AI for Postoperative Pain and Opioid Stewardship in Pediatric Surgery: A Systematic Review.

Abstract

To synthesize evidence on artificial intelligence (AI)-enabled postoperative pain assessment and prediction, opioid-related risk prediction, patient-controlled analgesia (PCA) monitoring, and implementation considerations in pediatric surgical care. A systematic review with narrative synthesis of evidence concerning AI applications in pediatric postoperative pain management and opioid stewardship. Biomedical, nursing, engineering, computer science, trial-registry, Google Scholar, and citation sources were systematically searched through March 2026. Eligible studies evaluated AI, machine learning, deep learning, computer vision, multimodal sensing, or AI-enabled tools in pediatric surgical or postoperative populations. A single reviewer performed study selection, data extraction, and methodological appraisal, with consultation from an external subject-matter expert. Risk of bias and applicability were assessed using design-appropriate appraisal tools. Findings were narratively synthesized because clinical and methodological heterogeneity precluded meta-analysis. Of 198 identified records, 137 were screened, and 20 full-text reports were assessed. Eleven studies were included in the qualitative synthesis, alongside one companion dataset report. Applications included automated pain assessment, home-based pain monitoring, prolonged opioid-use prediction, and prediction of pain or adverse effects during PCA. Most studies evaluated facial expressions using computer vision or multimodal combinations of physiological, video, and audio data. Reported performance varied substantially, with validation predominantly limited to internal datasets. No randomized trial or controlled clinical-impact study demonstrated reductions in pain intensity, opioid exposure or prescribing, adverse events, readmissions, or length of stay attributable to AI-guided care. AI-enabled tools show potential to support pediatric postoperative pain assessment and opioid-related risk stratification. However, current evidence is insufficient to justify routine clinical implementation. External validation, prospective clinical evaluation, fairness assessment, and workflow-integrated effectiveness studies are needed. AI-enabled assessment and prediction tools may complement pediatric nurses' clinical assessment, pain monitoring, opioid-related risk identification, patient and family education, and timely escalation of care. Nevertheless, their use should remain supplementary to professional nursing judgment until clinical effectiveness, safety, and generalizability are established.

Abstract as supplied by the indexing source. © the authors / publisher.