Shuya Lu

Background: Surgical site infections (SSIs) remain a major cause of health care–associated infections, and early prediction is essential for improving patient outcomes. Machine learning (ML) has shown potential for SSI prediction; however, clinical implementation requires models that are both accurate and explainable. Despite recent progress in explainable ML, its clinical application to SSI prediction remains limited. Objective: This study aimed to map explainable ML models for SSI prediction from a clinical perspective and examine their use of structured and unstructured data across the dimensions of data, methodology, and explanation output. Methods: We conducted a scoping review following PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) and Joanna Briggs Institute (JBI) guidance, and registered the protocol in PROSPERO. Six databases were searched for eligible studies published from January 2010 onward, without language restrictions. The search was conducted on August 9, 2025, and updated on July 14, 2026. We included studies that developed or validated an explainable ML model for predicting SSI in adults. Two reviewers (RS and YL) independently screened studies and extracted data. Findings were narratively synthesized and presented in evidence maps. Methodological quality was assessed using the PROBAST+AI (Prediction model Risk Of Bias Assessment Tool for prediction models using regression or artificial intelligence) methods. Results: Overall, 77 studies reporting 98 ML models were included. Most models were prognostic (72/98, 73.5%), whereas 26 focused on postoperative SSI diagnosis. A total of 81.8% (63/77) of studies addressed a single surgical specialty, most commonly gastrointestinal surgery (27/63, 42.9%). Overall, 51.9% (40/77) of studies addressed composite SSI predictions. Among all models, % (40/98) were black-box models explained by post hoc methods; SHAP combined with ensemble learning was the leading approach (18/40, 45%). Regression models accounted for half of the inherently interpretable models, interpreted using coefficients. Prognostic models commonly included health and lifestyle (60/72, 83.3%), individual characteristics, and surgical process details (both 57/72, 79.2%); health and lifestyle factors were most frequently important across SSI types. Diagnostic models commonly included surgical process details (13/26, 50%), administrative codes, and individual characteristics (both 10/26, 38.5%). Key diagnostic predictors varied by SSI types: postoperative clinical interventions predominated for composite SSI; vital signs, postoperative interventions, and administrative codes for superficial SSI; postoperative recovery status for deep SSI; and vital signs for organ-space SSI. Conclusions: Extending previous reviews focusing on model performance, this review mapped explainability methods and important features in SSI prediction, identifying recurring predictor patterns and substantial methodological heterogeneity across prognostic and diagnostic settings. Incomplete reporting of feature definitions and explanatory rationale, together with limited clinical relevance, constrained clinical interpretation and actionability. Clinician-informed reporting frameworks and validation of explanation fidelity and clinical relevance are needed to improve the trustworthiness and utility of SSI prediction models.

Background . Hundreds of coronavirus disease 2019 (COVID-19) clinical practice guidelines (CPGs) and expert consensus statements have been developed and published since the outbreak of the epidemic. However, these CPGs are of widely variable quality. So, this review is aimed at systematically evaluating the methodological and reporting qualities of COVID-19 CPGs, exploring factors that may influence their quality, and analyzing the change of recommendations in CPGs with evidence published. Methods . We searched five electronic databases and five websites from 1 January to 31 December 2020 to retrieve all COVID-19 CPGs. The assessment of the methodological and reporting qualities of CPGs was performed using the AGREE II instrument and RIGHT checklist. Recommendations and evidence used to make recommendations in the CPGs regarding some treatments for COVID-19 (remdesivir, glucocorticoids, hydroxychloroquine/chloroquine, interferon, and lopinavir-ritonavir) were also systematically assessed. And the statistical inference was performed to identify factors associated with the quality of CPGs. Results . We included a total of 92 COVID-19 CPGs developed by 19 countries. Overall, the RIGHT checklist reporting rate of COVID-19 CPGs was 33.0%, and the AGREE II domain score was 30.4%. The overall methodological and reporting qualities of COVID-19 CPGs gradually improved during the year 2020. Factors associated with high methodological and reporting qualities included the evidence-based development process, management of conflicts of interest, and use of established rating systems to assess the quality of evidence and strength of recommendations. The recommendations of only seven (7.6%) CPGs were informed by a systematic review of evidence, and these seven CPGs have relatively high methodological and reporting qualities, in which six of them fully meet the Institute of Medicine (IOM) criteria of guidelines. Besides, a rapid advice CPG developed by the World Health Organization (WHO) of the seven CPGs got the highest overall scores in methodological (72.8%) and reporting qualities (83.8%). Many CPGs covered the same clinical questions (it refers to the clinical questions on the effectiveness of treatments of remdesivir, glucocorticoids, hydroxychloroquine/chloroquine, interferon, and lopinavir-ritonavir in COVID-19 patients) and were published by different countries or organizations. Although randomized controlled trials and systematic reviews on the effectiveness of treatments of remdesivir, glucocorticoids, hydroxychloroquine/chloroquine, interferon, and lopinavir-ritonavir for patients with COVID-19 have been published, the recommendations on those treatments still varied greatly across COVID-19 CPGs published in different countries or regions, which may suggest that the CPGs do not make sufficient use of the latest evidence. Conclusions . Both the methodological and reporting qualities of COVID-19 CPGs increased over time, but there is still room for further improvement. The lack of effective use of available evidence and management of conflicts of interest were the main reasons for the low quality of the CPGs. The use of formal rating systems for the quality of evidence and strength of recommendations may help to improve the quality of CPGs in the context of the COVID-19 pandemic. During the pandemic, we suggest developing a living guideline of which recommendations are supported by a systematic review for it can facilitate the timely translation of the latest research findings to clinical practice. We also suggest that CPG developers should register the guidelines in a registration platform at the beginning for it can reduce duplication development of guidelines on the same clinical question, increase the transparency of the development process, and promote cooperation among guideline developers all over the world. Since the International Practice Guideline Registry Platform has been created, developers could register guidelines prospectively and internationally on this platform.