AI-WAR: Warfarin Dosing Trial Data with Bi-LSTM Model Performance

624 patients participated in a randomized controlled trial to evaluate an AI-based warfarin management system. The AI-WAR software, developed by Yuan Li, integrates remote follow-up and a bidirectional LSTM dosing model. It significantly improved median Time in Therapeutic Range from 48.7% to 81.3% and reduced adverse event rates.

Use Cases

• Benchmark AI dosing models based on reported performance metrics like 80.3% dose prediction accuracy
• Study anticoagulation quality improvement using metrics such as Time in Therapeutic Range (TTR)
• Analyze subgroup performance of dosing models across different genotypes as mentioned in the results
• Evaluate the impact of remote management systems on clinical outcomes in primary care settings

Strengths

• Includes data from a randomized controlled trial with 624 participants and an external validation cohort of 200 patients
• Model performance is quantified with specific metrics: 81.3% median TTR and 93.2% stable dose prediction accuracy
• Results include subgroup analyses demonstrating performance across genotypes and with missing genetic information

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• The dataset is a 64.1 KB DOCX file, suggesting it may be a supplementary document rather than raw structured data

Provenance

Source

figshare

Collection Method

Data collected from a randomized controlled trial and a prospective registry study for model training, with an independent real-world cohort for validation.

Freshness

Last updated 2026-04-29 05:57:17; freshness should be verified