A growing strategy among insurance companies, which is increasingly mirrored in Medicare recovery efforts, is the systematic auditing of old charts, often years after care was delivered and paid, in search of technical documentation gaps that can justify clawbacks. Artificial intelligence has supercharged this approach, allowing rapid mining of historical records against today’s evolving standards rather than the clinical and documentation norms that existed at the time the care was provided.
Spinal cord stimulation (SCS) offers a clear and consequential example of how this broader problem manifests and continues to face targeted post-payment review under Recovery Audit Contractor (RAC) Topic 0207 (“Spinal Cord Neurostimulation: Medical Necessity and Documentation Requirements”). Reviewers apply the longstanding National Coverage Determination (NCD) 160.7 criteria: The therapy must represent a late or last resort after documented failure of conservative care, a psychological evaluation must be present, and a successful trial (at least 50 percent pain relief with functional improvement) must be clearly recorded before permanent implantation. Complex medical-record reviews examine the completeness of prior therapies, the exact language of trial outcome notes, consistency between operative reports and progress notes, and alignment with any applicable Local Coverage Determination.
Clinical evidence supporting SCS value
As mentioned above, the high-quality data demonstrate that properly selected patients obtain meaningful, durable benefit:
- The EVOKE randomized controlled trial of evoked compound action potential (ECAP)-controlled closed-loop SCS showed, at 36 months, 77.6 percent of patients achieving at least a 50 percent reduction in overall back and leg pain (versus 49.3 percent with open-loop) and 49.3 percent achieving at least an 80 percent reduction. Importantly, there were zero explants due to loss of efficacy in the closed-loop arm.
- Multiple systematic reviews and network meta-analyses confirm that both conventional and novel SCS modalities produce superior pain reduction, functional improvement, and quality-of-life gains compared with conventional medical management.
- Successful trials convert to permanent implants at high rates (typically 72 to 94 percent in well-selected cohorts), and a substantial proportion of responders maintain clinically meaningful relief at long-term follow-up.
These outcomes matter. When a patient who achieved sustained relief later generates a clawback demand based on a documentation technicality, the financial hit lands on the practice that delivered evidence-based care. Many patients achieve sustained relief that reduces medication reliance and restores daily function. Yet practices can still face overpayment demands long after the patient has already benefited just because an algorithm flagged a missing phrase about prior therapies, an incomplete sentence in a psychological evaluation, or a documentation element that did not exist in the same form when the care was planned, authorized, and delivered.
Real physician voices, twice a week
Free, and one click to unsubscribe.
Because these cases are high-dollar, they are repeatedly selected for additional documentation requests. A large share of alleged “improper payments” in Medicare and Medicaid audits are driven by documentation deficiencies rather than true absence of medical necessity. When AI tools are layered onto this process, whether for real-time claim mining or retrospective chart review, the system can flag technical gaps years after the care was delivered and paid. This is not isolated to neuromodulation. The same retrospective playbook is applied across high-value services: Flag historical claims, apply current or algorithmically derived standards, and recover money on technicalities rather than true absence of medical necessity. It occurs against a backdrop of ongoing payment rate cuts that already strain the economic viability of interventional and specialty practices.
Problems with AI-driven retrospective review
Evidence from related AI-assisted coverage processes raises serious concerns about accuracy and fairness:
- Studies of Medicare Advantage prior-authorization denials show overturn rates of approximately 80 to 82 percent when appealed, indicating that the large majority of challenged denials were ultimately judged incorrect.
- Early experience with CMS’s Wasteful and Inappropriate Service Reduction (WISeR) AI-assisted prior-authorization model has shown elevated denial rates compared with traditional processes.
- The majority of improper-payment findings across Medicare and Medicaid programs stem from insufficient documentation rather than fraud or clear lack of medical necessity.
- Physician surveys (including Medical Group Management Association [MGMA] data) consistently rank audits, appeals, and documentation demands among the top regulatory burdens, with strong associations to burnout and diversion of clinical resources.
When algorithms trained on current documentation expectations are applied retrospectively to care delivered under earlier norms, the result is often a technical “gotcha” rather than a genuine determination that the therapy was never medically necessary.
Impactful statement for physicians
Using artificial intelligence to mine years-old charts and claw back payment for spinal cord stimulation is not program integrity; it is retrospective second-guessing that punishes evidence-based care. This practice fails basic fairness. Physicians document in real time under the guidance and clinical standards then in force. They cannot time-travel to satisfy tomorrow’s template. When 80 percent of appealed Medicare Advantage denials are overturned, and when the majority of alleged improper payments are documentation-related rather than true lack of medical necessity, the system has stopped distinguishing between waste and the imperfect reality of clinical records. Contingency-fee incentives and look-back periods amplify the problem, converting legitimate high-value care into a perpetual financial risk. The human cost is substantial, and this process is a documented driver of physician burnout. Regulatory burden surveys, including those from the MGMA, rank audits, appeals, and documentation demands among the top stressors facing practices. Excessive administrative work is a leading driver of physician burnout, frequently ranking above financial pressure and staffing shortages.
The real-world consequence is a chilling effect. Practices become more hesitant to offer advanced neuromodulation to complex patients. Administrative resources that should support patient care are consumed by appeals. And patients who already improved are left as collateral damage in a process optimized for recovery rather than clinical truth.
True integrity requires prospective clarity, contemporaneous human clinical judgment, and standards that do not change after the fact. Weaponizing AI against historical charts to extract money from care that met the rules of its time, and that produced measurable patient benefit, is an abuse of authority that ultimately undermines the very patients the program claims to protect. Yet under current RAC processes and expanding AI-assisted review, a practice can still face an overpayment demand years after the patient has already benefited just because an algorithm flagged a missing phrase about prior therapies, an incomplete sentence in a psychological evaluation, or a documentation element that did not exist in the same form when the care was planned, authorized, and delivered.
This occurs against a backdrop of ongoing Medicare payment rate cuts that already squeeze the economic viability of interventional pain practices.
Kayvan Haddadan is a physiatrist and pain management physician, and president and medical director of Advanced Pain Diagnostic & Solutions, a multidisciplinary pain management practice in California that he founded in 2012. A physician and surgeon licensed by the Medical Board of California, he is double board-certified in pain medicine and physical medicine and rehabilitation. He is also certified in controlled substance registration through the DEA and serves as a qualified medical examiner through California’s Department of Industrial Relations Division of Workers’ Compensation.
Dr. Haddadan earned his Bachelor of Science degree from the College of Alborz in Tehran, Iran, and his medical degree from Shahid Beheshti University of Medical Sciences. He later received his Educational Commission for Foreign Medical Graduates certification in Philadelphia, completed an internship in medical surgery at Loyola University Medical Center’s Stritch School of Medicine in Illinois, and finished his residency in physical medicine and rehabilitation at the same institution. He completed his fellowship in pain medicine at California Pacific Medical Center’s Pacific Pain Treatment Center and also trained in medical acupuncture for physicians at the University of California, Los Angeles David Geffen School of Medicine.
Dr. Haddadan has contributed to 29 research publications across multiple specialties, including pain management, cardiology, pulmonology, endocrinology, gastroenterology, and infectious disease. His work has examined topics such as hyperlipidemia in high cardiovascular risk patients, hyperuricemia and gout management, type 2 diabetes and hypertension, chronic obstructive pulmonary disease and asthma therapies, influenza treatment, irritable bowel syndrome, and opioid related complications in chronic pain care. His research has also included clinical outcome studies in spinal cord stimulation and award-winning presentations on neuropathic pain management and neuromuscular disorders.
