Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
September 25, 2026DiagnosticsOpen Access

AI models predicting perioperative pulmonary complications lack robust validation, with only ~15% externally validated.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Despite advances in AI and machine learning models to predict perioperative pulmonary and respiratory complications, these models are rarely implemented in routine clinical practice due to a translation gap.

Population

59 unique publications with 63 phenotype-specific study-listings on PPRCs

Comparison

AI and machine learning models for predicting PPRCs vs current clinical practice

Design

Narrative review synthesizing evidence and proposing a translational framework

Key result

Artificial intelligence models for predicting perioperative pulmonary complications lacked robust validation, with only 15.3% (9 of 59) of publications reporting true external validation.

Authors

JLJi-Yeon LeeJLJi-Yeon LeeYIYong-Ho In

Discussion

Loading...

Member takes

Overview

AI models for PPRC prediction rarely reach clinical practice; leaves open the need for implementation and prospective validation studies.

Key Points

  • To review the current state of artificial intelligence models predicting perioperative pulmonary and respiratory complications and establish an implementation framework to bridge the translational gap into clinical practice.
  • Synthesized findings from a narrative review of 59 unique publications encompassing 63 phenotype-specific study-listings across six distinct perioperative pulmonary and respiratory complication phenotypes.
  • Assessed studies using a tiered outcome ascertainment quality framework (TIER) and evaluated multi-institutional external validation rates.
  • Classified 39 of 63 phenotype-specific study-listings (61.9%) as high outcome ascertainment quality (TIER A) based on the proposed tiered framework.
  • Identified that only 9 of the 59 unique publications (15.3%) conducted true cross-institutional external validation, demonstrating limited generalizability across health systems.

Structured PICO

P
Population
A narrative review of 59 unique publications evaluating artificial intelligence and machine learning models for predicting perioperative pulmonary and respiratory complications.
E
Exposure
Artificial intelligence and machine learning models

Near-term clinical translation of AI models for perioperative respiratory complications is more likely through institution-specific development and local adaptation rather than direct widespread deployment.

Limitations

  • Most models lacked robust validation, with only 15.3% reporting true cross-institutional external validation, thereby limiting their generalizability.

Cite This Study

Lee et al. (2026) conducted a review in Perioperative pulmonary and respiratory complications (n=59). Artificial intelligence and machine learning models was evaluated on Outcome ascertainment quality and external validation. Artificial intelligence models for predicting perioperative pulmonary complications lacked robust validation, with only 15.3% (9 of 59) of publications reporting true external validation.

synapsesocial.com/papers/6ab60feb406bf401c1468f6dhttps://doi.org/10.3390/diagnostics16193082
View Full Paper
Ask AI
Bookmark
Share

Also Consider

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

  1. 1From prediction to practice: closing the translation gap in artificial intelligence for anesthesia2026 · 2 citations
  2. 2Construction of an automated machine learning-based predictive model for postoperative pulmonary complications risk in non-small cell lung cancer patients undergoing thoracoscopic surgery2025 · 7 citations
  3. 3Developing Machine Learning Algorithms to Predict Pulmonary Complications After Emergency Gastrointestinal Surgery2021 · 25 citations
  4. 4Development and validation of a deep neural network model to predict postoperative mortality, acute kidney injury, and reintubation using a single feature set2020 · 81 citations
  5. 5Development and Evaluation of the Universal ACS NSQIP Surgical Risk Calculator: A Decision Aid and Informed Consent Tool for Patients and Surgeons2013 · 1,917 citations