Artificial Intelligence Versus Human Systematic Literature Review Into Negative-pressure Wound Therapy in Plastic Surgery

Abstract

Background: The potential of artificial intelligence (AI) to support physician evidence-based medicine is vast. We compared AI's ability to perform a systematic review of the literature to that of human investigators. Negative-pressure wound therapy (NPWT), a mainstay of wound management with a large but varied body of evidence, was therefore chosen as the subject of this investigation. Producing high-level evidence of NPWT's impact on wound healing has been challenging due to trial design issues, making a systematic review important and challenging. In this article, NPWT efficacy and the ability of AI to assess levels of evidence were evaluated.

Methods: A literature search was conducted using PubMed, SCOPUS, and CINAHL. The resulting articles were screened using Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. The Grading of Recommendations, Assessment, Development, and Evaluations criteria were applied by both humans and AI to analyze the quality and evidence of each article.

Results: Eighteen studies on 3131 patients were reviewed. Seven studies addressed length of stay; five showed shorter stays with NPWT. Fourteen studies examined infection rates. Eight found significant improvement with the use of NPWT. Twelve articles analyzed time to wound closure, and nine of those articles found reduced time when NPWT was utilized. AI generally assigned lower quality of evidence scores compared with humans.

Conclusions: AI is a promising tool but remains limited in accurately determining evidence quality. AI's lower scores may reflect reduced bias. Multiple confounders and the diversity of its application lead to a lack of high-level evidence of NPWT's efficacy.

Fig. 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analysis flowchart illustrating the process of article inclusion of 18 final articles.