Clinical Scorecard: Authors’ Response: Understanding Agreement in AI-Assisted ASA Classification
At a Glance
Category
Detail
Condition
Clinical topic addressed by the source article: Authors’ Response
Key Mechanisms
Mechanisms or clinical associations described in the source article.
Target Population
The bootstrap analysis sampled 600 patients with replacement. Category-level cross-tabulations included 57 clinician-assigned ASA III patients and 6 clinician-assigned ASA IV patients in the NHI-data condition.
Care Setting
Routine preanesthetic documentation
Key Highlights
Unweighted Cohen κ was 0.883 (95% CI 0.841-0.946) with NHI data and 0.518 (95% CI 0.412-0.691) without NHI data.
The paired-bootstrap difference in unweighted κ was 0.365 (95% percentile bootstrap CI 0.292-0.440), based on 10,000 patient-level replicates.
Linearly weighted Cohen κ was 0.893 with NHI data and 0.560 without NHI data.
All 9 disagreements among 63 patients with NHI data were between adjacent ASA categories; without NHI data, 16 of 17 disagreements were adjacent.
The authors note that κ results do not establish comparable performance across ASA categories; interpretation is limited by the large proportion of ASA II patients and small ASA III-to-IV subgroups.
Guideline-Based Recommendations
Diagnosis
Management
Monitoring & Follow-up
Risks
Patient & Prescribing Data
The analysis involved 600 patients. Among clinician-assigned ASA III patients, AI classified 50 of 57 as ASA III with NHI data and 44 patients as ASA III without NHI data. Among clinician-assigned ASA IV patients, 4 of 6 were classified as ASA IV with NHI data and 2 of 6 without NHI data.
No treatment or prescribing data are reported. The response evaluates classification agreement, not treatment effects.
