims Report 2025

AI and agentic systems in behavioral health billing

Step back and the through-line is clear: denials are rarely a single-point failure. A denial traces upstream to a benefits check that missed a session limit, a code that didn't match documented time, a note that didn't establish medical necessity, or a payer rule that changed last quarter. Fixing denials means fixing all of those places at once, continuously, and at the speed payers now operate.

That's the idea behind Supabill. It isn't a single billing tool. It's an integrated system of behavioral-health-trained agents that work together, 24×7, across the whole revenue cycle. The benefits verification agent pulls eligibility and surfaces session limits, prior-auth requirements, and copay structure before the first visit. It works across 5,000+ payers, and when a portal won't answer, it can place an actual call to the payer to get the benefits. When it learns a payer caps psychotherapy at 20 sessions a year, the claims-scrubbing agent flags that patient's submissions as they approach visit 18, before an avoidable denial. The claims agent holds state-by-state and payer-by-payer rules in its core database and checks every claim pre-submission: time-vs-code, modifier, POS, prior-auth match, bundled-code conflicts. Whatever still gets denied flows to the denials-management agent, which routes by code and payer, drafts appeals against matched clinical context, and tracks resubmission cycles. And when that agent learns a payer is rejecting a code for a specific missing element, that signal feeds back into the documentation guardrails in Supanote, so the note prompts for it next time. Each agent makes the others smarter. A denial reason learned once becomes a denial prevented everywhere.

The point isn't to replace your team. It's to give the provider side the same leverage the payer side already has, so a treatment center isn't bringing a stopwatch to an algorithm fight. Your people stop doing the rote, cross-payer rule-checking that humans get wrong at volume, and move to the judgment work that actually needs them.

And the honest limits, again: agents can't change payer policy, fix a wrong diagnosis, or rescue a broken clinical workflow. They fix the high-volume, rules-based, cross-team work, which happens to be exactly where automated payers are winning today.

If you're interested, book a demo here to learn more.

Quick wins you can act on this quarter

1. Calculate your true denial cost. Pull your last quarter: denial rate, average denial value, days in A/R, and an estimate of soft-denial leakage by comparing expected vs actual reimbursement per CPT. The number is usually bigger than leadership thinks, and it's the baseline for any model decision.
2. Audit your last 50 denials by cause (prior auth, coding, eligibility, medical necessity, documentation, timely filing, payer policy). The shape tells you whether your problem is a staffing problem, a tooling problem, or both.
3. Pressure-test any vendor or platform on behavioral-health level-of-care. Ask specifically how they handle IOP/PHP concurrent review and ASAM medical necessity. Generic answers are a red flag.
4. Re-verify eligibility and auth 24–48 hours before every visit. It's the single workflow change that closes the most front-end denials, whatever model you run.
5. Track first-pass clean claims rate as a leading KPI, not just denial rate. HFMA's target is 95%+; structured docs and pre-submission scrubbing get behavioral health centers to 92–95%.

FAQ

At what claim volume does an in-house RCM team beat outsourcing? There's no universal threshold, but most centers find a full in-house function hard to justify below roughly 1,000–1,500 claims/month. Below that, you can't keep specialized billers, coders, and appeals staff productive, and turnover risk is high. Above it, in-house plus tooling starts to win on control and feedback loops. The bigger driver than raw volume is payer-mix complexity and whether you have strong RCM leadership to run the team.

Is an agentic platform a replacement for billers, or a tool for them? A tool, and that distinction matters. The platform runs the rules-based, high-volume work (eligibility, scrubbing, appeal drafting); your team supervises exceptions and makes judgment calls. Centers that try to eliminate billing staff entirely usually find the edge cases pile up. The win is productivity per biller, not zero billers.

How much does a behavioral health billing company actually cost? Typically 4–9% of collections, sometimes higher for complex level-of-care work. Compare that not to a biller's salary but to the revenue the model captures. A cheaper vendor that ignores your low-dollar denials and soft downcoding can cost more than a pricier one that doesn't.

Why do IOP and PHP claims deny more than outpatient claims? They're the most prior-auth- and concurrent-review-intensive services in behavioral health, with denial rates around 20–30% versus 15–20% for individual outpatient. Medical necessity for continued stay is challenged frequently, step-down criteria are disputed, and group-billing adds complexity. Any model you choose has to be strong specifically here.

Are payers really using AI to deny claims, or is that overstated? It's documented. Reporting on Cigna's PxDx system showed hundreds of thousands of claims denied at roughly a second each, and UnitedHealth faced litigation over an algorithm with an alleged 90% error rate on post-acute denials. A 2024 Senate investigation found Medicare Advantage insurers used automation to deny prior auth. The takeaway for operators isn't outrage. It's that manual, one-at-a-time appeals don't scale against automated denial, so your billing model has to.

Will outsourcing fix my denial problem? It can fix a capacity problem fast. It doesn't automatically fix a prevention problem. A billing company appealing denials after the fact still leaves the upstream causes (eligibility gaps, doc gaps, coding errors) in place, and you lose visibility into them. The lowest denial rates come from preventing denials upstream, which requires tight feedback loops or continuous automated scrubbing.

How do I keep visibility if I outsource? Negotiate for it explicitly: direct access to your denial data, monthly reporting by denial cause and payer, and a defined process for feeding systemic issues back to your clinicians. If a vendor won't give you your own denial data in a usable form, treat that as disqualifying.

What KPIs should I watch regardless of model? First-pass clean claims rate (target 95%+), denial rate by payer and cause, days in A/R, net collection rate, and soft-denial leakage (expected vs actual reimbursement per CPT). Track them monthly and per payer. A single blended number hides the payer that's quietly downcoding you.

Can I blend models? Yes, and many centers do. For example: in-house front-end eligibility and auth, an agentic platform for scrubbing and denial routing, and a specialized vendor for a specific complex payer. The goal isn't purity. It's the lowest total revenue leakage for your volume and payer mix.

How does the payer-automation trend change the decision over the next few years? It raises the cost of any model that relies on humans manually keeping up with payer rules and manually appealing automated denials. Expect the gap between centers that automate their own revenue cycle and those that don't to widen. The asymmetry compounds. Build for a payer that adjudicates by machine, because that's the payer you have.

References

• AMA. (2024). Prior Authorization Physician Survey. https://www.ama-assn.org/practice-management/prior-authorization
• CareCloud Continuum. (2025). How Much Are Denials Really Costing You? https://carecloud.com/continuum/how-much-are-denials-really-costing-you/
• Centers for Medicare & Medicaid Services. (2024). Prior Authorization and Pre-Claim Review Program Statistics, FY 2024. https://www.cms.gov/files/document/pre-claim-review-program-statistics-document-fy-24.pdf
• Experian Health. (2025). State of Claims Report 2025. https://www.experian.com/blogs/healthcare/healthcare-claim-denials-statistics-state-of-claims-report/
• Global AHS. (2025). Behavioral Health Billing Denial Rate Benchmarks. https://www.globalahs.com/behavioral-health-billing-denial-rate-benchmarks/
• Healthcare Dive. (2025). More insurance claim denials are being overturned upon appeal. https://www.healthcaredive.com/news/insurance-denials-overturned-appeal-new-york-study-JAMA/817490/
• HFMA. (2025). Navigating the Rising Tide of Denials. https://www.hfma.org/revenue-cycle/denials-management/navigating-the-rising-tide-of-denials/
• HFMA. (2024). Claim Integrity Task Force: Standardizing Denial Metrics. https://www.hfma.org/guidance/standardizing-denial-metrics-revenue-cycle-benchmarking-process-improvement/
• HHS OIG. (2023). High Rates of Prior Authorization Denials in Medicaid Managed Care (OEI-09-19-00350). https://oig.hhs.gov/reports/all/2023/high-rates-of-prior-authorization-denials-by-some-plans-and-limited-state-oversight-raise-concerns-about-access-to-care-in-medicaid-managed-care/
• ICANotes. (2025). Behavioral Health Billing Metrics & KPIs 2025. https://www.icanotes.com/2025/09/24/behavioral-health-billing-metrics-kpis/
• KFF. (2025). Claims Denials and Appeals in ACA Marketplace Plans in 2024. https://www.kff.org/patient-consumer-protections/claims-denials-and-appeals-in-aca-marketplace-plans-in-2024/
• KFF. (2025). Medicare Advantage Insurers Made Nearly 53 Million Prior Authorization Determinations in 2024. https://www.kff.org/medicare/medicare-advantage-insurers-made-nearly-53-million-prior-authorization-determinations-in-2024/
• ProPublica. (2023). How Cigna Saves Millions by Having Its Doctors Reject Claims Without Reading Them. https://www.propublica.org/article/cigna-pxdx-medical-health-insurance-rejection-claims
• SimiTree. (2025). How to Calculate and Improve Your Behavioral Health Clean Claims Rate. https://simitreehc.com/simitree-blog/how-to-calculate-and-improve-your-behavioral-health-clean-claims-rate/
• STAT News. (2023). Denied by AI: How Medicare Advantage plans use algorithms to cut off care. https://www.statnews.com/2023/11/14/unitedhealth-algorithm-medicare-advantage-investigation/
• U.S. Senate Permanent Subcommittee on Investigations. (2024). Refusal of Recovery: How Medicare Advantage Insurers Have Denied Patients Access to Post-Acute Care. https://www.hsgac.senate.gov/subcommittees/investigations/
• Supahealth. (2026). Aggregate billing and denials data from 200+ behavioral health practices. Internal dataset.