A large, multi-specialty provider group with 150+ providers across diverse locations — managing high volumes of outpatient charts with a coding team that was meeting throughput targets but producing inconsistent results.
On paper, the client’s coding operation looked functional. Charts were being coded. Claims were going out. But the numbers underneath told a different story.
Coding-related denials were running well above benchmark — and when the client dug in, the root causes were varied: incorrect ICD-10 diagnosis assignments, misapplied modifiers, and inconsistent E/M leveling across providers and coders. No single issue dominated. It was a pattern of small inaccuracies compounding into significant revenue impact.
The more troubling finding came from a chart audit: billable CPT codes were being missed on a meaningful percentage of encounters — procedures that were performed, documented, and never billed. The client’s coding director estimated the revenue impact but couldn’t get the existing team to close the gap at scale. It wasn’t a training problem — it was a volume and consistency problem that manual coding couldn’t solve.
NDS started with a data analysis phase that proved critical. The client shared four weeks of historical charts with corresponding billing codes. NDS processed these through its baseline AI model and compared outputs against the manually coded data. The divergences revealed patterns the client hadn’t fully quantified — particularly around missed CPT codes and systematic E/M under-coding on certain specialties.
These divergences were used to build client-specific business rules and calibrate the AI before it touched a live chart. No major system changes were required.
The AI — using clinically trained generative AI and clinical NLP — evaluated complete patient encounters, interpreted provider intent, and assigned ICD-10-CM, CPT, and modifier combinations based on documentation content, AMA-licensed reference data, and built-in NCCI edits.
The capability that proved most valuable in production was one NDS hadn’t originally emphasized in the proposal: Dynamic Denial Prevention. NDS analyzed the client’s payer 835 ERA data to identify which specific coding behaviors were driving denials — then built customized validation rules that flagged those patterns before claim submission. This closed the loop between coding and payer adjudication in a way the client’s manual QA process never could.
A governed exception handling workflow routed low-confidence charts to certified medical coders. Their corrections fed back into the system through reinforcement learning. Over the first 90 days, the volume of charts requiring human review dropped meaningfully as AI accuracy improved.
DNFB was also elevated. Not because coders were slow, but because the review-and-correction cycle on complex charts created a tail that delayed claim submission by days.
Single specialty pilot on the client's highest-volume specialty. One early issue: the AI initially over-coded E/M levels on a subset of established patient visits. The validation rules caught this before claims were submitted, and the model was adjusted within two weeks.
Expanded to five additional specialties with specialty-specific rule configurations and ongoing quality audits.
Full adoption across all relevant specialties. Complex cases continued routing to human review.
Increase in captured revenue — the headline number, driven primarily by CPT codes that were previously being missed
Reduction in coding-related denials — improved accuracy and Dynamic Denial Prevention rules working together
Reduction in DNFB — faster throughput on routine charts freed up the review cycle
Reduction in coding costs — though the client noted this took a full quarter to realize as they redeployed staff
Unexpected outcome: The AI surfaced a pattern of modifier misapplication on a specific payer that had gone undetected for over a year. The client's billing team used this to recover underpayments on previously adjudicated claims — a recovery effort that wasn't part of the original project scope.
What the Client Said
"The denial reduction was what we signed up for. The revenue we found from missed CPT codes was what changed the conversation with our board. We didn't know how much we were leaving on the table until NDS showed us the data."
Discover the way forward through more case studies.
Share your sample charts and your organization’s business rules, and we’ll build a proof of concept showing how NDS handles your documentation — from clinical analysis to compliant code assignment — tailored to your specialties and payer mix.
© 2026 NDS InfoServ. All rights reserved.