
As healthcare providers adopt Artificial Intelligence (AI) to automate documentation, payers are simultaneously deploying AI to automate claim adjudications and denials. This article examines the resulting "co-evolutionary" arms race between provider and payer algorithms. We identify the emergence of "denial inflation," where the near-zero cost of AI-driven denials incentivizes payers to challenge more claims, and the "automation divide," which disproportionately penalizes small, under-resourced practices unable to afford advanced defensive AI. While the eventual endpoint of a "perfect note" system may eliminate administrative waste, the transition period threatens financial stability for many providers. This article argues that algorithmic transparency is essential to minimize market casualties and ensure this technological convergence benefits patients rather than merely escalating administrative conflict.
Objective: To evaluate a natural language processing (NLP) framework for harmonizing insurance terminologies across multiple payers, improving accuracy and efficiency in claims data integration, and improving analytics for value-based contracting. Study Design: Claims data dictionaries from major U.S. insurers were aligned using an NLP-based approach and compared with manual review and rule-based matching. The Medicare dictionary served as the reference standard. Methods: The framework included (1) automated preprocessing to standardize terms, (2) semantic alignment using advanced language models, and (3) expert verification. Performance was assessed using precision, recall, and F1 score, as well as processing time and cost estimates. Results: The NLP approach achieved an F1 score of 0.92 (95% CI, 0.90–0.94) versus 0.78 (95% CI, 0.74–0.82) for manual matching (p < 0.001). Precision and recall were 0.93 and 0.91, respectively. Processing time per 1,000 terms decreased from approximately 14 hours (manual) to less than 10 minutes with NLP, resulting in a reduction of estimated costs by over 40%. Conclusions: NLP-driven terminology alignment provides substantial operational and financial benefits, enabling faster and more accurate claims integration. Adopting this approach can support value-based contracting, improve reimbursement analysis, and enhance strategic decision-making for healthcare systems.
Background: In the United States, rising prescription drug costs are a growing concern for policymakers. In response, several states have established Prescription Drug Affordability Boards (PDABs) to evaluate high-cost drugs, assess their affordability for patients and payers, and, where authorized, establish upper payment limits (UPLs). Objectives: To characterize statutory and governance differences across PDABs, examine how each state defines and operationalizes “affordability” in policy and law, and synthesize quantitative data on drugs selected for affordability review or UPL consideration. Methods: We identified PBABs that were active as of October 24, 2025. We included state entities that had the authority to conduct drug affordability reviews or use strategies such as UPLs. We excluded entities that functioned exclusively as price transparency or data-reporting commissions. For each state, we reviewed enabling statutes and publicly released data files. We extracted data on statutory design, governance structure, affordability definitions, and drugs identified for affordability review, and summarized cross-state similarities and differences. Results: We identified 5 PDABs in Maryland, Colorado, Oregon, Washington, and Minnesota. Across states, boards required expertise in clinical medicine or healthcare economics. Drugs were identified for affordability reviews based on wholesale acquisition costs (WAC) thresholds, such as WAC increases of $300 or more or 200% or more in the prior 12 months, or an annual WAC of $30,000 or more. For biosimilars, 4 PDABs considered whether they were launched at a 15% or 20% discount relative to the reference biologic. Drugs that progressed to full review were high-cost specialty therapies with high per-patient spending or rapid price growth, such as Enbrel. Conclusion: PDABs share common structural features across states, but the extent of implementation differs. Although the long-term impact of PDABs on payer spending is uncertain, variation in transparency and prioritization across boards may influence how these policies evolve over time.
Objective: Primary care practices have attracted significant private capital investment, and prior studies in fee-for-service (FFS) settings link investment to higher costs. We conducted the first study to examine how cost trajectories differ before and after external investment in a value-based care (VBC) environment, with particular attention to selection dynamics into investor ownership. Ours is a correlational study with selection into PE ownership as the core mechanism. Methods: We analyzed 98,014 Medicare Advantage patient-year claims over a six-year period (2018–2024) from investor-backed primary care platforms using fixed effects and event-study models with controls for demographics, comorbidity, geography, payer, preventable conditions, time, and practice effects. Results: Average post-investment costs were statistically lower. Event-study analysis showed modest pre-investment cost declines and statistically significant additional cost reductions emerging three to four years after acquisition. This suggests that selection explains part of the observed association between investment and cost declines, and benefits from post-investment capability changes may manifest in the longer term. A temporary cost increase occurred during the investment year, likely due to amplified outreach. Unlike FFS, external investment did not reverse the downward trend. Conclusions: Findings suggest that investors may be selecting high-performing practices already engaged in cost-reduction strategies, scaling existing and perhaps bringing new capabilities to drive further cost reductions after 3-4 years. Policy makers should distinguish FFS from VBC and consider how private capital might help scale VBC models.
Hospitals are important health and economic entities their communities. Beyond the direct economic impact of hospitals that derives from the wages paid to employees, there are both indirect employment effects and induced effects of employment and cooperative arrangements. An input-output model was applied to data available for a critical access hospital in Louisiana. In fiscal year 2022, Lallie Kemp Regional Medical Center had 319 employees with earnings of $19.6 million and non-labor expenditures of $27.4 million. Including induced effects, the economic impact was over $85 million. Cooperative arrangements, notably a prescription drug program, provided an economic benefit to the community of nearly $15 million. The sum of $100 million for a small critical access hospital is a substantial amount. Applying this model to other hospitals is feasible with the general availability of data, though there are many assumptions implied by an input-output model that merit attention.