November 25, 2019PLoS ONEOpen Access

Semantic computational analysis of anticoagulation use in atrial fibrillation from real world data

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Atrial fibrillation (n=10030). Intervention: Oral anticoagulants (OAC) vs. Antiplatelet only or no antithrombotic medication. Primary outcome: OAC prescription rate in high-risk AF patients (CHA2DS2-VASc ≥ 2).

View Full Paper

Ask AI

Bookmark

Key result

OAC use in high-risk AF climbs to 62%, driven by DOAC availability.

n=10,030

Why the study?

The study aimed to develop and validate an open-source natural language processing pipeline to calculate stroke and bleeding risk scores from free-text electronic health record data in atrial fibrillation.

Does a natural language processing pipeline accurately calculate CHA2DS2-VASc and HAS-BLED scores from free-text electronic health records in patients with atrial fibrillation compared to expert manual scoring?

Population

10,030 AF patients identified from discharge summaries

Comparison

Automated NLP-derived risk scores vs manual scoring by two independent experts in 40 patients

Design

Validation and retrospective electronic health record cohort study

Authors

DBDaniel BeanKing's College London JTJames TeoKing's College London HWHonghan WuZunyi Medical University

Discussion

Member takes

Implication

Rising OAC use in high-risk AF may aid stroke prevention; extends EHR-based analyses but leaves causal effects open.

Study Design

Type

Observational (n=10,030)

Multicenter

Structured PICO

Population

10,030 patients with atrial fibrillation discharged from 1st January 2011 to 1st October 2017, average age 75.3 ± 12.3 years, 64.6% male.

Intervention

Natural language processing pipeline to identify risk factors in clinical text and calculate CHA2DS2-VASc and HAS-BLED scores.

Comparator

Manual scoring by two independent experts (in a 40-patient validation subset).

Outcome

Agreement between automatic risk scores and independent experts (measured by kappa statistic).

An open-source NLP pipeline can automatically calculate CHA2DS2-VASc scores from EHR free text with strong expert agreement, enabling large-scale analysis of anticoagulation trends in atrial fibrillation.

Limitations

Risk of data accuracy degradation due to the NLP algorithm, with lower inter-rater agreement for HAS-BLED scores
Retrospective assessment of HAS-BLED components is challenging and not all variables are reliably recorded
Routine EHR data may not capture all necessary clinical information, potentially underestimating co-morbidities
The NLP algorithm was tested on data from a single multi-site organization, requiring further external validation
Inpatient admission data may represent an older, frailer population than typical community AF cohorts
Did not distinguish between different temporal patterns of atrial fibrillation (permanent, persistent, paroxysmal)
Observational study design precludes concluding causal links between clinical variables and prescribing changes
Requires further validation

Cite This Study

Bean et al. (2019) conducted an observational in Atrial fibrillation (n=10,030). Oral anticoagulants (OAC) vs. Antiplatelet only or no antithrombotic medication was evaluated on OAC prescription rate in high-risk AF patients (CHA2DS2-VASc ≥ 2). In high-risk patients with atrial fibrillation, oral anticoagulant use increased significantly from 42% in 2011 to 62% in 2017, driven by the availability of direct oral anticoagulants.

synapsesocial.com/papers/6a08f464a2bc65e38873a811 https://doi.org/10.1371/journal.pone.0225625

Topics

Atrial fibrillation Anticoagulation in AF Persistent AF management

View Full Paper

Ask AI

Bookmark

Semantic computational analysis of anticoagulation use in atrial fibrillation from real world data

Discussion

Member takes

Implication

Study Design

Structured PICO

Limitations

Cite This Study

Topics

Semantic computational analysis of anticoagulation use in atrial fibrillation from real world data

Discussion

Member takes

Implication

Study Design

Structured PICO

Limitations

Cite This Study

Also Consider

Topics

Also Consider