Prashant Sharma

Mr. Prashant Sharma

orcid

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 October 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
PSPrashant Sharma
Prashant Sharma
Background: Multi-payer claims reconciliation in the United States healthcare system is a complex operational problem because eligibility, coordination-of-benefits (COB) rules, adjudication logic, remittance formats, and payment responsibilities vary across payer environments. These differences can create duplicate payments, crossover-processing failures, reconciliation discrepancies, and audit risk. Objective: This paper presents the Multi-Payer Claims Reconciliation Architecture (MPCRA), an original five-layer framework for AI-driven, near-real-time claims reconciliation across heterogeneous healthcare reimbursement environments. Methods: MPCRA was designed as a payer-agnostic architecture spanning eligibility and coverage harmonization, payer sequencing, cross-payer adjudication and payment modeling, discrepancy detection and resolution, and adaptive audit/compliance reporting. The evidence basis includes aggregated operational observations from U.S. Federal and State healthcare data programs during 2019-2025 in which MPCRA architectural principles were applied. The reported outcomes are observational and are not derived from a controlled experimental study. Results: Across the identified operational environments, observed improvement ranges were 38-61% for COB error reduction, 72-89% for duplicate-payment elimination, 41-67% for crossover-claims MPCRA Preprint v2.3-October 2026 processing efficiency, 55-78% for remittance reconciliation accuracy, 79-93% for real-time adjudication throughput, and 88-96% for regulatory audit compliance. These ranges should be interpreted as programlevel observations associated with application of MPCRA principles rather than as causal estimates attributable exclusively to MPCRA. Conclusion: MPCRA provides a unified architectural model for integrating eligibility harmonization, payer sequencing, payment prediction, discrepancy resolution, and adaptive compliance feedback. The framework is intended as a foundation for future empirical evaluation and implementation across Medicaid, Medicare, ACA Marketplace, commercial, and pharmacy-benefit environments.

29 July 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
PSPrashant Sharma
Prashant Sharma
Background: Government agencies administering federally funded benefits programs, including Medicaid, ACA Marketplace plans, and vocational rehabilitation services, maintain independent data systems with heterogeneous client identifier schemes, eligibility rule sets, and governance frameworks. The absence of unified reconciliation architecture generates systemic eligibility verification failures, improper payments, and recurring federal audit findings across programs serving millions of beneficiaries. Objective: This paper presents the Cross-Agency Benefits Eligibility Reconciliation Model (CABERM), an original conceptual architecture for AI-driven, real-time eligibility reconciliation across independent government data systems, developed through applied engagement with largescale state government benefits programs. Methods: CABERM introduces a five-layer architectural model integrating: (1) multi-system client identity resolution, (2) cross-agency eligibility data harmonization, (3) AI-driven eligibility rule adjudication, (4) real-time reconciliation and discrepancy resolution, and (5) federal compliance audit trail generation. The architecture is described at a conceptual level to support broad applicability and peer evaluation. Results: Programs guided by CABERM architectural principles demonstrated: 89-97% crossagency identity resolution accuracy; 44-68% eligibility determination error reduction; 61-79% case processing time reduction; 73-88% manual verification elimination; 52-74% cross-agency data consistency improvement; and 91-98% federal RSA/CMS audit compliance. Conclusion: CABERM provides a scalable, program-agnostic framework for proactive, AIdriven cross-agency eligibility reconciliation applicable across Medicaid, ACA Marketplace, vocational rehabilitation, and all federally funded benefits programs requiring cross-system eligibility coordination.

30 June 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
PSPrashant Sharma
Prashant Sharma
Background: Modern U.S. healthcare data environments span distributed multi-payer ecosystems, including Medicaid, Medicare, ACA Marketplace, and commercial payer systems, generating hundreds of millions of claims transactions annually. Ensuring data quality across these heterogeneous, high-velocity pipelines remains a critical unresolved challenge. Traditional rule-based EDI validation tools lack the semantic, temporal, and contextual intelligence required to govern data integrity at enterprise scale. Objective: This paper presents the Healthcare Data Quality System (HDQS), an original conceptual architecture for AI-driven, real-time healthcare data quality enforcement. HDQS introduces a six-layer architectural model integrating payer-aware metadata harmonization, ML-driven anomaly intelligence, multi-dimensional validation, temporal claims consistency assessment, and adaptive autonomous remediation. Methods: HDQS was designed through iterative applied research across multiple U.S. Federal and State healthcare data programs serving millions of patients. The architecture is described at a conceptual level, avoiding implementation-specific details, to support broad applicability and peer evaluation. Each architectural layer addresses a distinct class of healthcare data quality challenge identified through program-level operational experience. Results: Environments guided by HDQS architectural principles reported: 70-90% reduction in anomaly detection latency; 50-87% reduction in downstream pipeline incidents; 30-50% improvement in CMS regulatory audit preparation efficiency; and 20-40% improvement in healthcare AI model stability. Cross-payer reporting reconciliation cycles improved by 35-60%. Conclusion: HDQS provides a scalable, payer-aware, and temporally intelligent architectural blueprint for healthcare data quality governance. Its conceptual contributions advance the field beyond static EDI format validation toward dynamic, intelligence-driven clinical data integrity management, with direct applicability across Medicaid, Medicare, ACA Marketplace, pharmacy, and all-payer data environments.

17 November 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
PSPrashant Sharma
Prashant Sharma
Background: Cross-payer member identity resolution is foundational to healthcare interoperability and a persistent barrier to coordinated care. Duplicate and fragmented member records impair care coordination, administrative efficiency, and population health analytics across Medicaid, Medicare, and commercial payer environments. The growing adoption of Payer-to-Payer data exchange under CMS-0057-F and the Trusted Exchange Framework and Common Agreement (TEFCA) intensifies the need for governed, auditable member identity architectures. Objective: To specify AIM-HIM, a governed master data management architecture for healthcare member identity resolution and golden-record creation, and assess selected components through synthetic proof-of-concept experiments. Methods: The five-layer design combines ingestion, standardization, identity resolution, attribute survivorship, and governance. The architecture reflects patterns researched and implemented during professional practice as a Data Architect and Analyst (2017–2024), with AI enhancement since 2024. A linkage experiment used 1,200 synthetic persons with 480 held-out identities; a separate survivorship experiment evaluated 24,000 attribute decisions across five simulation seeds. Results: The hybrid matcher achieved precision 0.998, recall 0.944, and F1 0.970. Ungated logistic regression achieved F1 0.976; the hybrid's lower recall reflects a deliberate design tradeoff—routing DOB-contradictory pairs to human stewardship rather than automatic acceptance—at the cost of recall in this synthetic scenario. Confidence-weighted survivorship achieved 97.02% accuracy versus 89.50% and 92.11% for two fixed-priority baselines, conditional on correct identity links and an informative simulated quality signal. Discussion: The experiments illustrate component feasibility under simplified assumptions. Conclusion: AIM-HIM provides a testable architecture and evaluation protocol. Independent replication, stronger baselines, realistic benchmarks, and authorized real-world validation will further strengthen the evidence base.