What question did this study set out to answer?

Synthesize evidence on algorithms, datasets, evaluation metrics, and real-world implementation readiness for artificial intelligence systems designed to automate ICD-10 coding from clinical text.

August 16, 2026Frontiers in Digital HealthOpen Access

Artificial intelligence for automated ICD-10 coding: a systematic review of multi-label text classification in clinical narratives

View Full Paper

Ask AI

Bookmark

Authors

KTKamonrat TangudomkitPrince of Songkla University SCSawrawit ChairatPrince of Songkla University SCSitthichok ChaichuleePrince of Songkla University

Discussion

Member takes

Overview

Systematic review demonstrates that hybrid deep learning provides stable automated ICD-10 coding across clinical narratives, suggesting dataset structure heavily influences AI performance.

Key Points

Synthesize evidence on algorithms, datasets, evaluation metrics, and real-world implementation readiness for artificial intelligence systems designed to automate ICD-10 coding from clinical text.
Conducted a PRISMA-guided systematic review (OSF: https://osf.io/cegqk) searching 7 academic databases for studies published between January 1, 2020 and December 31, 2025.
Evaluated machine learning, deep learning, transformer, and large language model approaches across 24 eligible studies (extracted from 257 records) encompassing 296 experimental evaluations.
Hybrid deep learning was the most widely used primary approach and demonstrated the most stable performance under full-code evaluations, whereas large language models showed less consistent efficacy for structured multi-label classification.
F1-macro scores were consistently lower than F1-micro scores across studies reporting both metrics, and model performance varied directly based on the documents-per-label ratio.
Methodological quality varied widely across the included literature, with 7 studies assessed as high quality, 8 as moderate, and 9 constrained by technical or methodological limitations.

Cite This Study

Tangudomkit et al. (2026) studied this question.

synapsesocial.com/papers/6a81791ff2fb91fc834ac460 https://doi.org/10.3389/fdgth.2026.1905379

View Full Paper

Ask AI

Bookmark

August 16, 2026Frontiers in Digital HealthOpen Access

Artificial intelligence for automated ICD-10 coding: a systematic review of multi-label text classification in clinical narratives

View Full Paper

Ask AI

Bookmark

Authors

KTKamonrat TangudomkitPrince of Songkla University SCSawrawit ChairatPrince of Songkla University SCSitthichok ChaichuleePrince of Songkla University

Discussion

Member takes

Overview

Systematic review demonstrates that hybrid deep learning provides stable automated ICD-10 coding across clinical narratives, suggesting dataset structure heavily influences AI performance.

Key Points

Synthesize evidence on algorithms, datasets, evaluation metrics, and real-world implementation readiness for artificial intelligence systems designed to automate ICD-10 coding from clinical text.
Conducted a PRISMA-guided systematic review (OSF: https://osf.io/cegqk) searching 7 academic databases for studies published between January 1, 2020 and December 31, 2025.
Evaluated machine learning, deep learning, transformer, and large language model approaches across 24 eligible studies (extracted from 257 records) encompassing 296 experimental evaluations.
Hybrid deep learning was the most widely used primary approach and demonstrated the most stable performance under full-code evaluations, whereas large language models showed less consistent efficacy for structured multi-label classification.
F1-macro scores were consistently lower than F1-micro scores across studies reporting both metrics, and model performance varied directly based on the documents-per-label ratio.
Methodological quality varied widely across the included literature, with 7 studies assessed as high quality, 8 as moderate, and 9 constrained by technical or methodological limitations.

Cite This Study

Tangudomkit et al. (2026) studied this question.

synapsesocial.com/papers/6a81791ff2fb91fc834ac460 https://doi.org/10.3389/fdgth.2026.1905379

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark