Currently submitted to: JMIR Medical Informatics
Date Submitted: Nov 4, 2026
Open Peer Review Period: Nov 13, 2026 - Jan 13, 2027
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Performance and Processing Time of an AI-Assisted Quality-Control System for Detecting Documentation Defects in Cesarean Delivery Records: A Single-Center Retrospective Validation Study
ABSTRACT
Background:
Quality control of cesarean delivery records is labor-intensive and may miss time-sensitive or cross-document documentation defects. Whether an artificial intelligence-assisted system can provide high-throughput screening with acceptable detection performance remains uncertain.
Objective:
To evaluate the detection performance of an artificial intelligence-assisted quality-control system and compare its processing time with independent expert review of cesarean delivery records.
Methods:
We conducted a single-center retrospective validation study using a simple random sample of 3,077 cesarean admissions from records generated between January 1, 2025, and June 30, 2026. The AI-assisted system and two independent experts assessed 10 core documentation indicators, yielding 30,770 case-indicator evaluations. Discrepant assessments underwent joint adjudication. Detection metrics were calculated against the resulting consensus reference, and processing durations were compared within admission.
Results:
Consensus adjudication identified 3,407 defective evaluation units, corresponding to 11.07% of all evaluations. The AI-assisted system achieved a sensitivity of 84.94%, specificity of 93.64%, positive predictive value of 62.44%, negative predictive value of 98.04%, accuracy of 92.67%, and F1 score of 0.720. Median processing time was 1.8 min per admission for the AI system, compared with 14.0 min and 14.8 min for Experts A and B, respectively. The median within-admission duration reductions were 87.41% and 88.16%, respectively, with both paired comparisons yielding P < 0.001. Performance varied substantially by indicator, with sensitivity of 1.39% for the surgical risk assessment and safety checklist indicator and a positive predictive value of 7.52% for the preoperative ward-round indicator. At the admission level, sensitivity and specificity were 94.49% and 53.51%, respectively.
Conclusions:
AI-assisted system enabled rapid, relatively sensitive screening of documentation defects in cesarean delivery records, particularly for selected postoperative and discharge indicators. However, substantial indicator-level heterogeneity, false-alert burden, and missed defects in specific categories limit autonomous use. The system is best positioned as an adjunctive triage tool that prioritizes records for human review. Prospective and multicenter evaluations are needed before broader deployment.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.