Systematic Review

Comparative Analysis of the Diagnostic Performance of AI POCT and Traditional POCT: A Systematic Review

Abstract

Introduction: Artificial intelligence (AI) has increasingly been integrated into point-ofcare testing (POCT) systems to enhance diagnostic accuracy and overcome limitations inherent in traditional POCT methods, particularly those dependent on manual result interpretation. This systematic review compares the diagnostic performance of AI-enhanced POCT and traditional POCT across a range of medical conditions and clinical settings. Methods: Using PRISMA 2020 guidelines, a comprehensive search was conducted across four databases—PubMed, Google Scholar, ScienceDirect, IEEE Xplore and Semantic Scholar—covering publications from January 2015 to April 2024. Eligible studies reported at least one diagnostic accuracy metric, such as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), or area under the curve (AUC). Results: Eight studies met the inclusion criteria, targeting diseases such as COVID-19, influenza, diabetic retinopathy, streptococcal pharyngitis, myocardial infarction, and β- thalassemia. Across studies, AI-enhanced POCT demonstrated comparable and superior diagnostic performance compared to traditional methods. Sensitivity and specificity of AI-driven systems typically exceeded 94%, with some models reporting AUC values approaching 0.99. In contrast, traditional POCT devices showed greater variability in performance, with sensitivities ranging from 24% to 93% depending on disease type and user interpretation. AI integration reduced subjectivity, improved consistency, and enabled automated interpretation of complex data, particularly in imaging-based diagnostics and low- resource environments. However, challenges such as infrastructure demands, algorithm bias, and regulatory gaps persist, especially in under-resourced settings. Conclusion: Despite these hurdles, AI enhanced POCT offers significant potential to democratize access to expertlevel diagnostics, reduce diagnostic errors, and speed up decision-making in high-pressure clinical environments. Rather than replacing traditional POCT, AI-based systems should be viewed as complementary tools that fill critical gaps where precision, standardization, and automation are essential. The findings reveal the need for further head-to-head comparative studies and broader validation across diverse populations.

Authors affiliation:

Williams Tarimobowei Tabowei: Pathology Department, James Paget University Hospital, NR31 6LA, Lowestoft road, Gorleston, Norwich.
Chikadibia Fyneface Amadi (ORCID): Department of Medical Laboratory Science, Faculty of Allied Health Sciences, PAMO University of Medical Sciences, Rivers State, Nigeria.
Elaine Mitchell: James Paget University Hospital NHS Trust, Lowestoft Road, Gorleston-on-Sea Great Yarmouth, NR31 6LA, UK.
Godfrey Tabowei: Texas Tech University Health Science Center, Permian Basin, Odessa, Texas, United States.

How To Cite: W.T. Tabowei, C.F. Amadi, E. Mitchell, and G. Tabowei. Comparative Analysis of The Diagnostic Performance of AI POCT and Traditional POCT: A Systematic Review. Journal of Applied Health Sciences and Medicine, 6(11):17–28, 2026. https://doi.org/10.58614/jahsm6113