Assessing Total Hip Arthroplasty Outcomes and Generating an Orthopedic Research Outcome Database via a Natural...

Area of Science:

• Orthopedic Surgery
• Health Informatics
• Artificial Intelligence

Background:

• Electronic health records (EHRs) are complex and fragmented across multiple care sites.
• Extracting research-grade data from EHRs is time-consuming and costly.
• Existing data systems limit clinical practice insights.

Purpose of the Study:

• To develop an automated natural language processing (NLP) pipeline for extracting orthopedic clinical concepts from EHRs.
• To assess therapies and complications following total hip arthroplasty (THA).
• To create a unified clinical data warehouse from multicenter EHR data.

Main Methods:

• An AI-based NLP and machine learning pipeline processed EHRs from 1290 primary THA patients.
• A gold standard was established by 3 independent medical reviewers using 100 EHRs.
• The pipeline aggregated data from diverse EHR systems for analysis.

Main Results:

• The NLP pipeline achieved high accuracy, comparable to human reviewers (F1-score 0.92).
• 92.1% of patients experienced a complication-free recovery within the first year post-THA.
• Common early complications included lateral femoral cutaneous nerve sensitivity and intraoperative fractures.

Conclusions:

• Automated NLP effectively extracts clinical information from EHRs, rivaling human accuracy.
• This technology offers significant value for surgeon practice and patient counseling.
• The study supports a low early complication rate for direct anterior THA.