Scholarlink

This research proposes an innovative end-to-end predictive analytics pipeline for automated decision support, which significantly improves the workflow for preauthorization in healthcare systems. Incorporating structured claims data and unstructured content from electronic medical records (EMRs) enables our system to outperform traditional single-modality systems and commercial solutions with 89.4% accuracy, 88.7% precision, and 94.7% AUC-ROC in predictive performance versus the other system’s 83.6% accuracy and 82.9% precision. We employ a hybrid architecture that integrates gradient boosting approaches for tabular claims data with transformer-based natural language processing for clinical narratives. The system used exhibits extraordinary operational efficiency gains in terms of preauthorization decision time reduction by 73% from 4.2 days to 1.1 days, 42% reduction in administrative costs from $743 to $431 per case, and sustaining 89.4% agreement with expert clinical adjudication. This thorough evaluation using a multi-institutional dataset from five healthcare systems, consisting of 157,346 participants, demonstrates that the proposed pipeline is capable of effective generalization across varying patient populations and clinical scenarios. Key contributions include: A novel hybrid architecture that fuses structured claims and unstructured EMR data; A cross-system analysis in five disparate healthcare systems; Two new metrics—alignment score for the automation of preauthorization decisions against guideline-driven care automation frameworks, and clinical context alignment score for automated preauthorization decisions relative to evidence-based guidelines—to gauge the level of automation versus guideline integration; and Operational enhancement of workflow preauthorization demonstrated by reduced time and cost. This research demonstrates the operational efficiency and clinical appropriateness of using a payor-driven hybrid data architecture which combines unabridged unmanaged clinical data and structured claims data while dynamically adjusting administrative controls to clinical access frameworks within a scalable model for healthcare payors and providers.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

End-to-End Predictive Analytics Pipeline for Preauthorization Outcomes Using Claims and EMR Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rakesh Ramakrishna Pai,
  • Jothsna Praveena Pendyala

摘要

This research proposes an innovative end-to-end predictive analytics pipeline for automated decision support, which significantly improves the workflow for preauthorization in healthcare systems. Incorporating structured claims data and unstructured content from electronic medical records (EMRs) enables our system to outperform traditional single-modality systems and commercial solutions with 89.4% accuracy, 88.7% precision, and 94.7% AUC-ROC in predictive performance versus the other system’s 83.6% accuracy and 82.9% precision. We employ a hybrid architecture that integrates gradient boosting approaches for tabular claims data with transformer-based natural language processing for clinical narratives. The system used exhibits extraordinary operational efficiency gains in terms of preauthorization decision time reduction by 73% from 4.2 days to 1.1 days, 42% reduction in administrative costs from $743 to $431 per case, and sustaining 89.4% agreement with expert clinical adjudication. This thorough evaluation using a multi-institutional dataset from five healthcare systems, consisting of 157,346 participants, demonstrates that the proposed pipeline is capable of effective generalization across varying patient populations and clinical scenarios. Key contributions include: A novel hybrid architecture that fuses structured claims and unstructured EMR data; A cross-system analysis in five disparate healthcare systems; Two new metrics—alignment score for the automation of preauthorization decisions against guideline-driven care automation frameworks, and clinical context alignment score for automated preauthorization decisions relative to evidence-based guidelines—to gauge the level of automation versus guideline integration; and Operational enhancement of workflow preauthorization demonstrated by reduced time and cost. This research demonstrates the operational efficiency and clinical appropriateness of using a payor-driven hybrid data architecture which combines unabridged unmanaged clinical data and structured claims data while dynamically adjusting administrative controls to clinical access frameworks within a scalable model for healthcare payors and providers.