Case Study: How an 8-Cardiologist Group Cut Coding Denials by 62%

Cardiology billing is among the most technically demanding in all of medicine. The specialty combines high-volume office visits with complex procedural coding that spans cardiac catheterizations, electrophysiology studies, echocardiograms, nuclear stress tests, and interventional procedures — each with its own modifier requirements, bundling rules, and payer-specific documentation thresholds. A single cardiac catheterization encounter can generate six or more CPT codes with overlapping modifier requirements that differ by payer, and a single error can turn a $4,000 reimbursement into a denial.

This case study examines how an 8-cardiologist group — performing diagnostic and interventional catheterizations, electrophysiology studies, echocardiograms, stress tests, and office visits across two locations — reduced its claim denial rate from 22% to 8.4% using AI-powered cardiology-specific coding and automated prior authorization for cardiac imaging. The results represent aggregate outcomes observed over a 12-month implementation and optimization period.

Note: Specific metrics in this case study are representative figures based on composite customer outcomes. Individual results vary based on practice size, payer mix, and baseline performance.

Results at a Glance

Metric	Before	After	Change
Overall denial rate	22.0%	8.4%	-62%
Annual recovered revenue	—	$1.8M	—
Coding accuracy	78%	96%	+18 pts
Auth-related denials	11.2% of denials	<1%	Effectively eliminated
Denial management FTEs	5	2	3 redeployed
Avg. denial resolution time	38 days	11 days	-71%
Clean claim rate	74%	93.8%	+19.8 pts
Annual denial rework costs	$820K	$280K	-66%

The Challenge: A 22% Denial Rate in a High-Complexity Specialty

The Group's Profile

The group consisted of 8 cardiologists — 3 interventional cardiologists, 2 electrophysiologists, and 3 general/noninvasive cardiologists — operating across a main office with an in-office cath lab and a satellite clinic. The group performed approximately 2,400 cardiac catheterizations annually, 1,800 echocardiograms per month, 600 EP studies per year, 900 nuclear and exercise stress tests per month, and 3,200 office visits per month.

Total annual charges exceeded $38 million, with a payer mix of 42% Medicare, 28% commercial, 18% Medicare Advantage, 8% Medicaid, and 4% self-pay. The billing team consisted of 11 FTEs: 3 coders, 3 claims processors, 2 prior authorization specialists, 2 denial management staff, and 1 billing manager.

At a 22% denial rate, the group was generating approximately 1,900 denied claims per month. At an average claim value of $420, that represented $798,000 in monthly at-risk revenue — nearly $9.6 million annually at risk, with roughly 19% of denied claims ultimately written off.

Root Cause Analysis: Why Cardiology Coding Is Uniquely Prone to Denials

The group's denial landscape revealed four distinct problem areas, each rooted in the structural complexity of cardiology billing.

Modifier complexity — 26/TC split, multiple procedure discounting, and laterality (36% of all denials). Cardiology billing depends on modifier usage more than almost any other specialty. The professional/technical component split (modifier 26 for professional, modifier TC for technical) applies to virtually every diagnostic study — echocardiograms, stress tests, nuclear imaging, and catheterization interpretations. The group owned its own equipment and employed its own technicians, meaning every diagnostic study required a decision about whether to bill globally, or split into 26 and TC components — and the correct answer depended on the place of service, the patient's payer, and the specific study performed.

Multiple procedure discounting created another layer of complexity. When a patient underwent both a left heart catheterization and coronary angiography in the same session, CMS applied a 50% reduction to the lesser procedure. Some commercial payers followed CMS guidelines; others had their own discounting schedules. The group's coders were inconsistently applying discounting logic, sometimes billing codes that triggered automatic payer discounting when a different code combination would have yielded higher reimbursement without any discounting.

Laterality modifiers (LT/RT) for procedures like coronary stenting or peripheral vascular interventions were missed on 8% of applicable claims, triggering automatic denials from payers that required explicit laterality designation.

Cardiac catheterization bundling errors (28% of all denials). Cardiac catheterization coding underwent a major overhaul with CPT 2023 updates, consolidating dozens of catheterization codes into a smaller set of bundled codes. The group's coders were still manually navigating the catheterization coding matrix — left heart cath (93452), right heart cath (93451), combined left and right (93453), coronary angiography (93454-93461), left ventriculography (93565), and the various add-on codes for grafts (93564), right ventricular angiography, and catheter placement.

The bundling traps were numerous. Coronary angiography was bundled into certain catheterization codes but separately billable with others. Left ventriculography was an add-on to left heart catheterization but was frequently miscoded as a standalone procedure. Selective coronary artery injection codes had to align with the specific vessels imaged and the catheterization technique (radial vs. femoral access). The coders estimated that they spent more time on catheterization coding than any other procedure category, yet catheterization claims had the highest denial rate in the practice — 31% of cath claims were denied on first submission.

Echocardiography coding specificity (19% of all denials). Echocardiography coding requires precise differentiation between study types: transthoracic (93306, 93307, 93308), transesophageal (93312-93318), stress echo (93350-93351), and Doppler studies (93320, 93321, 93325). The most common error was failing to distinguish between complete and limited echocardiograms — billing a complete TTE (93306) when the documentation supported only a limited study (93308), or vice versa. Payers, particularly Medicare, had tightened their medical necessity requirements for echocardiography, and the documentation had to explicitly support the type and extent of study performed.

Color flow Doppler (93325) was an add-on code that required its own medical necessity justification when billed with a TTE. The group's cardiologists routinely performed color flow Doppler as part of every echocardiogram, but the documentation didn't always explicitly state why color flow was clinically necessary for that specific patient. Medicare denied color flow Doppler add-on codes at a rate of 17% for the group — nearly double the national average.

Prior authorization gaps for cardiac imaging (17% of all denials). Nuclear stress tests, cardiac MRI, cardiac CT angiography, and stress echocardiograms required prior authorization from most commercial payers and Medicare Advantage plans. The group's two prior authorization specialists managed approximately 280 authorization requests per month, but the volume exceeded their capacity during peak periods.

Authorization requirements varied dramatically by payer. One commercial payer required prior auth for all nuclear stress tests; another required it only for patients under 65. Some Medicare Advantage plans required prior auth for stress echocardiograms; others did not. The authorization specialists maintained a manual tracking spreadsheet, but the rules changed frequently enough that the spreadsheet was reliable only for the highest-volume payers. For lower-volume payers, the specialists often discovered the authorization requirement after the study was performed — when the claim was denied.

The group estimated that 11.2% of all denials were attributable to authorization gaps, representing approximately $360,000 in annual at-risk revenue.

The Staff Burden

Five of the group's 11 billing staff were dedicated to denial-related activities: 2 full-time denial management specialists, plus an estimated 60% of each coder's time spent on denial research, appeals preparation, and corrective coding rather than productive coding work. Staff turnover in the coding team was 40% annually — the group had lost two experienced cardiology coders in the preceding 18 months and replaced them with generalist coders who required 6-9 months of specialty training to reach baseline competency.

The turnover cycle was devastating. An experienced cardiology coder who understood catheterization bundling and modifier logic was difficult to replace. Each departure reduced coding accuracy and increased denials, which increased the workload on remaining staff, which increased burnout, which drove further turnover.

The Solution: QuickCode and QuickAuth for Cardiology

After evaluating three RCM technology platforms, the group selected QuickIntell's cardiology-specific configuration, deploying two products to address the root causes of their denial problem.

QuickCode: Cardiology-Specific AI Coding

QuickCode was deployed with a coding model trained specifically on cardiology documentation, procedure reports, and billing patterns. Unlike general-purpose coding AI, the cardiology model understood the clinical context of cardiac procedures and the coding relationships between them.

Cardiac catheterization coding engine. QuickCode's catheterization module parsed operative reports for key clinical elements — chambers catheterized, vessels injected, ventriculography, access site, and interventional procedures — then applied correct bundling logic (including post-2023 consolidated codes) and assigned appropriate modifiers. The system automatically handled diagnostic-to-interventional conversion cases, applying the correct code hierarchy to capture separately billable diagnostic components.

Modifier logic engine. A payer-specific modifier decision tree determined 26/TC vs. global billing, applied multiple procedure modifiers, added laterality modifiers, and verified modifier consistency. The engine's most valuable capability was payer-specific awareness — applying modifier 59 for one payer and XS for another on identical code pairs, and sequencing codes to minimize multiple procedure discounting impact.

Echo coding specificity. QuickCode distinguished complete from limited studies based on documentation content, verified Doppler and color flow Doppler medical necessity justification, and flagged ambiguous documentation for coder clarification before claim submission.

QuickAuth: Automated Cardiac Imaging Authorization

QuickAuth addressed the prior authorization gap by automating the determination, submission, and tracking of authorizations for cardiac imaging studies.

Payer-specific authorization determination. At the point of scheduling, QuickAuth checked the patient's specific insurance plan against its authorization requirement database to determine whether the ordered study required prior auth. The database was maintained at the plan level (not just the payer level), capturing the variation between a payer's HMO, PPO, and Medicare Advantage products.

Automated auth submission. When authorization was required, QuickAuth compiled the clinical information needed to support the request — the ordering physician's indication, relevant prior studies, symptoms, and risk factors — and submitted the request electronically to payers with electronic auth portals or via fax for payers without digital submission options.

Real-time status tracking. Authorization status was tracked automatically, with alerts for pending requests approaching expiration, approved authorizations nearing their valid date window, and denials requiring peer-to-peer review. The system ensured that no patient was scheduled for a study without active authorization, and no authorization expired before the study was performed.

Implementation Timeline

Phase 1: QuickCode Deployment (Months 1-4)

Month 1: Integration and shadow mode. QuickCode was integrated with the group's EHR and practice management system. All claims were processed through QuickCode in shadow mode — the AI generated codes alongside the human coders, and discrepancies were reviewed daily by the billing manager and the QuickIntell implementation team.

During shadow mode, QuickCode identified coding discrepancies in 24% of claims — consistent with the group's 22% denial rate plus an additional 2% of claims that had coding errors but were paid anyway (typically underpayments due to suboptimal code sequencing).

Month 2: Catheterization coding activation. QuickCode was activated for catheterization coding first, replacing manual cath report coding with AI-generated codes reviewed by the senior coder. The catheterization denial rate dropped from 31% to 14% in the first 30 days of active deployment.

Month 3: Full coding activation. QuickCode was activated for all procedure types — echocardiograms, stress tests, EP studies, and office visits. Coders transitioned from primary coding to AI code review, approving or modifying AI-generated codes rather than assigning codes from scratch.

Month 4: Optimization. The first round of model refinement based on the group's specific payer response patterns. QuickCode's accuracy improved from 93% to 95.2% as the model incorporated denials and payer-specific corrections from the first three months.

Phase 1 results (month 4): Denial rate at 14.6%. Coding accuracy at 95.2%. Catheterization-specific denial rate at 10.8%.

Phase 2: QuickAuth Deployment (Months 4-7)

Month 4-5: Authorization database configuration. QuickAuth's authorization requirement database was configured for the group's 32 contracted payers and plans. The two prior authorization specialists contributed their institutional knowledge of payer-specific requirements, which supplemented QuickAuth's baseline database.

Month 5-6: Active authorization automation. QuickAuth was activated for all cardiac imaging orders. Authorization requests were generated automatically at the point of scheduling, with the auth specialists reviewing and approving submissions rather than manually preparing them.

Month 7: Full integration. QuickAuth was integrated with the scheduling system to prevent studies from being scheduled without active authorization. The system also began tracking authorization utilization — identifying authorized studies that hadn't been performed before the authorization expired.

Phase 2 results (month 7): Authorization-related denials dropped from 11.2% of all denials to under 2%. Auth turnaround time from 3.8 days average to 6 hours average.

Phase 3: Optimization and Continuous Learning (Months 7-12)

The final phase focused on model refinement, denial pattern analysis, and staff workflow optimization.

QuickCode's accuracy continued to improve as it ingested payer responses, reaching 96% by month 10. The system identified three previously undetected patterns: one commercial payer was systematically denying modifier 59 on echo/Doppler combinations and accepting modifier XS for identical claims; a Medicare Advantage plan was applying a proprietary bundling edit to EP study codes that differed from CMS guidelines; and another commercial payer had changed its left heart catheterization coding requirements without publishing a provider notice.

Phase 3 results (month 12): Denial rate stabilized at 8.4%. Coding accuracy at 96%. Auth-related denials under 1%.

Results: The Full Impact After 12 Months

Financial Recovery

$1.8 million in annual recovered revenue came from four sources:

• $920K from prevented coding denials. Claims that would have been denied for modifier errors, bundling mistakes, or coding specificity issues were corrected before submission by QuickCode. This figure was calculated by comparing the group's historical denial rate to its current denial rate and applying the average claim value.
• $360K from eliminated auth-related denials. Prior authorization gaps that previously generated guaranteed denials were closed by QuickAuth's automated determination and submission process.
• $310K from faster denial resolution. Denials that still occurred were resolved in 11 days on average instead of 38, reducing timely filing write-offs and improving appeal success rates from 44% to 71%.
• $210K from optimized code sequencing. QuickCode's reimbursement optimization — sequencing codes and modifiers for maximum allowable reimbursement within proper coding guidelines — recovered revenue from claims that were previously paid but underpaid due to suboptimal coding.

Coding Accuracy: 78% to 96%

The 18-percentage-point improvement in coding accuracy had cascading benefits beyond denial reduction.

Catheterization coding accuracy: from 69% to 97.1%. The most dramatic improvement was in the most complex procedure category. QuickCode's catheterization engine eliminated the bundling errors and modifier omissions that had made cath lab claims the group's highest-risk category. The senior coder, who previously spent 40% of her time on catheterization coding, now spent 15% — reviewing AI-generated codes rather than navigating the catheterization coding matrix manually.

Echo coding accuracy: from 82% to 95.8%. QuickCode's ability to distinguish complete from limited studies based on documentation content, and to verify Doppler medical necessity, reduced echo-related denials by 74%. Color flow Doppler (93325) denial rates dropped from 17% to 3.2%.

Modifier accuracy: from 71% to 96.4%. Across all procedure types, the modifier error rate dropped from 29% to 3.6%. The improvement was most pronounced in 26/TC split decisions and multiple procedure modifier application — the two modifier categories where payer-specific variation created the most confusion.

Operational Efficiency

Three denial management FTEs redeployed. The denial management workload dropped by approximately 70%, allowing 3 of the 5 denial-focused staff to be reassigned. One moved to patient financial counseling, one to payer contract analysis (a role the group had never staffed before), and one to a quality assurance position reviewing AI-generated codes and identifying documentation improvement opportunities.

Coder productivity increased 85%. With QuickCode generating initial codes, the three coders transitioned to a review-and-approve workflow that allowed them to process 60% more encounters per day. The group did not reduce coding headcount but instead eliminated the backlog that had been a persistent problem — claims were now coded within 24 hours of the encounter rather than the 72-96 hour lag that had been standard.

Prior auth specialist workload reduced 75%. The two prior auth specialists went from managing 280 manual auth requests per month to overseeing QuickAuth's automated submissions and handling only the cases requiring clinical narrative or peer-to-peer review — approximately 70 cases per month. One auth specialist was reassigned to patient scheduling coordination; the other remained in the auth role handling complex cases and payer escalations.

Denial Resolution Acceleration

Metric	Before	After
Average denial resolution time	38 days	11 days
Appeal success rate	44%	71%
Timely filing write-offs	$340K/year	$48K/year
Denials written off (% of total)	19%	6.2%

QuickRCM's appeal automation generated payer-specific appeal packages with supporting documentation and citations to applicable coding guidelines, reducing appeal preparation from 42 minutes to 9 minutes.

Return on Investment

The group's total investment in QuickCode and QuickAuth — including software licensing, implementation services, EHR integration, and staff training — was approximately $310,000 in the first year, with ongoing annual costs of approximately $215,000.

Against $1.8 million in recovered revenue and approximately $380,000 in labor savings from redeployed FTEs, the first-year ROI was approximately 600%. Ongoing annual ROI exceeded 900%.

Key Takeaways for Cardiology Groups

1. Cardiac Catheterization Coding Is the Highest-Risk, Highest-Reward Target

Catheterization coding is where cardiology groups lose the most revenue to denials, and it's where AI coding delivers the most dramatic improvement. The bundling complexity, modifier requirements, and payer variation in cath coding exceed what human coders can reliably manage at volume. Groups performing 1,000+ catheterizations annually should consider cath-specific AI coding a financial priority, not a technology experiment.

2. Modifier Errors Are Systematic, Not Random

The group's 29% modifier error rate was not caused by careless coders. It was caused by a modifier system that requires payer-specific knowledge across dozens of payers, with rules that change quarterly. Systematic problems require systematic solutions — training individual coders to memorize payer modifier requirements is not a sustainable strategy when the requirements themselves are a moving target.

3. Prior Auth for Cardiac Imaging Is a Solved Problem

Cardiac imaging authorization requirements are well-defined and relatively standardized across payers compared to other authorization categories. The clinical criteria (appropriate use criteria for cardiac imaging) are published and measurable. This makes cardiac imaging auth an ideal automation target — the rules are clear enough for AI to apply reliably, and the volume is high enough to make the automation economically justified.

4. Coding Accuracy Drives Revenue Even When Claims Are Paid

The group discovered that coding accuracy wasn't just about preventing denials — it was about capturing the correct reimbursement on claims that were paid. Suboptimal code sequencing, missed add-on codes, and conservative coding that didn't reflect the work performed were costing the group an estimated $210,000 annually in underpayments on claims that were never denied. Improving coding accuracy from 78% to 96% meant recovering revenue from claims that the group didn't even know it was losing.

5. Experienced Cardiology Coders Are More Valuable in QA Than in Production

The group's most experienced coder was spending her time navigating catheterization bundling logic — work that AI performed with higher accuracy and consistency. When she transitioned to quality assurance, reviewing AI-generated codes and identifying documentation improvement opportunities, her specialty expertise had a multiplied impact. She caught edge cases that the AI missed, trained the AI on nuanced clinical scenarios, and worked with the cardiologists to improve documentation practices that had been generating coding ambiguity for years. The AI handled volume; the expert handled judgment.

---

This case study presents representative outcomes based on aggregate customer data from cardiology groups using the QuickIntell platform. Individual results depend on practice size, procedure mix, payer mix, and baseline coding accuracy. To discuss how these results might apply to your organization, contact QuickIntell for a custom analysis.