A data and knowledge cross-level fusion-driven learning framework for detecting missing diagnosis.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Documentation Improvement

Background:

• Diagnosis omission is prevalent in electronic medical records (EMRs), leading to inaccurate documentation.
• This inaccuracy impacts Diagnosis Related Group (DRG) assignments and financial reimbursements due to missed Complications and Comorbidities (CC) or Major Complications and Comorbidities (MCC).

Purpose of the Study:

• To develop and evaluate a data and knowledge cross-level fusion-driven learning framework for automated identification of missed diagnoses in EMRs.
• To assess the impact of missed diagnoses on DRG assignments and insurance reimbursements.

Main Methods:

• A novel data and knowledge cross-level fusion-driven learning framework was proposed for automated missed diagnosis identification.
• The model was evaluated on real-world EMR data from six Chinese hospitals, comparing its performance against expert systems, BERT, and other large language models (LLMs).
• A hybrid approach combining the proposed model with an expert system was used to minimize alert fatigue and improve precision.

Main Results:

• The proposed model demonstrated superior F1 scores compared to baseline methods in identifying missed diagnoses.
• Approximately 37.8% of EMRs were predicted to have missed diagnoses, affecting 9.0% of DRG groupings and 3.2% of insurance reimbursements.
• The hybrid approach boosted precision by 6.7-13.4%, with two human-machine coupling modes designed for practical application.

Conclusions:

• The fusion-driven learning framework effectively identifies missed diagnoses in EMRs, enhancing documentation accuracy and financial outcomes.
• Combining the model with expert systems offers a practical solution to mitigate alert fatigue and improve clinical workflow efficiency.
• The study highlights the significant impact of diagnosis omission and the potential of AI in improving healthcare documentation and reimbursement processes.