This guide is for the leaders who own that risk: member services, compliance, and vendor management. It explains what the rules cover and what changed for contract year 2027. It also shows where call centers most often slip and how to build compliance into daily operations. Throughout, it separates what CMS requires from the controls we recommend, and each rule cited links to its source.

What the Medicare Marketing Guidelines Cover

For Medicare Advantage, the Medicare marketing guidelines are the rules in 42 CFR Part 422, Subpart V. Part D plans follow parallel rules in 42 CFR Part 423, Subpart V, and CMS guidance sits alongside both. They govern how plans, and anyone acting for them, communicate with beneficiaries. That includes plan call centers, outsourced member services teams, and third-party marketing organizations (TPMOs).

The rules turn on one distinction. Under 42 CFR 422.2260, a communication becomes marketing when it meets two tests. First, it aims to draw attention to a plan or influence enrollment, including retention. Second, it addresses benefits, premiums, cost sharing, or rankings such as Star Ratings. Marketing carries far stricter rules than ordinary communications.

That distinction matters on the phone. A member services agent who answers a claims question is communicating. However, the same agent who compares benefits to keep a member from switching plans may be marketing. So leaders need call flows that keep agents on the right side of that line. Otherwise, the flow should hand the call to a licensed agent.

What Changed in the Medicare Marketing Guidelines for 2027

CMS published the contract year 2027 final rule in the Federal Register on April 6, 2026. Its marketing changes apply from October 1, 2026, according to a CMS partner tip sheet. Most changes reduce burden. Still, several require script and system updates before AEP calls peak.

The timing change for the disclaimer deserves attention. Agents trained to read it in the first minute may now read it too late. That happens whenever a caller raises benefits early. Update scripts so the disclaimer comes before any benefit question gets an answer. The current wording sits in 42 CFR 422.2267(e)(41), with a parallel Part D provision.

What Medicare Call Center Leaders Should Update Before and During AEP

The actions below are our operational recommendations, not additional CMS requirements. Each one also ties back to a rule covered in this guide. Together, they give compliance teams a short brief for internal teams and outsourced partners.

Test these controls against recorded calls, not only against written scripts. After all, live calls show whether agents follow the new order under real volume.

Medicare Marketing Rules That Did Not Change

Deregulation headlines can create false comfort. In fact, many core controls stay in place. Under 42 CFR 422.2264, plans still may not make unsolicited contact by cold call, robocall, text message, or voicemail. Contact is not unsolicited when the beneficiary consents to it or initiates it. That is the main route for permitted prospecting, though the regulation sets out further detail. Calls to current members about plan business follow a separate set of rules. Plans may make those calls, but they must give members an annual written notice explaining how to opt out.

The scope of appointment requirement also survives. Plans must still agree on and record a scope of appointment before any personal marketing appointment. In-person appointments need it in writing. What changed is timing alone: the 48-hour wait between completing the scope and holding the appointment is gone. Once completed, a scope of appointment stays valid for 12 months after the beneficiary signs it.

TPMO oversight rules remain strict as well. Under 42 CFR 422.2274, TPMO contracts must require recording of all marketing and sales calls. That includes the audio portion of web-based calls. TPMOs must report staff disciplinary actions and violations to the plan monthly. They must also tell a beneficiary when a transfer to a licensed agent happens. Agents and brokers still face annual training and testing, with a passing score of 85% or higher.

Note the scope of that recording rule. It applies to marketing and sales calls under TPMO arrangements. It does not, on its own, cover routine member service calls about claims or ID cards. Plans set policies for those calls under their broader record retention obligations.

CMS Call Center Standards Still Apply During AEP

Removing two Star Ratings measures does not repeal the underlying standards. For Medicare Advantage, 42 CFR 422.111(h) still sets performance floors for plan customer call centers. Part D plans face matching standards under 42 CFR 423.128(d). As a result, these floors matter most during AEP, when call volume peaks and hiring lags.

Interpreter access is where many plans feel pressure. Spanish is usually the largest need, so language coverage drives this metric. A team of native bilingual agents removes the interpreter step for those calls entirely. Our article on the Spanish-language service gap explains why that demand keeps growing.

Where Call Centers Most Often Break the Medicare Marketing Guidelines

Compliance failures rarely come from bad intent. They come from scripts, transfers, and systems that drift away from the rules under volume pressure. The table shows the patterns compliance teams find most often. The controls are our recommendations, not methods CMS prescribes.

Notice that most controls sit in systems and scripts, not in agent memory. Agents forget rules under pressure, especially in their first AEP. Therefore, the safest programs make the compliant path the default path.

How to Build Medicare Compliance Into Call Center Operations

Start with scripts and call flows. Review every flow against the 2027 changes before AEP traffic peaks. Then put the rules into quality assurance. Our guide to call center quality assurance shows how to make missed disclaimers and silent transfers auto-fail items. Auto-fail scoring is a recommended safeguard, not a CMS-mandated method. Also score a higher share of sales and retention calls than routine service calls, because their risk is higher.

Next, revisit recording and storage. The shorter retention period lowers storage cost, and transcripts can replace audio in years four to six. However, a transcript must be complete and accurate to qualify. Confirm that your platform can produce one before you delete any audio.

Finally, plan for oversight of every partner. Plans remain accountable for their first tier, downstream, and related entities. Partners outside the United States add a separate obligation. Since 2008, CMS has required sponsors to report offshore subcontractors that use beneficiary PHI. Sponsors also attest to how they protect that information. The 2008 memo does not define offshore. Even so, a nearshore center in Latin America sits outside the United States, so plans should confirm with their compliance team whether reporting applies.

Responsibilities here split three ways. The plan files any required report and attestation with CMS. The partner supplies accurate facts and cooperates with audits. A business associate agreement is a separate HIPAA requirement, and it does not satisfy the CMS reporting duty by itself. Our article on HIPAA and PCI compliance questions for nearshore BPOs covers the security side.

What to Require From a Medicare Call Center Partner

Evaluate a partner on evidence, not promises. Ask to see scripts already updated for the 2027 disclaimer timing. Ask how the partner records web-based calls and how it produces transcripts for later retention years. Request a sample of the monthly violation report it would send you.

Next, test capacity. AEP demands a fast ramp, and the call center standards apply on the first day. Ask how the partner will hit 80% of calls within 30 seconds while new agents learn. Also ask how many native Spanish speakers it can staff, and how it covers TTY calls. Finally, confirm that the partner will supply the information your offshore reporting needs. Separately, put a business associate agreement in place wherever HIPAA requires one.

A partner should also know where its role ends. Licensed sales belong with licensed, appointed agents. Member services, grievance intake, and outreach to current members can sit with a trained partner team. Our Medicare outsourcing services follow that division, and our health plan outsourcing page covers the wider member journey.

Conclusion

The 2027 Medicare marketing guidelines ease several burdens, but they do not ease accountability. Disclaimer timing changed, retention shortened, and the 48-hour wait disappeared. Meanwhile, the bans on unsolicited contact, the recording mandate, and the call center standards all remain. Plans that update scripts, systems, and partner contracts now will enter AEP with fewer surprises.

The broader market raises the stakes further. Our analysis of the Medicare Advantage enrollment shift shows members switching plans more actively. Every retention call is therefore both an opportunity and a compliance test. You can review the standards we hold on our certifications page.

Frequently Asked Questions

What are the Medicare marketing guidelines?

They are the CMS rules for how Medicare Advantage and Part D plans communicate with beneficiaries. They also govern marketing to them. Medicare Advantage rules sit in 42 CFR Part 422, Subpart V, and Part D rules in Part 423, Subpart V. They apply to plans, their call centers, and third-party marketing organizations acting for them.

When must the TPMO disclaimer be read on a call?

From October 1, 2026, the disclaimer must be read during sales calls before any discussion of benefits. The earlier rule required it within the first minute. The wording now refers callers to Medicare.gov or 1-800-MEDICARE and no longer mentions SHIPs.

How long must Medicare sales calls be recorded and kept?

TPMOs must record all marketing and sales calls, including the audio of web-based calls. Recordings must be kept for six years. The first three years must be audio. After that, a complete, accurate transcript may replace audio in years four to six. This rule covers marketing and sales calls, not routine member service calls.

Is the 48-hour scope of appointment rule still in effect?

No. The contract year 2027 rule removed the 48-hour waiting period. A personal marketing appointment can now follow a completed scope of appointment at any time. However, the plan must still complete a scope of appointment before the appointment. Once completed, it stays valid for 12 months.

Can a health plan call center make outbound calls to prospects?

Generally, only when the beneficiary has consented or initiated contact. Cold calls, robocalls, unsolicited texts, and voicemails remain prohibited. Calls to current members about plan business follow separate rules, including an annual written opt-out notice.

Do member services calls count as marketing?

Not usually. A call becomes marketing when it aims to influence enrollment or retention. It must also address benefits, premiums, cost sharing, or rankings. Service calls about claims or ID cards are communications. Retention conversations that compare benefits can cross into marketing.

The post Medicare Marketing Guidelines for 2027: What Health Plan Call Centers Must Get Right appeared first on SkyCom Call Center.

This guide shows how to close it. It treats QA as an operating system, not a scoring form. You will learn how to write observable standards, then build and pilot a weighted scorecard. It also covers risk-based sampling, calibration, governance, coaching, AI, and a QA scorecard template you can adapt. The aim is simple: a program where a better score means a better customer experience.

What Call Center Quality Assurance Actually Measures

Call center quality assurance is the process of reviewing customer interactions against a defined standard. Evaluators listen to calls or read chats, score them on a scorecard, and use the results to coach agents. Quality monitoring is the narrower step of observing and recording those interactions. QA adds the standard, the scoring, and the follow-through. Done well, it answers two questions. Did the agent do the right things? And did those things help the customer?

Most programs answer only the first question. They check whether the agent used the greeting, verified identity, and read the closing script. Those behaviors matter, but they are inputs. The outcome is whether the customer got an accurate answer with little effort. A strong program measures both, and it weights them so the outcome counts most.

Standards bodies treat quality as part of the whole operation, not a side task. ISO 18295-1 sets service requirements for customer contact centers, whether in-house or outsourced. In other words, QA is one control inside a wider service system. It works best when it connects to training, workforce planning, and process fixes.

Why Most QA Programs Fail to Move CSAT

Customers notice when quality stalls. Deloitte Digital published its 2026 Global Contact Center Survey this year. More than half of consumers said service quality stayed the same or got worse in 2025. Fewer than one in five leaders felt their current strategies were fully effective. Meanwhile, consumers estimated they spend 36% more with companies that give great service.

Five patterns usually explain the gap. First, the scorecard rewards script compliance more than resolution. Second, the criteria are vague, so evaluators score the same call differently. Third, the sample is too small or too random to show real risk. Fourth, scores never turn into coaching, so behavior does not change. Finally, nobody owns the scorecard, so disputes go unresolved and agents stop trusting the result. Each problem has a fix, and the steps below address them in order.

How to Build a Call Center Quality Assurance Program in 10 Steps

A QA program is a closed loop, not a form. Customer outcomes set the standards, and the standards shape the scorecard. The scorecard then drives sampling, calibration, and coaching. Finally, coaching has to show up in customer results, which feeds the next review. The ten steps below follow that loop.

Step 1: Define the Outcomes You Want

Begin with the customer, not the script. List the three or four outcomes that matter most for your contacts. Typical outcomes are an accurate answer, a resolution on the first contact, low customer effort, and full compliance. Then decide which agent behaviors drive each outcome. Those behaviors become the criteria on your scorecard.

Step 2: Define Observable Quality Standards

Every criterion should pass one test. What would an evaluator hear, see, or verify that proves the agent met it? Criteria such as “agent was professional” fail that test. Two evaluators will read them differently, and agents cannot tell what to change. Observable criteria fix both problems. They also make automated scoring possible later.

Rewrite each criterion until two evaluators would score it the same way without discussion. This takes time up front. However, it saves months of disputes once scores start to count.

Step 3: Build a Weighted Quality Assurance Scorecard

Weighting tells agents what matters. If every item counts equally, a polite greeting carries the same weight as a wrong answer. Instead, give the largest weight to resolution and accuracy. Then add a short list of auto-fail items that zero the score, whatever else happened. The template later in this guide shows one way to structure it.

Step 4: Pilot the Scorecard Before Using It for Performance Management

Do not attach a new scorecard to performance reviews right away. Test it on real interactions first. Have at least two evaluators score a representative set of contacts across your main contact types. Then look for criteria that everyone passes and criteria that cause frequent disagreement. Also check whether high scores line up with good survey results.

Revise the scorecard before you connect scores to coaching plans, incentives, or corrective action. Run the pilot long enough to cover your main contact types. Include both peak and quiet periods, because agent behavior changes under load.

Step 5: Build a Risk-Based Sampling Strategy

Sampling is a measurement strategy, not a quota. Consider an agent who handles 1,000 contacts a month. Reviewing five of them covers 0.5% of that agent’s work. Pooled across a team, small samples can still show trends. However, they are too thin for a high-stakes decision about one person.

So build three layers. Random sampling measures overall performance without reviewer bias. Targeted sampling pulls escalations, repeat contacts, low CSAT, and unusually long interactions. It also covers new agents and recently coached agents. Risk-based sampling, the third layer, covers contacts where failure costs the most.

Risk depends on your business. Payment handling, identity verification, regulated disclosures, collections, and healthcare calls carry compliance risk. Cancellations, refund requests, repeated complaints, and competitor mentions carry churn risk. Not every interaction has the same impact, so coverage should follow the cost of failure. Report random and targeted results separately, because targeted samples skew low by design.

Step 6: Calibrate Evaluators and Build a Golden Interaction Library

Calibration keeps the score honest. Each month, several evaluators score the same interactions independently. Then they compare results and agree on the right answer for each criterion. Set a tolerance, for example five points, and retrain anyone outside it. Agents also benefit from joining a session now and then, because it shows them how scoring works.

Calibration works better with a shared reference set. Build a golden interaction library of scored, annotated examples. Include excellent calls, borderline calls, compliance failures, poor resolutions, and strong recoveries. If you use AI, add weak bot-to-agent handoffs too. Then use the library to train new evaluators, test scorecard changes, and onboard agents. Over time, it becomes your quality standard in practice.

Step 7: Set Up QA Governance and Agent Appeals

QA scores affect coaching, incentives, and sometimes employment decisions. That makes governance essential. Name one owner for the scorecard, with operations, training, and compliance as reviewers. Define who can change a criterion and who approves each change. Then version the scorecard and tell agents what changed and why.

Agents also need a formal appeal path. Let them dispute an evaluation within a set window, and have a second evaluator review it. Record each calibration ruling and appeal outcome as a scoring precedent. Those precedents settle future disputes quickly. Track the appeal rate, too. A rising rate usually points to an unclear criterion, not to difficult agents.

Step 8: Turn QA Findings Into Coaching

A score without a conversation changes nothing. Supervisors should coach on one or two behaviors at a time, using the recorded interaction as evidence. Agree on a specific action, then check it in the next review cycle. Recognize improvement as well as gaps. As a result, agents see QA as support rather than surveillance.

Measure the agent side of QA as well. Useful signals include the time from a QA finding to coaching and repeat coaching on the same issue. Score change after coaching matters too. A program can lift customer results while it erodes agent trust. Timely, consistent, and fair feedback prevents that.

Step 9: Connect QA to Customer Outcomes

Compare QA scores with what customers actually report. Match evaluated interactions to their post-contact survey results where you can. Track customer effort score (CES) alongside CSAT, because effort captures friction that satisfaction can miss. Repeat contacts within a few days are another honest signal. If high QA scores sit beside low CSAT, the scorecard measures the wrong things. In that case, revise the criteria, not the agents.

Step 10: Review and Update the Program Every Quarter

Products, policies, and channels change, so the scorecard must change too. Each quarter, remove criteria that every agent passes. Add criteria for new failure points that coaching and complaints reveal. For example, Gartner found that 60% of agents fail to promote self-service. When agents did mention it, 12% made negative remarks. If self-service matters to your strategy, that behavior belongs on the scorecard. Update the golden library and the scorecard version at the same time.

A Quality Assurance Scorecard Template You Can Adapt

The template below works for most voice and chat programs. Weights total 100, and auto-fail items override everything else. Each section uses observable criteria, so evaluators and automated tools can score it the same way. Adjust the sections to match your outcomes from Step 1.

Score each criterion as met, partly met, or not met. Allow a “not applicable” option too. That keeps evaluators from penalizing an agent for steps the call never needed. The downloadable version breaks these sections into 18 criteria and calculates the weighted score automatically. It also includes calibration and pilot worksheets.

How to Extend Call Center Quality Assurance Across Every Channel

Customers do not experience voice QA or chat QA. They experience one brand across every channel they use. So keep one set of core quality principles, then adapt the criteria to each channel. Accuracy and resolution apply everywhere. Response time, concurrency, and consent rules differ by channel.

Pay special attention to handoffs between channels and from bots to agents. Customers dislike repeating themselves. So score whether the agent picked up the history and took ownership of the outcome.

Compliance Items Every QA Scorecard Needs

Quality assurance is also a compliance control, so the auto-fail list should reflect your regulators. Payment card data is the most common example. The PCI Security Standards Council states that call recordings must not contain sensitive authentication data after authorization. Therefore QA teams should confirm that recording pauses or redaction worked. A captured security code should be an auto-fail.

Regulated industries add their own checks. Health plans face direct monitoring of call quality. CMS runs a call center monitoring program, described in its memo for the 2022 study. It scores the accuracy of plan information that agents give. It also tests interpreter access within eight minutes, and those results feed Star Ratings. Healthcare teams should also score identity verification before any discussion of protected health information.

Collections programs need yet another layer. Scorecards there should check required disclosures and contact-frequency rules. Our guide to Regulation F call frequency limits explains one of those rules. For healthcare, our article on HIPAA-compliant patient support services covers the verification steps worth scoring.

Where AI Fits in Call Center Quality Assurance

AI changes the coverage problem from Step 5. Speech and text analytics can review every interaction, not a small sample. They flag missed disclosures, long silences, negative sentiment, and repeat contacts automatically. Still, AI should expand QA coverage, not replace QA judgment. The table compares the four common models.

What AI Should Score and What Humans Should Judge

Automation works best when a criterion has observable evidence. That is why Step 2 matters so much. A criterion such as “verified identity before accessing the account” is easy to detect. A vague one such as “showed empathy” is not.

Real-Time QA and Coaching

Post-call QA finds what went wrong after the fact. Real-time tools can prompt agents during the interaction instead. Typical triggers are a missed verification step, rising frustration, or long silence. Reserve live prompts for clear, high-confidence signals, since constant alerts distract agents. Our article on AI QMS call center solutions explains how these tools work in practice.

However, AI does not replace calibration. Automated scores still need human review against the same standard, or they drift. AI also creates new interactions to assure. Gartner found that 87% of customers say companies using GenAI must offer a human agent. So QA should cover the handoff from bot to agent, including whether the agent picked up the context.

Metrics That Show Your QA Program Is Working

A QA program needs its own scorecard. The average QA score alone proves little, because scorecards can be easy. Instead, track three groups of measures together. Our industry guides to telecom contact center KPIs and technical support KPIs cover the wider operational metrics.

Review customer outcomes with leadership each month. Program health metrics belong with QA and training leads. Agent metrics belong in coaching conversations, where they can drive action. Similarly, a falling auto-fail rate shows compliance improving, even when overall scores move slowly.

What If QA Scores Rise but CSAT Falls?

This example is illustrative, not drawn from a client. Suppose a contact center lifts its average QA score from 84% to 93% over two quarters. In the same period, CSAT falls from 88% to 82%. Leadership asks why agents score better while customers feel worse. The table shows how to read the signals.

The wrong response is to push agents for an even higher QA score. Instead, revisit the scorecard. Shift weight from scripted steps toward resolution and effort, then pilot the change as in Step 4. A higher QA score is not the same as higher quality. Only customer outcomes can prove that.

Conclusion

The goal of call center quality assurance is not a higher QA score. The goal is a measurable improvement in customer and business outcomes. That takes a closed loop: observable standards, a piloted scorecard, risk-based sampling, and regular calibration. It also takes fair governance, coaching that changes behavior, and steady checks against CSAT and effort. Without those pieces, QA becomes a monthly report that nobody acts on.

Building that program takes evaluators, analysts, and supervisors with time to coach. If your team lacks that capacity, a partner can run it alongside your operation. Our nearshore call center services include QA, calibration, and coaching in every program. You can also review the standards we hold on our certifications page.

Frequently Asked Questions

What is call center quality assurance?

Call center quality assurance is the review of customer interactions against a defined standard. Evaluators score calls and chats on a scorecard, then use the results to coach agents. A strong program measures both agent behaviors and customer outcomes. It also checks compliance with rules such as payment card and privacy requirements.

What is the difference between quality assurance and quality monitoring?

Quality monitoring is the process of observing, recording, or analyzing interactions. Quality assurance evaluates those interactions against defined standards and uses the findings to improve performance. Modern QA combines monitoring with structured scoring, calibration, coaching, and outcome measurement.

What should a quality assurance scorecard include?

A good scorecard covers opening and verification, discovery, resolution and accuracy, customer experience, documentation, and closing. It weights resolution and accuracy most heavily. Every criterion should be observable, so two evaluators score it the same way. It also lists auto-fail items, such as disclosing account details before verification.

How many calls should QA review per agent?

There is no single right number, because sampling is a measurement strategy rather than a quota. Combine random samples with targeted and risk-based samples. Targeted samples include escalations, repeat contacts, and low survey scores. Risk-based samples cover payments, disclosures, cancellations, and other costly failures.

Should call center QA review every call?

Not through manual review. Manual QA remains best for judgment-heavy evaluations and coaching. Automated QA can expand coverage across very large interaction volumes. The right model depends on risk, volume, technology, and what the measurement is for.

Can QA scores be used for agent performance management?

They can, but only with safeguards. Scores need validated criteria, regular calibration, representative samples, and a formal appeal process. Small samples work well for coaching. However, they can be inappropriate for high-stakes decisions about one agent.

What is QA calibration in a call center?

Calibration is a regular session where several evaluators score the same interactions independently. They then compare results and agree on the correct score for each criterion. Evaluators outside an agreed tolerance get retraining. A golden library of scored example interactions makes these sessions faster and more consistent.

How does AI change call center QA?

AI can review every interaction instead of a small sample, and it flags compliance misses and negative sentiment automatically. Human evaluators then focus on interactions that need judgment. However, automated scores still need calibration against human review. QA should also cover handoffs from AI tools to human agents.

The post Call Center Quality Assurance: How to Build a QA Program That Moves CSAT appeared first on SkyCom Call Center.

This guide maps telehealth CPT codes for 2026. It covers the codes, the place of service rules, the modifiers, and the Medicare policies behind them. Every rule links to its source, because payer policies differ and change. Treat it as orientation, and confirm specifics against the current CPT code set and each payer’s policy.

How Telehealth CPT Codes Work in 2026

A telehealth claim rests on three parts. The CPT or HCPCS code says what service the clinician delivered. The place of service code says where the patient was. The modifier says how the visit happened, such as video or audio only. A claim pays cleanly only when all three agree with the payer’s rules.

The rules for telehealth CPT codes are not uniform. The American Medical Association maintains the CPT code set, but each payer decides which codes it accepts. Medicare publishes its own list of telehealth services and updates it every January. Commercial plans and Medicaid programs then set separate policies. Therefore, the same visit can need different coding for different payers.

That variation is why support work matters before the claim exists. Scheduling teams capture the patient’s location, the visit type, and consent at booking. Our telehealth support services handle that front-end work, so the billing team starts with clean data.

The Telehealth CPT Codes Billing Teams Use Most

Most telehealth volume falls into a handful of code families. The table below groups them and notes how Medicare treats each one. Always check the current descriptor before you bill, because time thresholds differ by code.

Medicare’s list covers far more than these families. It held more than 250 codes at the start of 2025, according to HHS. For 2026, CMS added services including multiple-family group psychotherapy and group behavioral counseling for obesity. So check the list itself before you assume a service qualifies.

Why Payers Split on CPT Codes 98000-98015

The AMA introduced 17 telemedicine codes in the 2025 code set. According to the AMA, codes 98000 to 98007 cover audio-video visits, and 98008 to 98015 cover audio-only visits. Clinicians select them by medical decision-making or total time. However, Medicare did not adopt them.

Medicare contractor guidance is explicit on this point. Noridian states that Medicare does not recognize codes 98000 to 98015. Under the 2025 fee schedule, they carry status indicator I, meaning invalid. Instead, practices bill Medicare the standard office visit codes with a telehealth place of service. We found no CMS notice reversing that for 2026, but confirm it in the current fee schedule. Other payers vary, so check each plan’s policy before you choose a code family.

Telephone Codes 99441-99443 No Longer Exist

The telephone visit codes are gone. The AMA deleted codes 99441 to 99443 when it created the new telemedicine set. As a result, claims that still carry them will be rejected. For Medicare, an audio-only visit now uses the office visit code with modifier 93. For other payers, the audio-only codes 98008 to 98015 may apply. Update charge masters and templates, because old favorites lists often keep deleted codes alive.

Place of Service Codes for Telehealth

Place of service causes frequent confusion. Two codes apply to telehealth, and they pay differently under Medicare. The difference depends on where the patient sits during the visit, not where the clinician sits.

CMS defines the home as a private residence, not a hospital or other facility. Since January 1, 2024, Medicare has paid home telehealth claims at the non-facility rate. Its claims manual lists POS 02 among facility-rate settings and POS 10 among non-facility settings. Consequently, the wrong code changes the payment, not just the paperwork. Capture the patient’s location at scheduling, and confirm it at the start of the visit.

Telehealth Modifiers and When to Use Each

Modifiers tell the payer how the service reached the patient. Medicare uses fewer of them than many billing teams expect. Other payers set their own requirements, so keep a payer-specific grid.

Audio-only deserves extra care. Since January 1, 2025, Medicare allows audio-only telehealth when the patient is at home and cannot use video. The same applies when the patient does not consent to video. The clinician must still be capable of video. According to HHS, practices then append modifier 93, or FQ for rural health clinics and health centers. Document the reason for audio-only in the record, because the modifier attests to it.

Modifier 95 causes a lot of rework. CMS guidance limits its Medicare use to outpatient therapy by hospital-employed therapists. Other payers set their own rules for it, and some require it on video visits. So never copy one payer’s modifier rule to another.

Medicare Telehealth Rules That Affect Coding in 2026

Coding follows coverage, so the policy calendar matters. Congress extended the main Medicare telehealth flexibilities again in early 2026. According to HHS, patients can receive non-behavioral telehealth at home through December 31, 2027. Geographic limits on the originating site stay lifted until the same date. Audio-only delivery for those services also continues until then.

Behavioral health works differently. Home access, audio-only delivery, and the removal of geographic limits are permanent for behavioral and mental health telehealth. The in-person visit requirement remains on hold through December 31, 2027. CMS explains in its telehealth FAQ that the requirement applies after that date. Rural health clinics and health centers can also keep billing non-behavioral telehealth with HCPCS code G2025 through 2027.

The 2026 fee schedule changed several details too. CMS removed frequency limits for subsequent inpatient visits, subsequent nursing facility visits, and critical care consultations. It also made virtual direct supervision permanent, through real-time audio and video. For the originating site facility fee, code Q3014, Medicare pays 80% of the lesser of the charge or $31.85. Unless Congress acts again, the older restrictions return on January 1, 2028.

Remote Patient Monitoring CPT Codes

Remote monitoring sits beside telehealth, but it follows separate rules. The codes pay for device supply and for management time, not for a visit. For 2026, the AMA added codes for shorter monitoring periods, covering 2 to 15 days of data. It also added management codes that require only 10 minutes a month.

CMS sets guardrails that billing teams should build into edits. Its remote monitoring guidance ties each supply code to a day count. Data must cover 2 to 15 days, or 16 or more. Only one practitioner can bill remote monitoring for a patient in a 30-day period. In addition, you cannot bill remote physiologic monitoring and remote therapeutic monitoring together. Track the day count automatically, because manual counts fail at scale.

Where Telehealth Claims Get Denied

Many telehealth denials trace back to a mismatch, not a clinical dispute. Federal data shows the wider pattern. In 2024, administrative reasons drove 25% of in-network claim denials on HealthCare.gov plans. Telehealth adds more ways for the administrative details to go wrong.

Five errors are worth building edits for. First, the place of service does not match the patient’s actual location. Second, a team bills codes 98000 to 98015 to a payer that rejects them. Third, an audio-only visit goes out without the audio-only modifier. Fourth, the service is not on the payer’s telehealth list. Finally, a remote monitoring claim misses its day threshold.

Upstream steps prevent many of these. Accurate insurance verification confirms telehealth benefits before the visit. Timely prior authorization stops avoidable denials for services that need approval. Our article on the CMS prior authorization rule explains the new decision deadlines. Then our claims processing team applies payer-specific edits before submission.

A Pre-Submission Checklist for Telehealth Claims

A short checklist catches many errors before they cost a resubmission. Run it for every telehealth claim until you automate the edits. The table below pairs each check with the reason it matters.

Assign one owner for the payer grid, and review it each January. That is when Medicare updates its telehealth list and fee schedule. Mid-year changes also happen, as the 2026 extension showed.

Conclusion

Getting telehealth CPT codes right in 2026 comes down to agreement between three fields. The code, the place of service, and the modifier must match each other and the payer’s policy. Practices still bill Medicare with standard office visit codes. Medicare pays by patient location and uses modifier 93 for audio-only visits. Meanwhile, its main flexibilities now run through December 31, 2027.

So build the rules into your workflow, not your memory. Capture location and modality at scheduling, keep a payer grid, and automate the edits. If your team lacks the capacity, our revenue cycle management and appointment scheduling teams can carry both ends. This guide is informational, so confirm each rule against current payer policy before you bill.

Frequently Asked Questions

What CPT codes are used for telehealth?

The most common telehealth CPT codes are office and outpatient codes 99202 to 99215. The CPT code set also includes telemedicine codes 98000 to 98015 and the brief check-in code 98016. Online digital visits use 99421 to 99423. However, payers accept different code families. Therefore, check the payer’s telehealth policy before you choose.

Does Medicare pay for CPT codes 98000-98015?

No, based on the contractor guidance available. Medicare does not recognize codes 98000 to 98015, and its 2025 fee schedule lists them as invalid. Instead, bill Medicare the standard office visit code with place of service 02 or 10. Then confirm the status in the current fee schedule, because CMS can change it.

What is the difference between POS 02 and POS 10?

POS 10 means the patient received telehealth at home. POS 02 means the patient was somewhere other than home, such as a clinic. Medicare pays POS 10 claims at the non-facility rate and POS 02 claims at the facility rate. So the patient’s location directly affects payment.

Which modifier is used for audio-only telehealth?

Modifier 93 identifies a synchronous audio-only service. HHS guidance pairs it with Medicare audio-only telehealth claims. Rural health clinics and federally qualified health centers use Medicare modifier FQ instead. In each case, document why the visit was audio-only.

Are Medicare telehealth flexibilities still in place in 2026?

Yes. According to HHS, the main flexibilities for non-behavioral telehealth run through December 31, 2027. They include home as an originating site, no geographic limits, and audio-only delivery. Several behavioral health flexibilities are permanent. Unless Congress extends them again, the older limits return on January 1, 2028.

The post Telehealth CPT Codes and Modifiers: A Billing Team’s Guide appeared first on SkyCom Call Center.

That split confuses almost everyone at first. The TPA’s logo is on the ID card, and the employer’s money is in the account. This guide sets out the roles and responsibilities a TPA takes on and the ones it cannot take. It also covers the gaps between them. It is an operational explainer, not legal advice.

What Does a TPA Do? The Short Answer

In a self-funded health plan, the employer pays medical claims from its own funds. It does not buy an insurance policy. That model now dominates. According to KFF’s 2025 Employer Health Benefits Survey, 67 percent of covered workers are in self-funded plans. The share reaches 80 percent at larger firms and 27 percent at firms with 10 to 199 workers. Few of those employers want to build a claims department, so they hire an administrator.

State insurance law gives the role a formal shape. Under the NAIC model act, a TPA “adjusts or settles claims” and collects charges or premiums for a payor. The model also requires a license and a written agreement with that payor. Actual rules vary by state, because the NAIC text is a template for regulators. In plain terms, a TPA is a licensed contractor that handles the plan’s paperwork, payments, and phone calls.

The scale is large. A Department of Labor report based on 2022 filings counted 22,436 large self-insured plans with 38.3 million participants. Another 4,539 mixed-funded plans covered 30.1 million more. Those figures leave out most small plans, which do not have to file.

TPA Roles and Responsibilities in a Self-Funded Health Plan

TPA services fall into six groups. The table shows each one beside the part that stays with the employer. Plan documents call the employer the plan sponsor. The right-hand column matters as much as the middle one.

TPA Claims Administration: The Core of the Job

Claims are the center of what a TPA does. Each claim arrives, gets priced against the network contract, and meets the plan’s rules. The administrator then pays it from the employer’s account or issues a denial under the plan’s terms. The clock is federal. Under the DOL claims procedure rule, a group health plan must decide urgent care claims within 72 hours. Pre-service claims get 15 days, and post-service claims get 30. For those two categories, the rule allows one extension of up to 15 days. It applies only to “matters beyond the control of the plan,” and the claimant must receive notice. Urgent care claims have no such extension. Members then have at least 180 days to appeal a denial. Our overview of claims processing outsourcing covers the workflow in more depth.

Eligibility, Enrollment, and Member Services

Eligibility work looks dull and decides everything downstream. The administrator loads enrollment files, applies changes, and issues ID cards. A wrong termination date becomes a wrongly paid claim, so eligibility verification has to be exact. Meanwhile, the service desk answers members and providers by phone. Because that work involves health data, HIPAA applies. HHS says a TPA is generally a business associate of the group health plan.

Vendor Coordination and Reporting

A self-funded plan is a bundle of contracts, and the administrator sits in the middle. It connects the provider network, the pharmacy benefit manager, and the stop-loss carrier. Stop-loss insurance reimburses the employer when claims pass a set threshold, and the administrator usually files those claims. Reporting completes the picture. For example, plans must attest each year, by December 31, that their contracts contain no gag clauses. The plan owns that duty, but it leans on the administrator’s data to meet it.

What a TPA Does Not Do: The Fiduciary Line

Hiring a TPA moves the work, not the responsibility. Under ERISA, the federal law covering most private employer plans, the employer usually remains a fiduciary. DOL’s guide for group health plan fiduciaries explains the test. Fiduciary status “is based on the functions performed for the plan, not just a person’s title.” Anyone who uses discretion in running the plan is a fiduciary to that extent.

The same guide says a TPA “who performs solely ministerial tasks is not a fiduciary.” A DOL interpretive bulletin lists such tasks, including applying eligibility rules, preparing employee communications, and processing claims. However, the guide adds a warning. That status “may change if they exercise discretion in deciding on a participant’s eligibility for benefits.” An administrator that decides appeals with real discretion may therefore be a fiduciary for those decisions.

The Employer’s Duty to Monitor the TPA

Selection is a fiduciary act, and so is oversight. Employers must “monitor the service provider periodically,” the DOL guide says, and confirm that administration stays prudent. In practice, that means reviewing performance, reading the reports, checking the fees, and following up on participant complaints. PwC’s 2027 cost outlook makes the same point for large self-funded employers. It recommends linking “vendor oversight, carrier performance, and benefit strategy directly to claims experience.” An administrator should therefore expect questions, and a good one welcomes them.

TPA, ASO, or Insurance Carrier: Who Carries the Risk?

Three arrangements get confused, and the difference is who holds the risk. In a fully insured plan, the carrier generally bears the claims risk under the policy. In a self-funded plan, the employer generally funds the claims and hires an administrator. That administrator can be an independent TPA or an insurer selling administrative services only, known as an ASO. The table shows typical arrangements, and individual contracts differ.

Smaller employers often choose a middle path. KFF found that 37 percent of covered workers at firms with 10 to 199 workers are in level-funded plans. Those pair a small self-funded layer with stop-loss coverage. Either way, the administrator’s job looks much the same. That is why the question applies well beyond pure self-funding.

Where the Gaps Appear in TPA Administration

None of this describes a failing. The gaps in TPA administration are structural, and most TPAs would name them without prompting. They come from four fixed points that meet a workload nobody can fix in place.

A Fixed Fee Meets Variable Work

A TPA typically earns a fixed administrative fee per employee per month. Claim volume, call volume, and appeals do not stay fixed. Meanwhile, costs keep climbing. PwC projects a group medical cost trend of 9 percent for 2027. Higher costs tend to bring more scrutiny from plan sponsors and more questions from members. As a result, every extra manual touch comes straight out of the margin.

The January 1 Surge

Many plan years start on January 1. Enrollment files, new ID cards, and benefit questions therefore arrive together. New client implementations land in the same weeks. It may be the only business whose busiest day is also a public holiday. Staffing for that peak all year is expensive, and staffing for the average leaves January short.

Language and After-Hours Coverage

Members do not all call in English, or at noon. The federal claims and appeals rule requires “oral language services (such as a telephone customer assistance hotline).” The duty applies in counties where 10 percent or more of residents are literate only in one non-English language. For many plans, that language is Spanish. Consequently, a bilingual service desk is a compliance matter as well as a courtesy.

Federal Deadlines With Little Room to Flex

The federal claim deadlines do not stretch because a team is short-staffed. An urgent care claim still needs a decision within 72 hours. An appeal still needs a full and fair review. The one permitted extension is narrow, as noted above. Therefore, backlogs carry legal weight, not just service cost. A fixed fee does not know it is January, and neither does the regulation.

How TPAs Can Identify Administrative Capacity Gaps

A temporary spike and a lasting capacity problem look alike in the first week. TPAs can tell them apart by tracking four signals together, not one at a time. The table sets them out.

Read the four together. For example, rising volume with a flat backlog is a busy month. Rising volume with a growing backlog and more repeat contacts is a capacity gap. That second pattern is the point to add capacity for routine work, before deadlines come under pressure.

How TPAs Close the Gaps With Support Partners

The fiduciary rules point to the answer. The DOL bulletin on ministerial work gives the test. The bulletin covers work “within a framework of policies, interpretations, rules, practices and procedures made by other persons.” That is also a fair description of what an administrator can hand to a partner. The administrator writes the rules and keeps the decisions. The partner carries out the defined steps.

In practice, the work that moves is high in volume and low in discretion. It includes member and provider calls, eligibility data entry, claim status research, and document indexing. It also includes outbound follow-up for missing information. By contrast, adjudication authority, appeal decisions, and plan interpretation stay in-house. A partner that offers to take those is offering the wrong thing. That division is the basis of a sound TPA outsourcing arrangement.

What to Ask a TPA Support Partner

Start with data protection. A partner that handles member data for an administrator takes on business associate obligations as well. Ask for the agreement and the controls behind it. Our guide to HIPAA-compliant support covers the safeguards. Next, ask how the partner logs each interaction, because your plan sponsors will audit you. Finally, check the scope behind each credential on its certifications page, not just the logo.

Where SkyCom Fits in TPA Administration

SkyCom works on the ministerial side of that line. Our nearshore teams provide TPA support services in English and Spanish. That covers claims support and research, eligibility and enrollment processing, and member and employer services. The same teams support claims processing and back-office work for other healthcare clients. We follow the client’s procedures, and the client keeps every determination.

Conclusion: What a TPA Does, and Where It Needs Support

So what does a TPA do? It turns an employer’s promise to pay for care into a working operation. It pays claims, keeps eligibility straight, answers the phones, and reports back. Also, it does not carry the insurance risk, and it does not remove the employer’s fiduciary duty.

The gaps sit where fixed things meet moving ones. Fees, deadlines, and plan-year dates hold still while volume, cost, and language needs shift. TPAs that separate judgment from routine work handle that best. They keep the judgment and find reliable capacity for the rest.

Frequently Asked Questions: What Does a TPA Do?

What does a TPA do in health insurance?

A TPA administers a self-funded health plan for an employer. It processes claims, maintains eligibility records, serves members and providers, and produces reports. The employer pays the claims and remains responsible for the plan.

Is a TPA an insurance company?

No. A TPA does not take on insurance risk or collect premiums for its own account. It provides administration for a fee. Some insurers also sell administration alone, which the market calls administrative services only.

Who pays the claims in a self-funded health plan?

The employer does, from its own funds or a plan trust. The administrator issues the payments from that account. Many employers also buy stop-loss insurance, which reimburses them when claims pass a set threshold.

Is a TPA a fiduciary under ERISA?

It depends on what the TPA does. DOL guidance says a TPA performing solely ministerial tasks is not a fiduciary. That can change if it uses discretion, for example, when deciding benefit eligibility or appeals.

What is the difference between a TPA and an ASO?

Both administer self-funded plans for a fee. An ASO arrangement comes from an insurance carrier and usually uses that carrier’s network. An independent TPA is typically not tied to one carrier, and its network access depends on the arrangement.

Are TPAs licensed?

In many states, yes. The NAIC model for state regulators requires a TPA license and a written agreement with the payor. The details differ by state, so check the rules where the plan’s members live.

Does HIPAA apply to a TPA?

Yes. HHS says a TPA generally acts as a business associate of the group health plan. It must therefore protect member health information under a business associate agreement.

How is a TPA paid?

Most TPAs charge a fixed administrative fee for each covered employee each month. Some services carry separate charges. Because the main fee is fixed, extra manual work reduces the margin.

The post What Does a TPA Do for a Self-Funded Health Plan? Roles, Responsibilities, and Gaps appeared first on SkyCom Call Center.

This guide answers three practical questions. First, what counts as a complaint under the current medical device complaint handling process? Second, who has to log it, and how fast? Third, what happens once the record exists? Along the way, it separates complaint intake from medical device reporting, which runs on its own clock. It also draws an honest line around what a support team should never decide.

What Medical Device Complaint Handling Means Under the QMSR

For nearly three decades, FDA complaint handling requirements lived in one section of the Quality System Regulation. The QMSR final rule replaced that structure. Instead of writing its own text, FDA incorporated ISO 13485:2016 by reference. As a result, the current Part 820 now holds only six sections. The regulation did not get lighter. It moved house and left a forwarding address in Geneva.

QMSR complaint handling now sits in two places. Clause 8.2.2 of ISO 13485 sets the base process. In addition, 21 CFR 820.35(a) adds FDA’s own record requirements on top. Therefore, a procedure that still cites 820.198 points to a section that is gone. Auditors notice that kind of thing, usually on the first page.

Certification does not replace inspection, either. FDA’s QMSR FAQ is blunt on the point. “A certificate of conformance to ISO 13485 will not exempt a manufacturer from an FDA inspection.” Furthermore, the agency withdrew its old inspection technique on the same date. Investigators can now review management review, quality audit, and supplier audit reports. Consequently, your audit of an outsourced intake vendor can land on an investigator’s desk.

What Counts as a Complaint in Medical Device Complaint Handling

A complaint rarely introduces itself as one. Nobody calls to report a possible failure to meet specifications. They say the pump beeps at night for no reason. ISO 13485 defines the term broadly. In plain terms, it covers any written, electronic, or oral communication alleging a deficiency in a released device. That reaches quality, durability, reliability, usability, safety, and performance.

FDA’s record rule then names three targets. Under 820.35(a), manufacturers keep records for complaints “involving the possible failure of a device, labeling, or packaging.” Note the word possible. The caller does not need to be right, and the device does not need to be broken. Nothing needs proving at intake. Moreover, labeling and packaging count alongside the hardware. A confusing instruction sheet can therefore start a complaint record.

Complaint Sources and User Error

The channel does not matter. FDA’s MDR guidance for manufacturers lists telephone calls, faxes, letters, and email as complaint sources. That 2013 guidance predates the QMSR and still cites the old section number. However, its reporting logic rests on Part 803, which did not change. Chat and social messages arrive the same way today, and the ISO definition covers electronic communication.

User error deserves special attention, because intake teams dismiss it most easily. Part 803 lists user error among the ways a device may have “caused or contributed” to harm. In other words, a patient who admits a mistake has not closed the matter. The manufacturer still has to evaluate what happened. An agent who says “that one is on you” and ends the call has lost a record.

Complaint, Inquiry, or Service Issue: Worked Examples

The table below shows how common contacts tend to sort. It is illustrative only, and your written complaint handling procedure governs. When in doubt, the safer habit is simple: capture everything and let quality decide.

Notice how many rows start as ordinary support calls. For that reason, complaint detection belongs in technical support workflows, not only in quality. Similarly, durable medical equipment suppliers hear device problems daily during resupply and billing calls. Each of those conversations is a potential intake point.

Medical Device Complaint Handling and MDR Reporting Run on Different Clocks

Complaint handling and medical device reporting are related, but they are not the same duty. Every complaint needs evaluation, yet only some become reports to FDA. Under 21 CFR Part 803, a reportable event has two main forms. First, the device may have caused or contributed to a death or serious injury. Second, it malfunctioned in a way likely to cause such harm if it recurred.

The definition of malfunction is wider than most support teams assume. Part 803 calls it “the failure of a device to meet its performance specifications or otherwise perform as intended.” It then adds a sentence that marketing teams should read twice. “Performance specifications include all claims made in the labeling for the device.” Consequently, a device that works but falls short of a labeled claim can qualify.

The volume shows how much traffic the system carries. According to FDA, the agency receives “over two million medical device reports” each year. Those cover suspected deaths, serious injuries, and malfunctions. However, FDA also cautions that a report is not proof of fault. Submitting one “is not evidence that the device caused or contributed” to the outcome. Reporting is a surveillance duty, not a confession.

The MDR Reporting Clock Starts at Awareness

The MDR reporting deadlines are short. Manufacturers generally have 30 calendar days to report after they become aware of an event. That window shrinks to five work days when remedial action must prevent an unreasonable risk of substantial harm. The key phrase is “become aware.” For the 30-day report, the clock starts when “any of your employees becomes aware of a reportable event.” It does not wait for the quality team to open a ticket.

What about people who answer on your behalf? The regulation speaks of employees and stays silent on vendors. FDA’s guidance fills part of that gap. It says MDR procedures “should identify the roles and responsibilities of contractors or consultants” who process complaints. Therefore, the cautious approach is to treat the vendor’s first contact as your awareness date. Write that into the contract before anyone tests the alternative.

Who Logs It: Roles in the Medical Device Complaint Handling Process

The practical model has two halves. Anyone who receives possible complaint information forwards it. Designated quality and regulatory staff then evaluate it. FDA’s guidance states the first half plainly. Manufacturers should tell all employees to “immediately forward adverse event information to the appropriate person.” The guidance names marketing, sales, engineering, legal, installation, and service staff. In short, nobody may sit on a complaint because it arrived in the wrong inbox.

The second half belongs to the quality and regulatory roles your procedure names. They evaluate each complaint, decide whether to investigate, and determine reportability. Those calls need medical, technical, and regulatory judgment. A frontline agent should make none of them. Instead, the agent’s job is narrower and still essential: recognize, capture, and route.

Three Roles in Medical Device Complaint Handling

In practice, a workable complaint handling procedure splits into three roles. Intake recognizes a possible complaint and records the facts in the caller’s own words. Next, a complaint unit reviews the record, requests missing details, and classifies it. Finally, regulatory affairs makes the MDR decision and files where the rule demands it. Trouble starts when the first role quietly performs the second. An agent who decides something is “not a real complaint” has made a regulatory call without the authority.

Complaint intake rarely lives in one place. A medtech support operation typically runs phones, chat, email, and field service at once. Meanwhile, distributors and clinical educators hear problems that never reach a queue. Each channel needs the same trigger questions and the same route to quality. Otherwise, your complaint data reflects your org chart more than your product.

The split between intake and decision leads to the most practical point in this guide. You can outsource intake, but you cannot outsource accountability. A partner can recognize, capture, and route on your behalf. However, the evaluation and the reporting decision still answer to your quality system. The final section covers where that line sits.

What Happens Next: From Complaint Intake to MDR Decision

Logging is the start of the process, not the end. Once a record exists, four more questions follow, and each has a different owner. The table below shows the usual stages. Owners are typical, so your complaint handling procedure should name the actual roles.

One point in that table deserves care. The stages read as a neat sequence, but two of them overlap. The 30-day MDR clock runs from awareness, not from the end of the investigation. Therefore, regulatory affairs often has to assess reportability while the investigation is still open. Waiting for a root cause before asking the reporting question is how deadlines slip.

The Complaint Handling Record: Seven Fields in 21 CFR 820.35(a)

This is where the regulation gets usefully specific. Section 820.35(a) lists the information a manufacturer must record. The list applies to complaints that require an FDA report and to complaints the manufacturer investigates. Intake cannot know in advance which complaints those will be. Therefore, capture the fields every time. The table maps each requirement to what an agent actually does.

Where Complaint Records Fall Short

Two fields cause the most damage when they go missing. The first is the received date, because it anchors the reporting clock. The second is device identification, because investigation stalls without a lot or serial number. Moreover, a caller who has thrown the box away rarely calls back with the number. Ask for it while they are still on the line.

The same section also permits a shortcut, with a condition. If a similar complaint already received an investigation, another one is not necessary. However, the manufacturer must document the justification for skipping it. That decision belongs to quality, and it depends on consistent intake wording. Sloppy descriptions make similar complaints look different, and different ones look the same.

Where the Complaint Handling Process Breaks Down

FDA’s enforcement data shows where complaint systems have struggled. In its calendar 2016 quality system data, FDA recorded 3,027 inspection observations against Part 820. The corrective and preventive action subsystem drew 1,017 of them, or 34 percent. Complaint procedures alone accounted for 286 observations. Those figures are a decade old and cite the former section number. Still, they show where investigators have historically looked first.

A well-known example shows the cost of logging without analysis. In a May 2022 proposed order, FDA described how Philips Respironics handled foam degradation complaints. Philips’ own review confirmed 1,254 such complaints between 2014 and April 2021. Of those, 110 arrived between 2014 and 2017. FDA stated that Philips “failed to adequately evaluate this data and incorporate it into its CAPA system.” Those statements come from FDA’s proposed order, not from a court ruling. Even so, the lesson holds: the complaints sat in the system while the pattern went unread.

FDA’s own QMSR overview deck borrows a line it attributes to John Ruskin. “Quality is never an accident; it is always the result of intelligent effort.” Complaint handling is the least glamorous proof of that idea. It fails quietly, in small omissions, long before anyone issues a recall.

Four Intake Failures in Medical Device Complaint Handling

Four breakdowns tend to appear at the intake stage. First, agents resolve the issue and never record it, because a fixed problem feels finished. Second, no-fault-found returns close without a complaint review. A no-fault-found result describes the test bench, not necessarily the patient’s experience. Third, after-hours calls go to voicemail, and the received date slips by a weekend. Fourth, non-English callers get a shortened version of the script.

The language point deserves a sentence more. A complaint made in Spanish is still a complaint, and the clock does not pause for translation. Therefore, Spanish-language support needs the same trigger questions and the same record fields. Relaying the call through an interpreter adds a step where details vanish.

What an Outsourced Team Should and Should Not Do in Complaint Handling

Many manufacturers route first contact through an outside partner. That can work well, within limits. An inbound contact center can answer every channel, apply trigger questions, and capture the seven fields. It can also timestamp first contact and escalate the same day. Those are intake tasks, and trained agents perform them reliably at scale.

The limits matter more than the capabilities. An outsourced team should not decide whether a contact is a complaint. Nor should it judge reportability, assess causation, or close an investigation. In addition, it should never offer clinical advice or speculate about what went wrong. Those decisions stay with the manufacturer’s quality and regulatory functions. As noted earlier, the task can move, but the accountability stays on your letterhead.

What the Vendor Contract Should Cover

Because accountability stays put, the contract has to carry the detail. Define the escalation trigger, the route, and the maximum time from first contact to handoff. Specify who owns the recordings and how long the vendor keeps them. For reference, Part 803 requires MDR event files for two years or the device’s expected life, whichever is greater. In addition, plan to audit the vendor, since FDA can now review supplier audit reports. Ask for the scope behind each credential on a vendor’s certifications page, not just the logo.

Privacy needs a paragraph of its own. Complaint calls frequently include health details, because patients explain why they use the device. Whether HIPAA applies depends on the manufacturer’s role and the data flow. Either way, the record deserves the safeguards covered in our guide to HIPAA-compliant patient support. Collect what the complaint record needs and nothing more.

Where SkyCom Fits in Medical Device Complaint Handling

SkyCom works on the intake side of the line this guide draws. Our nearshore teams provide medical device support and complaint intake across phone, chat, and email, in English and Spanish. That work sits within our wider healthcare practice. We capture, timestamp, and route, and your quality team decides. That division is deliberate. It is also the only version we recommend.

Conclusion: Medical Device Complaint Handling Starts With the First Hello

Medical device complaint handling has a new legal home, but its logic has not changed. A complaint is any allegation that a released device, its labeling, or its packaging fell short. It counts whether it arrives by phone, chat, or a technician’s note. Furthermore, it counts even when the caller blames themselves.

The answer to who logs it is equally plain. Anyone who hears it forwards it, and intake records it the same way every time. Quality and regulatory affairs then decide what it means. Get the first step right, and every later step has something to work with. Get it wrong, and the best CAPA system in the industry is analyzing an empty file.

Frequently Asked Questions

What is medical device complaint handling?

It is the process a manufacturer uses to receive, record, evaluate, and investigate allegations about a released device. Under the QMSR, the process follows ISO 13485 Clause 8.2.2 and 21 CFR 820.35. It also feeds the separate decision on whether to report an event to FDA.

Is 21 CFR 820.198 still in effect?

No. The QMSR took effect on February 2, 2026 and restructured Part 820. FDA complaint handling requirements now come from ISO 13485 Clause 8.2.2, with record rules in 21 CFR 820.35. Procedures that still cite 820.198 need an update.

Does every complaint have to go to FDA?

No. Every complaint needs evaluation, but only reportable events go to FDA under Part 803. Those involve a death, a serious injury, or a malfunction likely to cause either if it recurred. The manufacturer’s regulatory function makes that determination.

How quickly must a manufacturer report a device event?

Manufacturers generally have 30 calendar days after becoming aware of a reportable event. The deadline drops to five work days when remedial action must prevent an unreasonable risk of substantial harm. Awareness can begin with any employee, so the first contact date matters.

Does user error count as a complaint?

It can. Part 803 includes user error among the ways a device may have caused or contributed to an event. FDA’s guidance says an event caused solely by user error, with no death or serious injury, needs no MDR. However, that conclusion belongs to the manufacturer after evaluation, not to the agent on the call.

Can a call center handle medical device complaints?

A call center can handle complaint intake. That means recognizing a possible complaint, capturing the required fields, and escalating promptly. It should not decide whether a contact is a complaint, whether it is reportable, or what caused it. The manufacturer stays accountable for the whole process, so the contract should define roles precisely.

What information must a complaint record include?

Section 820.35(a) lists seven items for investigated and reportable complaints. They cover the device name, received date, device identifiers, complainant contact details, complaint details, corrective action, and any reply. Capturing all seven at intake avoids chasing callers later.

How long must MDR records stay on file?

Part 803 requires manufacturers to keep MDR event files for two years from the event date. If the device’s expected life is longer, that longer period applies. Check your own procedure as well, because other record rules may extend it.

The post Medical Device Complaint Handling: What Counts, Who Logs It, and What Happens Next appeared first on SkyCom Call Center.

The difference matters in both directions. Seven calls can still break the law, and a collector can sometimes defend an eighth. This guide explains what the rule presumes, how to count, and what falls outside it. It also covers who the rule reaches and where state rules bite harder. One caution first: this is an operational explainer, not legal advice.

What Regulation F Call Frequency Limits Actually Say

Regulation F implements the Fair Debt Collection Practices Act. CFPB issued the final rule in late 2020, and it took effect on November 30, 2021. The core prohibition sits in 12 CFR 1006.14(b). A debt collector must not place calls “repeatedly or continuously with intent to annoy, abuse, or harass.” Intent is hard to prove, so the rule adds a numeric shortcut.

That shortcut is a pair of presumptions. Below a stated call frequency, the rule presumes compliance. Above it, the rule presumes a violation. When CFPB announced the rule, Director Kathleen Kraninger said the aim was “clear rules of the road.” The road came with a speed limit sign that reads “probably fine under seven.”

The Two Prongs of the 7-in-7 Rule

The presumption of compliance has two conditions, and a collector must meet both. First, the collector calls a person about a debt no “more than seven times within seven consecutive days.” Second, no call follows “within a period of seven consecutive days after having had a telephone conversation.” Break either condition, and the presumption flips to a violation. In other words, the popular name describes only half the rule.

CFPB is direct about what this structure is not. Its debt collection rule FAQs say the rule does not impose a specific “limit” or “cap” on calls. Instead, the numbers shift the burden of argument. Under seven, a regulator or plaintiff must show why the pattern was still harassing. Over seven, the collector must show why it was not.

How to Count Calls Under Regulation F Call Frequency Limits

Counting sounds simple until a dialer has to do it. The rule turns on three questions. What counts as a call? When do the seven days run? Whose count is it? CFPB’s official interpretation answers each one. The answers are more specific than many teams expect.

What Counts Toward the Seven Calls

A call counts when it connects to the dialed number. That includes a phone that rings unanswered and a call that reaches voicemail. It counts even if the agent cannot leave a message. Ringless voicemail counts too, because the commentary treats it as placing a telephone call. By contrast, a busy signal or an out-of-service notice does not connect, so it does not count.

Two rows deserve a second look. A limited-content message is still a voicemail, so it uses one of the seven. Meanwhile, consent has a short shelf life. It must come directly from the person, and it covers calls for no more than seven days.

When the Seven Days Start and End

The window is seven consecutive days, not a calendar week. Therefore, a dialer that resets every Monday can fail without anyone noticing. Seven calls from Friday to Sunday, then seven more from Monday to Wednesday, makes fourteen in six days. Each calendar week looks clean, yet the rolling count has doubled the threshold.

The conversation prong has its own arithmetic. The commentary says the date of the conversation is the first day of the period. Suppose an agent speaks with a consumer on a Monday. Monday is day one, and Sunday is day seven. As a result, the next presumptively compliant call falls on the following Monday.

Per Person, Per Debt, Not Per Phone Number

CFPB’s FAQs say the presumptions apply “per person, per debt, regardless of how many telephone numbers” a person has. So four calls to a mobile and four to a landline make eight, not two sets of four. The count also follows people other than the debtor. The presumptions apply to “all persons, not just to the consumer.”

The per-debt rule cuts the other way as well. A consumer with three separate debts at one agency has three separate counts. The arithmetic therefore allows up to 21 calls in a week. Whether that pattern survives scrutiny is a different question, covered below. Student loans are the exception. Loans that shared one account number when the collector obtained them count as one debt.

What Regulation F Call Frequency Limits Do Not Cover

The 7-in-7 presumptions apply to telephone calls only. CFPB’s FAQs state that the provision “does not apply to other media types,” including text messages and email. However, that is not a free pass for digital channels. The general ban on harassing conduct in 1006.14(a) still applies to every medium. In addition, a person can ask a collector to stop using a particular medium, and the collector must comply.

Inbound calls sit outside the count as well. When a consumer calls in, CFPB says that call is “not a telephone call placed by the debt collector.” Time of day is a separate rule entirely. Regulation F presumes that calls before 8 in the morning or after 9 at night are inconvenient. A collector can therefore respect the weekly count and still call at the wrong hour.

Who Regulation F Call Frequency Limits Apply To

The rule binds debt collectors as Regulation F defines them. That means a business whose principal purpose is collecting debts, or one that regularly collects debts owed to another. The debt must also be consumer debt, arising from personal, family, or household transactions. Medical bills qualify, so patient balances placed with an agency fall inside the rule. Consequently, business-to-business collections fall outside it. So do a creditor’s own employees collecting in the creditor’s name.

That boundary is narrower than it looks. An outsourced team working first-party accounts raises a real legal question about where it sits. Our guide to first-party versus third-party collections covers the distinction. In addition, some state and city rules apply their own limits to creditors directly. Ask counsel before assuming an exemption.

Why Seven Calls Is Not a Safe Harbor Under the 7-in-7 Rule

A presumption can fall, and the commentary lists how. The first factor is the frequency and pattern of calls, including the intervals between them. CFPB’s example is blunt: “two unanswered telephone calls to the same telephone number within five minutes.” Other factors include what the person said in earlier contacts and how the collector behaved. A prior request to stop calling weighs heavily.

This is not a theoretical risk. CFPB’s 2025 FDCPA annual report describes a supervision finding on exactly this point. Collectors stayed within the stated frequencies, so they started with the presumption. However, they placed “over 100 calls to the consumer after being specifically asked to stop.” Examiners found that the conduct “overcame that presumption and had the effect of harassing the consumer.” Seven is a presumption, not a permission slip.

When a Collector Can Defend an Eighth Call

The presumption of violation can fall too. The commentary names four situations that may justify an extra call. First, the law may require the call, such as a notice about loss mitigation options. Second, the call may relate directly to active litigation. Third, the consumer may have asked for it. Finally, it may carry time-sensitive information that prevents a demonstrably negative effect on the person. Each of these needs documentation at the time, not a story assembled later.

State Rules That Go Further Than Regulation F Call Frequency Limits

Federal law sets a floor here, not a ceiling. CFPB’s FAQs confirm that the rule does not preempt a state law “that affords greater protection to consumers.” That expressly includes “more restrictive presumptions related to telephone call frequency.” Two examples show how wide the gap can run.

Massachusetts and New York City Call Frequency Limits

Massachusetts is the long-standing outlier. Its debt collection regulations bar more than two communications in each seven-day period to a debtor’s personal numbers. The count covers calls, text messages, and recorded audio messages. Moreover, the definition of creditor reaches a business collecting its own debts.

New York City is the newest. In February 2026, the Department of Consumer and Worker Protection announced rules it called the strongest in the nation. The department’s FAQ describes three communications per account in seven days, across all media except mailed letters. The rules cover original creditors as well as third-party collectors. The same FAQ lists January 1, 2027 as the effective date, after earlier dates slipped. Confirm the date before you build around it.

The practical rule is simple to state and tedious to run. Apply the strictest standard that covers each account, based on where the consumer lives. For a consumer lending portfolio spread across states, that means state-level dialer rules. A single national setting of seven will overshoot in Massachusetts by a wide margin.

What the Data Says About Call Frequency Complaints

CFPB surveyed consumers before it wrote the rule. In that survey, 37 percent of consumers contacted about a debt reported at least four contact attempts a week. Seventeen percent reported at least eight. Close to two-thirds, 63 percent, said they were contacted too often. The fieldwork ran from December 2014 to March 2015, so treat it as a pre-rule baseline.

The complaint data after the rule tells a similar story. CFPB received about 207,800 debt collection complaints in 2024, or seven percent of all complaints. Among complaints about communication tactics, 51 percent concerned frequent or repeated calls. That makes call frequency the top communication grievance three years after the rule took effect. A presumption, it turns out, does not stop the phone from ringing.

Turning Regulation F Call Frequency Limits Into Dialer Rules

Policy fails at the point where it meets the dialer. The first decision is the counting key. Count by person and debt, never by phone number or by account alone. Next, use a rolling seven-day window in place of a weekly reset. Then map dispositions carefully, because a busy signal and an unanswered ring are different events under the rule.

Consent and conversations need timestamps. A callback request should record who gave consent, when, and for which debt. The system should also release that consent after seven days or after a conversation. Similarly, a completed conversation should lock the account until day eight. Agents should not have to do that arithmetic mid-shift.

Contact Quality Beats Contact Volume

The rule rewards better targeting more than faster dialing. With only seven presumptively safe attempts, each one should land when the consumer is likely to answer. That shifts attention to right-party contact rate, which our debt collection KPIs guide covers. It also raises the value of text and email, used within the general harassment rules. For Spanish-speaking consumers, a bilingual agent on the first attempt avoids spending a second one.

Where an Outsourced Collections Team Fits

An outsourced team inherits your contact policy; it does not replace it. The creditor or agency of record sets the frequency rules, the state overlays, and the escalation path. The partner’s job is to execute them and prove it with call-level records. Therefore, ask any vendor how its dialer counts, and ask to see the audit log. Ask for the scope behind its certifications as well. SkyCom runs collections programs for US clients on that basis, with outbound teams working to each client’s written policy.

Conclusion: Regulation F Call Frequency Limits Are a Starting Point

Regulation F call frequency limits work as a burden-shifting device, not a quota. Stay within seven calls, and wait seven days after a conversation, and the rule presumes compliance. Go beyond either, and it presumes the opposite. Both presumptions can fall on the facts.

So the useful question is not how many calls the rule permits. It is whether your records could explain each call to an examiner. Count by person and debt, roll the window, respect the stricter state rules, and document everything. Then the number seven becomes what the rule makes it: a presumption, with judgment still required.

Frequently Asked Questions About Regulation F Call Frequency Limits

What is the 7-in-7 rule in Regulation F?

It is the common name for the telephone call frequency presumptions in 12 CFR 1006.14(b). A collector presumptively complies by calling a person about a debt no more than seven times in seven days. The collector must also wait seven days after a telephone conversation about that debt.

Is seven calls a week a hard limit under Regulation F?

No. CFPB says the rule does not impose a specific limit or cap. Seven is the line between a presumption of compliance and a presumption of violation. Either presumption can fall, depending on the pattern of calls and the circumstances.

Do voicemails and unanswered calls count toward the seven?

Yes. A call counts if it connects to the dialed number, which includes ringing unanswered and reaching voicemail. Ringless voicemails and limited-content messages count as well. A busy signal or an out-of-service notice does not count.

Do texts and emails count toward the 7-in-7 rule?

No. The presumptions apply to telephone calls only. However, the general prohibition on harassing conduct covers every channel. A person can also ask a collector to stop using a specific medium.

Does the rule apply per debt or per consumer?

It applies per person and per debt, regardless of how many phone numbers the person has. A consumer with several debts therefore has a separate count for each. Student loans that shared one account number when the collector obtained them count as a single debt.

When can a collector call again after speaking with the consumer?

The date of the conversation counts as day one of a seven-day period. After a conversation on a Monday, the next presumptively compliant call falls on the following Monday. Calls made with the person’s direct prior consent sit outside that count.

Do Regulation F call frequency limits apply to original creditors?

Generally not. The rule covers debt collectors as the regulation defines them. It excludes a creditor’s own employees collecting in the creditor’s name. However, some state and city rules reach creditors directly. Outsourced first-party programs should take legal advice on their status.

Can states set stricter call frequency limits?

Yes. CFPB confirms the federal rule does not preempt state laws that give consumers greater protection. Massachusetts allows no more than two communications in seven days to personal numbers. New York City’s rules describe three communications per account in seven days.

The post Regulation F Call Frequency Limits: What the 7-in-7 Rule Actually Permits appeared first on SkyCom Call Center.

Most changing call center provider projects do not fail at provider selection. They fail in the handover. Knowledge sits in the heads of agents who are about to lose the account. This guide covers what the available evidence supports and what your contract probably says that you have not read. It also covers how incumbents behave once notice lands. Finally, it sets out a call center transition plan that holds service while it runs.

Why Companies Start Changing Call Center Providers

The stated reason is usually cost. The actual reason is usually something that happened eighteen months earlier and never got fixed. Perhaps a quality problem nobody escalated, or a reporting pack that stopped answering questions. Perhaps an account manager left, and nobody replaced them properly. Cost becomes the language because cost is the line item a CFO recognizes. The decision itself is rarely about rate. Rebuilding a call center outsourcing arrangement on price alone tends to reproduce the original problem.

Service quality decline is the most common trigger, and it rarely arrives suddenly. Attrition climbs on your account, tenure drops, and quality scores drift while the monthly report still shows green. By the time a client notices, the provider has usually been understaffing the account for two or three quarters.

Scope mismatch is the second trigger. You signed for voice, then added chat, email, and social. The provider staffed all of it with the same agent profile. Growth into new channels or languages exposes capability gaps that did not exist when you signed. That is a legitimate reason to move. It is also a reason to scope the replacement more carefully than the original.

Governance failure is the quietest trigger and the most expensive. KPMG observes that governance models are under strain as delivery becomes more hybrid. Internal governance, it notes, has not kept pace. Many clients discover during a switch that they never had a working governance model, only a monthly call. Consequently, they repeat the same mistake with the next provider.

What the Data on Provider Replacement Actually Says

Here we need to be direct with you about the evidence, because most articles on this topic are not. No analyst firm, Big Four publisher, or government body reports switching or transition-failure rates for contact center outsourcing. The confident percentages elsewhere on this keyword generally trace to vendor marketing, with no followable citation.

The closest defensible evidence comes from adjacent outsourcing categories. Everest Group examined service transitions and used incumbent replacement as a proxy for transition trouble. Among 116 infrastructure outsourcing deals observed in 2013, clients replaced the incumbent in 37 percent. Among 136 application outsourcing engagements, the figure was 24 percent, according to Joiner and Lade.

Read those numbers with two caveats firmly attached. They describe IT infrastructure and application work rather than contact center delivery, and the underlying observations date from 2013. So treat them as evidence that provider replacement is common across outsourcing generally, not as a contact center benchmark. Anyone presenting them as a call center churn rate is overreaching.

What the Everest analysis does offer is a durable diagnosis of why transitions go wrong. The authors point to misaligned objectives after signature, organizational readiness gaps, unclear accountability and scope, and inadequate planning. They also note a structural tension: buyers prioritize cost and quality while providers work to recover margin. None of that has aged, because it describes incentives rather than technology.

Their sharpest line is worth keeping in view throughout a switch. “Transition is too important to be left to the providers,” the authors argue. That applies doubly when two providers share the work, and only one wants the project to succeed.

The Real Risk Sits in the Handover, Not the Hire

Buyers spend most of their switching effort on provider selection and almost none on handover design. That allocation is backwards. Selection errors surface in month four and remain fixable. Handover errors surface in week two and hit customers immediately.

Consider what actually lives inside an incumbent operation after two years. Agents hold undocumented workarounds for your order system and informal escalation paths to specific people on your team. They also carry pattern knowledge about which complaints need careful handling. Very little of that sits in the knowledge base you are contractually entitled to receive.

Call center knowledge transfer therefore has to extract tacit knowledge from people whose jobs are ending. That is an awkward human situation, and naming it works better than managing around it. Some incumbents handle it professionally. Others do the contractual minimum, which is legally fine and operationally useless.

So treat any documentation the incumbent hands over as a starting point rather than a deliverable. Plan to rebuild your knowledge base during transition, using live call observation rather than inherited articles. Teams that budget for this finish stronger than they started. The exercise exposes process debt nobody had examined in years.

Read Your Contract Before You Give Notice

This is the step most companies skip, and it reliably costs them leverage. Your current agreement contains provisions that determine how much control you have over the exit. Reading them after serving notice means negotiating from a weaker position.

Start with the termination clause and the notice period. Many BPO contracts require 90 to 180 days of notice, and some auto-renew if notice lands late. Find out whether you have termination for convenience or only termination for cause. The second obligates you to document a breach.

Then look for the exit assistance obligations, sometimes called a transition-out or disengagement clause. A strong clause obliges the incumbent to cooperate with your new provider. It also holds them to service levels through the notice period. It also names the artifacts they must deliver. A weak clause says the parties will cooperate in good faith, which means whatever the incumbent decides it means.

Data provisions deserve particular attention and usually get none. Establish who owns call recordings and transcripts, in what format they return to you, and how quickly. Then establish what the provider deletes, and when. Establish the same for your knowledge base, quality scorecards, training material, and any workflow built on the provider’s platform. Where regulated data applies, confirm that deletion certification arrives in writing. Our guide to HIPAA and PCI compliance in nearshore BPO covers questions that matter as much on exit.

Finally, check what happens to telephony numbers, IVR configuration, and any licence held in the provider’s name. Porting a toll-free number is routine but not instant. A number held in the incumbent’s name becomes a hostage you would rather not create.

What the Incumbent Does During the Notice Period

No provider-published article covers this honestly, so here is the uncomfortable version. Once an incumbent knows the account is leaving, its incentives change immediately. Your service feels that change before your reports show it.

Account attrition usually accelerates first. Agents hear the program is ending and start applying internally or externally. Your most tenured people therefore leave earliest. The provider may also reassign strong performers to accounts with a future, replacing them with newer agents. Both responses are commercially rational, and both degrade your service during exactly the period when you need stability.

Backfilling also tends to slow or stop. Hiring for a program that ends in four months is hard to justify. Vacancies sit open, and the remaining team absorbs the volume. Handle time rises, abandonment rises, and quality slips. Your dashboard may lag these effects by several weeks.

You can manage this, though only if you plan for it before serving notice. Agree on a staffing floor in writing for the notice period and tie it to the final invoices. Increase your own monitoring frequency rather than relying on provider reporting. Ask for named continuity of the supervisory layer. Supervisors carry more institutional knowledge than anybody else on the floor.

Where the relationship is cordial, consider a retention incentive for key incumbent staff during transition. Paying to keep people you are leaving feels strange. It usually costs less than the service recovery you avoid.

Three Ways to Cut Over, and When Each One Fits

Cutover design is the central decision in any call center migration. Most buyers treat it as a technical detail. It is really a cost and risk trade-off, so make it deliberately.

Often your circumstances make this choice for you. A hostile exit removes the phased options. Gradual approaches need the incumbent to keep routing and reporting working while losing revenue weekly. Therefore, assess incumbent cooperation honestly before committing to a design that depends on it.

Avoid cutting over during your seasonal peak, which sounds obvious and happens constantly. Contract end dates land where they land. Teams then talk themselves into a December migration because the paperwork says so. Extending the incumbent by one quarter costs less than failing through peak.

A Realistic Call Center Transition Timeline

Providers quote aggressive timelines because fast transitions win deals. Buyers accept them because the current situation is painful. A program of any complexity needs more time than either party initially suggests.

Expect roughly two to four weeks for discovery and scope documentation before anything else begins. The new provider needs process documentation, volume and arrival patterns, and handle times by contact type. It also needs systems access requirements and quality definitions. Gaps found here are the cheapest gaps you will ever find.

Systems access and integration usually consume four to eight weeks, and it is the step that slips most. Access provisioning in regulated environments involves your security team, background check cycles, and sometimes client-side approvals nobody remembered. Start it on day one rather than after process design.

Training and certification needs three to six weeks for most programs, longer where licensing or clinical content applies. Insist on certification against live or recorded contacts rather than classroom scores. A team that passes a quiz is not a team that can handle your angriest customer.

Then allow four to eight weeks for ramp and stabilization after first live contact. Performance dips at go-live even in well-run transitions, so plan for the dip rather than meeting it unprepared. Tell your internal stakeholders it is coming, because stakeholders escalate an unannounced dip as a failure.

Added together, a serious transition runs three to six months from decision to steady state. Providers with established nearshore delivery capacity can compress the hiring and facilities portion. Discovery, access, and certification work resists compression regardless of who performs it.

What to Evaluate in the Replacement Provider

Deloitte surveyed over 500 business and technology leaders on how they assess service providers. More than 150 were C-suite executives. Transparency and trustworthiness tied at the top, each cited by 54 percent. Understanding of the buyer’s business followed at 40 percent, per the Global Outsourcing Survey 2022.

Now look at what those same executives rated low, because the gap is instructive. Only 27 percent treated collaboration with providers as a strategic priority. Just 22 percent weighed cultural fit during selection. Buyers say they want transparency and trust, then select without weighting the two things that produce either.

That inconsistency is, in our experience, the single best predictor of a future switch. Transparency is not a vendor attribute you can verify in a pitch. It emerges from governance cadence and escalation design. It also depends on whether your account team has reason to raise bad news early. Evaluate the operating relationship, not the promise.

Deloitte also records a conclusion worth quoting at your next vendor review. Service level agreements alone are not effective in maximizing benefits from a vendor relationship. Most clients who switch had SLAs that were technically met right up to the month they decided to leave.

Beyond the relationship, verify the concrete items. Confirm certifications directly rather than accepting a logo on a slide. Check them against the provider’s published compliance credentials. Visit the delivery site, or tour it virtually. A named delivery location tells you more about continuity and labor supply than a country name does. Ask for attrition by program rather than site average. Request references from clients who switched to them, not clients who started with them.

Our checklist of questions to ask before outsourcing a healthcare call center applies well beyond healthcare. The structures in our guide to nearshore call center pricing matter more during a switch than a first engagement. You now have a baseline to compare against, which you did not have the first time.

Budget for the Switch, Not Just the Rate

Switching costs money that does not appear in the new provider’s rate card. Finance teams approve a lower hourly rate, then meet the transition bill. That sequence damages trust in the whole project. Put the full number forward at the start.

Account for parallel running if you choose it, internal program management time, and recruitment or training fees billed separately. Add technology and integration work, plus any early termination charge. Add the cost of the performance dip, which is real even if nobody books it. Then compare that total against the annual savings to get an honest payback period.

KPMG notes that outsourcing deals today are shorter in duration but broader in scope, which changes this arithmetic. A shorter contract shortens the window over which transition costs amortize. Consequently, a fourteen-month payback reads differently on a three-year term than on a five-year one.

That shift also explains why buyer expectations have moved. In KPMG’s research, 81 percent of companies want providers to act as strategic collaborators rather than vendors. Roughly three in four seek transformational outcomes rather than transactional cost reduction, per The Future of Outsourcing. A switch justified purely on rate will disappoint against that expectation, whoever you choose.

Conclusion

Switching call center providers reads as a procurement exercise and behaves like an operations project. The selection decision gets the attention because it feels consequential and fits a familiar process. Meanwhile, the handover, where customers actually feel the change, falls to whoever has capacity. That inversion explains most switches that go badly.

The evidence base here is thinner than the internet suggests, and pretending otherwise does buyers no favors. Provider replacement is demonstrably common across outsourcing. The diagnosed causes have held steady for over a decade. Objectives drift after signature, accountability blurs, and planning falls to whoever has the least incentive to plan. Those are governance problems rather than vendor problems, which is encouraging, because governance is within your control.

So do the unglamorous work first. Read the exit provisions before you serve notice, and assume your tenured agents will leave during it. Choose a cutover design matched to incumbent cooperation, and budget the transition rather than only the rate. Expect a performance dip and tell your stakeholders it is coming. A switch handled this way takes longer than the timeline your provider quoted. It also holds service while it happens.

Frequently Asked Questions

How long does it take to switch call center providers?

A serious transition runs three to six months from decision to steady state. Discovery takes two to four weeks, and systems access takes four to eight. Training needs three to six weeks, then ramp another four to eight. Programs with licensing or clinical content take longer, and systems access slips more often than any other stage.

What is the biggest risk when changing call center providers?

Knowledge transfer, not provider selection. Tenured agents hold undocumented workarounds and escalation paths that never reach the knowledge base. Those agents often leave during the notice period because their account is ending. Treat inherited documentation as a starting point and rebuild your knowledge base from live call observation.

Should I run both providers in parallel during the transition?

Parallel running is the safest design, and you pay twice during the overlap. It suits regulated work, high volume, and low error tolerance. Phasing by queue or by volume share costs less and works for most programs. Parallel and phased designs both need incumbent cooperation, so assess that honestly before committing.

What should I check in my contract before giving notice?

Check the notice period and whether the contract auto-renews. Confirm whether you hold termination for convenience or only for cause, and what exit assistance the incumbent owes. Then confirm who owns call recordings, transcripts, knowledge base content, and quality data. Check the return format and what the provider deletes. Also check which telephony numbers and licences sit in the provider’s name.

Will service quality drop during a call center transition?

Expect a dip at go-live even in well-run transitions. Service often degrades earlier, during the incumbent’s notice period, as attrition climbs and backfilling slows. Agree on a staffing floor in writing for that period. Increase your own monitoring rather than relying on provider reporting, and ask for named continuity of supervisors.

How much does switching BPO providers cost?

Budget beyond the new hourly rate. Include parallel running if chosen, internal program management time, and recruitment or training fees. Add integration work, early termination charges, and the cost of the performance dip. Compare that total against the annual savings for an honest payback period. Shorter contracts give transition costs less time to amortize.

Are there reliable statistics on call center provider churn?

Not from authoritative sources. No analyst firm, Big Four publisher, or government body reports switching or transition-failure rates specific to contact center outsourcing. Everest Group found clients replaced the incumbent in 37 percent of 116 infrastructure outsourcing deals. The figure was 24 percent across 136 application outsourcing engagements, observed in 2013. Those cover different services and should not be read as contact center benchmarks.

When is the worst time to switch providers?

During your seasonal peak, which teams attempt surprisingly often because contract dates dictate the calendar. Extending the incumbent by a quarter almost always costs less than failing through peak volume. Plan the cutover around your demand curve rather than your paperwork.

The post How to Switch Call Center Providers Without Breaking Service appeared first on SkyCom Call Center.

Most discussion of loan processing delays jumps straight to automation as the answer. That instinct is reasonable but incomplete, because the federal and GSE data tell a more interesting story. Manual loan processing costs real money. Yet the gap between fast and slow lenders is wider than any single technology explains. This article works through what the evidence shows about lending process bottlenecks. It covers where they sit in the loan processing workflow, and what closing them achieves.

Average cost to originate sat near 11,600 dollars per loan, according to the 2024 Cost to Originate Study. That figure had climbed roughly 35 percent over three years. The increase came to about 3,000 dollars per file. Origination volume fell across that same window, so the cost rise reflects structural expense rather than growth.

The composition of that cost explains why bottlenecks matter so much. Personnel expenses account for roughly two thirds of total production cost. Every hour a file spends in a queue is an hour somebody is paid to manage, chase, or re-check. Loan processing bottlenecks are therefore not merely a service problem, because they convert directly into payroll consumed per file.

The Gap Between Fast and Slow Lenders Is Wider Than Any Average Suggests

Averages hide the finding that should actually drive strategy. The same Freddie Mac research broke cost to originate into performance quartiles, and the spread is remarkable. Top quartile lenders originated at roughly 6,900 dollars per loan. Bottom quartile lenders spent about 16,500 dollars for the same product.

That is a difference of roughly 9,600 dollars per file between the best and worst performers. Put differently, the worst quartile spends about 2.4 times what the best quartile spends. No single piece of software accounts for a gap that size, and no market condition does either. These lenders operate in the same rate environment and face the same regulatory requirements.

The gap is operational. It reflects how often staff touch a file and how long conditions sit unresolved. Rework volume per application matters just as much. A lender benchmarking itself against the industry average is therefore benchmarking against a number that describes almost nobody. The useful question is which quartile your loan application processing currently sits in.

This framing also changes where improvement effort should go. Chasing a five percent efficiency gain looks reasonable against an average. It looks unambitious against a 140 percent spread that other lenders have already closed.

Automation Shortens Cycle Time by Days, Not Weeks

Here is where the evidence contradicts most marketing on this topic. Vendors promise that automation transforms lending operations, and buyers reasonably expect cycle times to collapse. The measured effect is smaller and more specific than that.

Freddie Mac compared lenders making heavy use of digital origination tools against those using them minimally. The heavy users closed in roughly 34 days. The light users closed in roughly 39 days. Five days separate maximum digital adoption from minimal digital adoption, which is meaningful but hardly a transformation.

The cost effect is similarly bounded. Lenders maximizing those digital capabilities achieved about 1,500 dollars lower cost per loan, a saving of roughly 14 percent. Technology clearly helps, and 14 percent is worth having. Yet 14 percent cannot explain a 140 percent spread between quartiles. Something else drives most of that difference.

That something else is process discipline and available capacity. The study found that leveraging those tools can eliminate between 2.2 and 12.36 hours of production time per loan. Notice the width of that range, because it is the real finding. The same technology saves one lender two hours and another lender twelve. The tool, therefore, is not the variable.

Why Manual Loan Processing Is a Capacity Problem First

Most lenders diagnose bottlenecks as workflow design problems and buy software accordingly. Frequently the actual constraint is simpler. There are not enough trained people to work the queue during the hours when borrowers and third parties respond.

Consider what a conditions list actually requires. Somebody must read the underwriter’s conditions and translate them into plain language. Then that person calls the borrower, explains the requirement, and confirms receipt. Then somebody re-requests the bank statement that has since gone stale, which is the part borrowers find so endearing. No software automates that in the way vendors imply. It is a conversation rather than a data transfer.

With personnel at two thirds of production cost, capacity decisions dominate the economics. A lender that cannot staff follow-up during business hours will show long cycle times regardless of its technology stack. This is where back office processing support changes the arithmetic. It adds trained capacity without adding fixed headcount to a cyclical business.

Origination volume swings hard with rates, which makes permanent staffing for peak volume financially painful. Most lenders therefore staff somewhere below peak and absorb the delay during busy periods. That decision is rational at the budget level and expensive at the file level.

Where Lending Process Bottlenecks Actually Sit

Bottlenecks cluster in predictable places across consumer, commercial, and mortgage lending. Each has a different character, and each responds to a different intervention.

Document Collection at Intake

The first stall happens before underwriting ever sees the file. Borrowers submit incomplete packages, upload illegible scans, or misunderstand which statements you meant. Every round trip adds days, and documents expire while the round trips continue. Front-loading a structured intake conversation removes more delay than any later optimization.

The Conditions Loop

Underwriting issues conditions, and the file waits. This is the single most expensive queue in most mortgage operations, because conditions frequently cascade. One answer produces two new questions, and the borrower hears from three different people about related items. Owning the conditions conversation end to end shortens this loop considerably.

Third-Party Verification

Appraisals, title work, employment verification, and insurance binders all sit outside your control. They still sit inside your cycle time, which is what borrowers experience. Lenders that track third-party aging as a first-class metric close faster than those treating it as somebody else’s delay.

Financial Analysis in Commercial Files

Commercial loan processing carries a heavier analytical burden than consumer lending. Spreading financials, normalizing adjustments, and verifying entity structures takes skilled time that resists hurrying. Teams handling commercial lending operations usually find the bottleneck here rather than at intake.

Pre-Closing Coordination

The final week generates a surprising share of delay. Closing disclosures, funding conditions, and scheduling all converge at once. A file that moved efficiently for a month can lose four days in the last stretch. Nobody owned the calendar.

Thin Margins Make Loan Processing Efficiency Urgent

The cost data only becomes alarming when you set it against current margins. Independent mortgage banks reported a pre-tax net production profit of 973 dollars per loan in Q2 2026. That comes from the Mortgage Bankers Association. The preceding quarter produced 727 dollars per loan.

Hold those two numbers next to each other for a moment. Profit per loan sits in the high hundreds while cost to originate sits in five figures. A few hundred dollars of avoidable processing expense therefore consumes a substantial share of the margin on that file. Efficiency at this point is not an optimization exercise but a profitability requirement.

Note that these two figures measure different things and come from different sources. Freddie Mac’s cost to originate and the MBA’s production profit are built on separate methodologies. The comparison is directional rather than arithmetic, though the direction is unambiguous enough to act on.

Speed Expectations Have Shifted Underneath Lenders

Borrower patience has contracted while processing complexity has grown. Federal data captures part of this shift. Consider the FDIC’s 2024 Small Business Lending Survey. Large banks are much more likely to report deciding a loan in one business day or less.

The same survey found something more encouraging for smaller institutions. Small and large banks are about equally likely to approve a loan within five business days. Scale wins the sprint, in other words, but it does not dominate the ordinary case. That finding should reassure community lenders who assume they cannot compete on speed.

Competitive pressure nonetheless keeps rising. Digital-first lenders have trained borrowers to expect same-day decisions, which reshapes expectations across every channel. Established lenders exploring fintech-style operating models are usually responding to that expectation rather than to a technology fashion.

Section 1071 Will Add Work to Every Small Business Application

Here is a factor almost absent from other articles on lending bottlenecks. The CFPB’s Section 1071 rule requires covered lenders to collect and report application data. It covers credit applications from small, women-owned, and minority-owned businesses. That obligation lands squarely on loan processing staff.

The Bureau issued a final reconsideration rule on May 1, 2026, which revised the rule’s scope and extended compliance. The current compliance date is January 1, 2028. That revision narrowed coverage, adjusted the small business definition, and streamlined which demographic data get collected.

Lenders reading the extension as breathing room may be reading it wrong. Every covered application will require extra data capture, validation, and reporting. That work falls on the same people already managing conditions queues. A processing workflow running at capacity today will not absorb new required fields gracefully. Teams handling consumer and small business lending should treat 2027 as the build year rather than 2028.

How to Diagnose Your Own Loan Processing Workflow

Most lenders track cycle time and pull-through, which describe outcomes rather than causes. Diagnosing bottlenecks requires measuring the queues between milestones instead. The metrics below isolate where files actually wait, and each points at a specific intervention.

Measure these by loan type rather than in aggregate. Lenders adding capacity through nearshore delivery teams should track the same queues before and after. The effect then becomes visible. Mortgage, consumer, and commercial files bottleneck in genuinely different places, so a blended number describes none of them accurately. Lenders running several product lines across banking and financial services operations usually find one product dragging the whole average.

Conclusion

Loan processing bottlenecks persist because they hide inside averages. A blended cycle time looks acceptable while individual files sit for days in queues nobody monitors. The cost of those queues surfaces as payroll rather than as visible delay. Finance teams therefore rarely flag them. Borrowers notice long before the reporting does.

The evidence points somewhere more useful than another technology purchase. Digital tools deliver roughly 14 percent cost improvement and about five days of cycle time, both worth capturing. Yet the 2.4x spread between best and worst quartile lenders proves something else. Most of the opportunity sits in how teams sequence, own, and staff the work. Those levers cost less than a platform migration and move faster.

Margins make the timing urgent rather than optional. Production profit runs in the high hundreds per loan. A few hundred dollars of avoidable processing expense is therefore no rounding error. Section 1071 data collection then lands on the same teams before 2028. Lenders running at capacity today face a harder year ahead. Measuring your queues now is considerably cheaper than discovering them later.

Frequently Asked Questions

What causes loan processing bottlenecks?

The most common causes are incomplete document collection at intake and unresolved underwriting conditions. Third-party turnaround on appraisals and verifications adds more. Capacity constraints compound all three, because personnel account for roughly two thirds of production cost. Files wait when nobody has time to chase them rather than because the workflow is poorly designed.

How much does it cost to originate a loan?

Freddie Mac’s 2024 Cost to Originate Study put the average near 11,600 dollars per loan. It used third quarter 2023 data from 203 institutions. Top quartile lenders achieved roughly 6,900 dollars while bottom quartile lenders spent about 16,500 dollars. That spread matters more than the average.

Does automation actually fix loan processing delays?

Automation helps measurably but less dramatically than vendors suggest. Lenders with heavy digital tool usage closed in roughly 34 days versus 39 days for light users. They also achieved about 14 percent lower cost per loan. The remaining gap between fast and slow lenders comes from process discipline and staffing capacity.

Where do mortgage processing delays usually occur?

The conditions loop after underwriting generates the most delay in most mortgage operations. Conditions frequently cascade into further follow-up questions. Third-party items such as appraisal and title sit outside your control while remaining inside your cycle time. Pre-closing coordination costs more days than teams expect.

How is commercial loan processing different?

Commercial files carry a heavier analytical burden than consumer lending. Spreading financials, normalizing adjustments, and verifying entity structures require skilled time that resists compression. Consequently, commercial bottlenecks tend to appear in analysis rather than at document intake.

How does Section 1071 affect lending operations?

Covered lenders must collect and report data on small business credit applications. That adds capture and validation work to every file. The CFPB issued a final reconsideration rule on May 1, 2026, setting compliance for January 1, 2028. That work will fall on staff already managing existing processing queues.

Should lenders outsource loan processing work?

Outsourcing suits the parts of the workflow that scale with volume rather than the parts requiring credit judgment. Document chasing, condition follow-up, status calls, and verification coordination absorb enormous staff time without needing underwriting authority. Keeping decisions in-house while adding external capacity handles volume swings without permanent headcount.

The post Loan Processing Bottlenecks: Where Manual Work Slows Lending Decisions appeared first on SkyCom Call Center.

Healthcare referral leakage is not primarily a loyalty problem or a marketing problem. It is a workflow problem, and the evidence on that point is unusually strong. Peer-reviewed research has tracked what happens to referrals after they leave a primary care office. The findings are worse than most administrators expect. This article walks through that evidence and shows where the referral-to-appointment process actually breaks. It then sets out what referral leakage prevention requires in practice.

What Referral Leakage in Healthcare Actually Means

Patient referral leakage describes any referral that leaves the intended network or never reaches care at all. Most definitions in circulation emphasize the first half. A patient receives an in-network referral, chooses an outside provider instead, and the originating system loses the revenue. That version of the definition suits vendors selling network-steering tools, so it dominates the search results.

The operational definition is broader and more useful. Leakage includes every referral that fails to produce a completed encounter, whoever ends up providing the care. A referral that goes nowhere is a worse outcome than a referral that goes to a competitor. In the first case, the patient is untreated, while in the second, the patient at least received care somewhere.

Referral management in healthcare therefore covers two distinct failures. One is leakage out of network, which is a competitive and contracting issue. The other is leakage out of care entirely, which is a clinical safety issue. Organizations that conflate the two tend to buy steering software when they need a working patient access process instead.

The Referral-to-Appointment Process Fails More Often Than Leaders Assume

The most rigorous public data comes from a 2018 study in the Journal of General Internal Medicine. Researchers at a large academic medical center examined what happened to referrals across an entire fiscal year. The scale matters here, because it removes the usual objection that referral problems are anecdotal.

The team analyzed 103,737 referral scheduling attempts drawn from 90,437 referral orders. Those referrals went to 20 high-volume specialties from 24 primary care sites. Of those scheduling attempts, only 34.8 percent resulted in a documented complete appointment, according to Patel and colleagues. Nearly 39 percent carried no documented appointment date at all.

Read that figure carefully, because it is easy to overstate. It measures documented completion inside one system’s records, not a national leakage rate. Some patients in that missing majority certainly received care elsewhere without it appearing in the chart. Even allowing generously for documentation gaps, the number describes a referral workflow that loses track of most referrals.

That distinction between “went elsewhere” and “went nowhere” is exactly what most organizations cannot measure. If your reporting cannot separate the two, your improvement plan is guesswork. A patient who quietly gave up looks identical to one who saw a specialist across town.

Why the Healthcare Referral Workflow Breaks Down

The causes of patient referral leakage are boringly operational. None of them involve patients disliking the health system, which is the explanation executives reach for first. The referral simply encounters friction, and friction wins.

The Handoff Loses Information

The classic account appeared in The Milbank Quarterly in 2011, in a paper titled “Dropping the Baton.” Its authors examined the full referral process and reached a blunt conclusion. “There are breakdowns and inefficiencies in all components of the specialty-referral process,” wrote Mehrotra, Forrest and Lin.

They were equally direct about what gets lost in transit. “Many referrals do not include a transfer of information, either to or from the specialist; and when they do, it often contains insufficient data for medical decision making.” A specialist who receives a name and a diagnosis code cannot triage that patient sensibly. So the scheduler books the appointment late, or books the wrong one entirely.

Nobody Owns the Next Step

The same paper identified a gap that no software fixes on its own. “PCPs often do not know whether a patient actually went to the specialist,” the authors observed, “or what the specialist recommended.” Ownership of the referral effectively ends at the moment of the order. After that, the patient becomes responsible for navigating a system they do not understand.

Most patients handle this badly, which is not a criticism of patients. We ask them to call an unfamiliar office, decode their coverage, and outlast a hold queue while unwell. Reliable appointment scheduling support exists precisely because that expectation is unrealistic.

Coverage and Authorization Stop the Referral Cold

Two administrative steps quietly kill a large share of referrals. The first is insurance verification, because a patient who cannot confirm coverage usually postpones rather than proceeds. Delayed care has a way of becoming abandoned care. Front-loading eligibility verification removes that stall before the patient ever feels it.

The second is prior authorization, which introduces a waiting period nobody owns. The referring office assumes the specialist is handling it, and the specialist assumes the reverse. Meanwhile the patient hears nothing for three weeks and concludes the whole thing was optional. Treating prior authorization as a tracked task rather than background paperwork closes that gap.

The Patient Is Never Contacted Again

A referral order is not a communication. It is an entry in a chart that the patient may or may not have understood. Many patients leave the visit unsure whether someone will call them or whether they should call. That ambiguity alone accounts for a meaningful share of leakage, and it costs nothing to fix.

The Communication Gap Between Referring and Receiving Clinicians

There is a second failure running alongside the patient-facing one, and it is arguably stranger. Clinicians on both sides of a referral believe they are communicating well. The data says only one side can be right.

Researchers at the Center for Studying Health System Change surveyed 4,720 physicians. Each provided at least 20 hours of weekly patient care. Their results appeared in Archives of Internal Medicine in 2011. Among primary care physicians, 69.3 percent said they always or usually send patient history and the reason for consultation. Only 34.8 percent of specialists said they receive that information with similar frequency, per O’Malley and Reschovsky.

The return leg shows the same pattern. Specialists reported sending consultation results back 80.6 percent of the time. Only 62.2 percent of primary care physicians reported receiving them. Both gaps point the same direction. Information is leaving one office and not arriving at the other, and neither party knows it.

Note that these percentages describe a different thing than the completion figure quoted earlier. The coincidence of two studies both landing on 34.8 percent is genuinely accidental. One measures appointments that happened and the other measures records that arrived. Merging them would produce a statistic that means nothing.

Referral Leakage Prevention Is Already a Federal Quality Measure

Here is the part that most referral leakage articles omit entirely. Closing the referral loop is not merely a best practice that consultants recommend. It is a scored quality measure inside the Medicare payment system, which changes the business case considerably.

CMS maintains MIPS Measure 374, “Closing the Referral Loop: Receipt of Specialist Report.” CMS classifies it as a process measure and flags it as high priority. The denominator covers patients who received a referral during the performance period. The numerator counts those for whom the referring clinician actually received a report back.

That report must be a written document. It has to contain findings, a care summary, an assessment, or a treatment plan. A notation that the patient did not attend also qualifies. The last option deserves attention from anyone building referral tracking. Under this measure, documenting non-attendance closes the loop. Knowing the patient never arrived carries real clinical value.

The measure specification also cites encouraging improvement evidence. Enhanced electronic health record capability combined with process redesign raised closure rates to 76.8 percent in the cited work. That figure matters strategically. It demonstrates that referral loop closure responds to deliberate operational effort rather than remaining stubbornly fixed.

What Patient Referral Leakage Costs, and Why Nobody Can Tell You Precisely

Search for the cost of referral leakage and you will find confident figures everywhere. A commonly repeated claim puts losses between 200 and 500 million dollars annually per health system. Another says 55 to 65 percent of referrals leak out of network. These numbers appear across dozens of blogs, usually without a citation anyone can follow.

We are not going to repeat them, and it is worth explaining why. Those figures trace back to vendor marketing rather than to peer-reviewed research or government reporting. A number nobody can trace to its source is not evidence, however often blogs republish it. Build a business case on an untraceable statistic and finance will eventually ask where it came from.

The mechanism, however, is straightforward enough to model with your own data. Each referral represents an expected downstream encounter, and often a chain of them. A cardiology referral may carry diagnostics, a procedure, and follow-up visits behind it. When the referral fails, the entire chain disappears from the forecast without appearing anywhere as a loss.

That is the genuinely dangerous property of healthcare referral leakage. It never shows up as a line item in revenue cycle reporting. Denied claims appear in reporting, bad debt appears in reporting, and leaked referrals appear nowhere at all. Multiply your average downstream value per referral by your documented non-completion count. That gives you a defensible internal figure rather than a borrowed one.

Building a Closed-Loop Referral Management Process

Referral leakage prevention is less about technology than most vendors suggest. The organizations that improve share a few structural habits, and none of them require replacing your electronic health record.

The first habit is assigning ownership past the order. Somebody must own the referral until a confirmed appointment exists or the patient formally declines. Without a named owner, the referral belongs to the patient by default. That arrangement fails reliably, as the completion data demonstrates.

Track Every Referral as a Workflow Status

The second habit is treating the referral as a tracked object with a status, not as a completed task. A referral should sit in one of a small number of states at any moment. It might be sent, acknowledged, scheduled, completed, declined, or stalled. Any referral without a status change for a defined period should surface automatically for follow-up.

The third habit is outbound contact rather than passive waiting. Someone should call the patient within a short window, confirm they understood the referral, and help them book. Patient engagement support makes this practical at volume, because most practices cannot absorb the call load internally. The outreach does not need to be elaborate, and a short confirming call resolves a surprising share of stalls.

Language is the habit most organizations forget entirely. A Spanish-speaking patient who receives an English voicemail has effectively received nothing. That patient then appears in your data as unresponsive rather than as unreached. The distinction sends your improvement effort in opposite directions. Staffing follow-up with Spanish-language agents closes a gap that reminder software alone cannot.

The fourth habit is closing the loop back to the referring clinician. This is where MIPS Measure 374 and good medicine point in the same direction. The referring physician needs the specialist’s report, or at minimum needs to know the visit never happened. Both outcomes are actionable, whereas silence is not.

The fifth habit is measuring the process rather than the outcome alone. Track time from order to first patient contact, and time from order to scheduled appointment. Those two intervals predict completion better than any satisfaction score. They are also the two things an operations team can directly change.

How to Measure Patient Referral Management Performance

Most organizations track referral volume and little else. Volume tells you how much work entered the pipe, not how much came out the other end. The following metrics give a clearer picture of referral workflow health, and each one maps to a specific intervention.

Large organizations tend to need this reporting at the service-line level rather than in aggregate. A hospital or health system usually finds that leakage concentrates in two or three specialties. It rarely spreads evenly. Independent medical groups face a different version of the problem. They lack the internal staffing to chase every stalled referral, so leakage spreads thinly across everything.

Frequently Asked Questions

What is referral leakage in healthcare?

Referral leakage in healthcare occurs when a referred patient does not complete care within the intended network. The broader operational definition includes any referral that never produces a completed encounter anywhere. The second category matters more clinically, because those patients receive no care at all rather than care elsewhere.

What percentage of referrals are actually completed?

A 2018 study in the Journal of General Internal Medicine analyzed 103,737 referral scheduling attempts. The setting was a large academic medical center. Only 34.8 percent resulted in a documented complete appointment. That figure reflects documentation within one system rather than a national rate. It still indicates that most referrals leave no trace of completion.

What causes patient referral leakage?

The main causes are operational rather than reputational. Referrals lose clinical information in transit, ownership ends at the order, and coverage or authorization steps stall the process. Many patients also never receive a follow-up contact, so they never learn what they were supposed to do next.

Is referral loop closure a regulatory requirement?

Closing the referral loop is a scored quality measure rather than a strict mandate. CMS maintains MIPS Measure 374, “Closing the Referral Loop: Receipt of Specialist Report,” as a high-priority process measure. Documenting that a patient did not attend also satisfies the measure, which many organizations overlook.

How much does referral leakage cost a health system?

Widely quoted dollar figures generally trace to vendor marketing rather than peer-reviewed or government sources. We therefore avoid repeating them. Calculate your own estimate instead. Multiply your average downstream revenue per referral by your documented non-completion count for a defensible internal number.

How do you prevent referral leakage?

Assign ownership of each referral beyond the order, and track it as an object with a status. Contact patients outbound within a short window rather than waiting for them to call. Resolve eligibility and authorization early, and confirm that the specialist report returns to the referring clinician.

What is a closed-loop referral workflow?

A closed-loop referral workflow tracks a referral from order through completion and back to the referring clinician. The loop closes only when that clinician receives the specialist report or a documented notation of non-attendance. Anything less leaves the referring physician managing a patient without knowing what happened.

Conclusion

Referral leakage in healthcare persists because it is invisible in the places organizations normally look. It produces no denial, no complaint, and no line item in a variance report. The patient simply stops appearing, and the referring physician rarely finds out. That silence is what makes the problem durable rather than any lack of concern from the people involved.

The peer-reviewed evidence points consistently toward process rather than preference. Referrals fail when information does not travel and when nobody owns the next step. They also fail when administrative steps stall unattended. Each of those failures responds to ordinary operational discipline. The improvement data in the MIPS specification confirms that closure rates move when organizations work at them.

Treating patient referral management as a tracked workflow rather than a completed order changes the outcome. Assign ownership, give every referral a status, and contact patients before they drift. Then close the loop back to the clinician who started it. None of that requires new software. All of it requires somebody accountable for referrals that go quiet.

Close the Loop on Your Referral Workflow

Can your organization separate referrals that went elsewhere from referrals that went nowhere? If not, that gap is the place to start. SkyCom staffs referral follow-up, appointment scheduling, insurance verification, and prior authorization from nearshore delivery centres across Latin America. Those are the four points where the referral workflow most often stalls. Our teams work US hours in English and Spanish.

Share your referral volume, your specialty mix, and where your current process loses visibility. You will get an assessment of where leakage concentrates and what closing it would take operationally. Get a Referral Workflow Assessment, and we will walk through your referral data.

The post Referral Leakage in Healthcare: Where Patients Get Lost Before Care Begins appeared first on SkyCom Call Center.

The comparison is usually framed as a question about accent or language quality. That framing is misleading. Both markets support English-language voice programs for US customers. The more consequential differences turn on four things. Those are time zone behavior, available scale, hurricane exposure, and workforce depth. This guide works through each of those in turn.

The Two Markets Are Moving in Opposite Directions

Jamaica built the larger industry first. For roughly a decade it was the default Caribbean answer for US voice programs. Its sector grew steadily on that reputation. That growth has now reversed. Reporting on the Global Services Association of Jamaica puts current sector employment near 50,000. Two years earlier that figure stood at roughly 62,000. Sector spending fell over the same period.

The causes are worth understanding, because buyers frequently guess wrong about them. GSAJ president Yoni Epstein addressed the assumption directly: “The job loss has been caused by a few things, but artificial intelligence isn’t one.” The contraction traces instead to US onshoring decisions and general cost pressure on outsourcing budgets. Geopolitical shifts in sourcing strategy and disruption from Hurricane Melissa did the rest.

Belize moved the other way, though from a much smaller base. Its industry grew under a deliberate government programme rather than organic market pull. Consequently, the two markets now present very different profiles. Jamaica offers established depth in a contracting market. Belize offers a smaller but policy-supported market with fewer incumbent operators.

Neither direction is automatically good or bad for you. A contracting market can mean available experienced talent and softer pricing. A growing market can mean constrained supply as demand catches up. What matters is matching the market’s shape to what your program actually needs.

Time Zone Is the Most Underrated Difference

This is where the two countries diverge most cleanly, and it is the factor buyers most often overlook. Jamaica runs on Eastern Standard Time and does not observe daylight saving. Belize runs on Central Standard Time and has not observed daylight saving since 1983. Both are stable year-round, which is itself an advantage over markets that shift twice a year.

The practical consequence appears in shift coverage. Jamaica aligns with US Eastern time for half the year. It falls an hour behind during US daylight saving months. Belize aligns with US Central time in winter and sits an hour behind Central during summer. Therefore, an East Coast program gets cleaner morning coverage from Jamaica. A program weighted toward Central and Mountain customers fits Belize better.

The gap is only an hour, so neither site is unworkable for either customer base. Still, an hour matters at the edges of a shift. Morning peak volume is where a one-hour offset either costs you an early-shift premium or does not.

That last row deserves emphasis. Buyers treat the comparison as a choice, but the offset can be used deliberately. Two sites an hour apart stretch your staffed window without paying anyone a night differential.

Scale Versus Scarcity

Jamaica’s larger labor pool is its clearest structural advantage. That scale makes larger ramps easier to support. It matters most when a program needs hundreds of seats or rapid seasonal expansion. It also means a deeper bench of supervisors, trainers, and quality analysts who have already worked US programs.

Belize cannot match that depth. Its population is a fraction of Jamaica’s, and its outsourcing sector is correspondingly smaller. Government support runs through BELTRAIDE and its Designated Process Area programme, which grants qualifying operators specific incentives. That programme has attracted a set of operators, but the absolute pool remains limited.

For a 30-seat program with modest growth plans, that limitation rarely binds. For a program that needs 200 agents by next quarter, it does. Smaller markets also concentrate risk. When one large operator ramps, wage pressure and attrition spread quickly.

On the other hand, a smaller market can make your program a priority rather than one account among many. That is a real operational difference, and it shows up in how quickly issues get senior attention.

Why Belize Workforce Numbers Are Hard to Compare

Buyers deserve honesty about the evidence base here, because the Belize numbers in circulation are unreliable. Published workforce figures range from roughly 6,000 seats to 20,000 workers depending on the source. Those sources are overwhelmingly vendor marketing pages rather than government statistics or analyst research.

The same caution applies to accent claims. Accent should be evaluated at the provider level rather than assumed from the country. Ask to hear calls from the actual delivery team and review the QA standards behind them. Test customer comprehension rather than trusting country-level marketing.

The practical lesson is to verify workforce claims at the provider level rather than the country level. Ask how many agents the provider actually employs in that country. What its attrition rate has been over the past twelve months. Ask how long its longest-running US account has operated. Those answers are checkable. National headcount estimates generally are not.

Hurricane Exposure and Continuity Planning

Both countries sit in the Atlantic hurricane belt, so neither offers weather immunity. Jamaica’s recent experience with Hurricane Melissa demonstrated how directly a storm can affect sector capacity. Belize faces comparable seasonal exposure along its coast.

What separates providers is continuity design rather than geography. Infrastructure resilience varies by provider and by site, not by country. Ask any provider in either country three specific questions. How does the site fail over when local power goes down? What is the work-from-home activation process, and how fast does it run? Which other site absorbs volume if the country goes dark for a week?

A provider operating in several countries answers that third question easily. A single-country provider cannot answer it at all. Consequently, multi-country delivery is worth more in this region than in markets with lower weather risk.

Data protection adds another consideration. Jamaica’s Data Protection Act governs personal data handling and gives US buyers a recognizable framework to contract against. Confirm how your provider maps its controls to whichever regime applies. Get that mapping in writing before volume moves.

Cost, Attrition, and What Actually Drives Total Spend

Hourly rates in both markets sit well below US onshore levels and above the lowest far-offshore options. The spread between Belize and Jamaica is narrower than buyers expect. It rarely decides the engagement on its own.

Attrition matters more. It can materially change the economics of a program, because recurring recruitment and training costs compound as turnover rises. Lower attrition also protects quality scores and preserves the product knowledge that makes tenured agents valuable.

Ask for attrition by program rather than by site. Site-level averages hide the pattern that matters. A stable long-running account and a churning new one blend into a number describing neither. Our guide to nearshore call center services sets out the broader regional framework. Our inbound call center and customer engagement pages cover the program types these sites typically run.

Four Questions That Settle the Choice

Most buyers can resolve this decision quickly by answering four questions honestly.

First, where are your customers concentrated? An East Coast base leans Jamaica, and a Central base leans Belize. A national base makes coverage design matter more than the country.

Second, how fast do you need to scale? A ramp beyond roughly a hundred seats within a quarter argues for the deeper labor market. A steady 20- to 50-seat program fits either.

Third, what happens if the site goes offline for a week? If that would seriously damage your business, single-country delivery is the wrong structure. The country you pick does not change that.

Fourth, how specialized is the work? Programs requiring regulated-industry experience, long ramp curves, or complex technical support benefit from the larger supervisory bench. Generalist programs are far less sensitive to that difference.

Read the right column as an operating model rather than a winner. The answers tell you which delivery structure fits, and for many programs that structure involves more than one country.

Not Sure Whether Belize, Jamaica, or a Multi-Site Model Fits?

SkyCom operates nearshore delivery across several Latin American countries. Therefore, we have no incentive to push you toward one site. We will tell you which site fits your customer distribution, ramp plan, and continuity requirements. Sometimes that answer is a split across two sites.

Share your customer geography, monthly volume, required coverage hours, and growth plan. SkyCom will map those requirements to an appropriate nearshore delivery model and tell you where the trade-offs sit. You get a staffing model rather than a brochure.

Get a Nearshore Site Assessment

Frequently Asked Questions

Is Belize or Jamaica better for a US call center program?

Neither is universally better. Jamaica offers a deeper labor pool and Eastern time alignment, which suits larger programs and East Coast customer bases. Belize offers Central Time alignment and a smaller, government-supported market. That suits mid-sized programs weighted toward Central and Mountain customers. Your customer geography and ramp speed should decide it.

What is the time difference between Belize, Jamaica and the US?

Jamaica stays on Eastern Standard Time year-round, and Belize stays on Central Standard Time year-round. Neither observes daylight saving. Consequently, each site shifts by one hour relative to your US teams when US daylight saving begins and ends.

Which is better for a large call center operation, Belize or Jamaica?

Jamaica is substantially larger, with sector employment around 50,000 according to industry association reporting. That figure has declined from roughly 62,000 two years earlier. Reliable national workforce figures for Belize are harder to find. Therefore, verify headcount at the provider level rather than trusting country-wide estimates.

Why has Jamaica’s call center employment fallen?

Industry leadership attributes the decline to US onshoring and cost pressure on outsourcing budgets. Geopolitical sourcing shifts and hurricane disruption also contributed. The Global Services Association of Jamaica has explicitly said artificial intelligence is not the cause.

Should a US company use one nearshore location or multiple sites?

That depends on how much a country-level outage would cost you. A single site is simpler to manage and usually cheaper to run. Two sites an hour apart widen your staffed coverage window. They also give volume somewhere to go during a storm or a power event. Programs with strict uptime commitments generally justify the second site.

How should companies compare Belize and Jamaica call center costs?

Compare total cost of the engagement rather than the hourly rate. The rate spread between the two markets is narrower than buyers expect. Ask for twelve-month attrition by program, ramp timelines, and the cost of the supervisory layer. Turnover and re-training usually move the total more than the headline rate does.

How should I handle hurricane risk in either country?

Evaluate the provider’s continuity design rather than the country’s weather. Ask how the site fails over during a power loss. Ask how fast work-from-home activates, and which other site absorbs volume during a long outage. Multi-country providers answer that last question; single-country providers cannot.

Do agents in Belize or Jamaica have better accents for US customers?

Available evidence does not support an accent advantage for either market. Both supply native English speakers, and outcomes depend on the provider’s recruiting standards and training rather than nationality. Evaluate recorded calls from the actual delivery team instead of trusting country-level claims.

What Actually Decides the Choice

Buyers who frame this comparison around language quality are answering a question that both markets already settled. The difference between two well-run sites comes down to recruiting and training, not to nationality.

The decisions that actually shape your outcome are less glamorous. They concern which time zone lines up with your morning peak. They also concern whether the local labor market supports your ramp, and what happens during a storm. Answer those and verify the numbers at the provider level rather than the country level. Once those factors are mapped against your program requirements, the appropriate delivery model becomes much easier to evaluate.

For a good number of programs, that evaluation lands on both sites rather than one. Splitting delivery extends your coverage window and gives you somewhere for volume to go when weather interrupts a country. We are happy to model that scenario alongside the single-site options so you can compare them properly.

The post Belize vs Jamaica Call Center: Which Nearshore Market Fits Your Program? appeared first on SkyCom Call Center.