A Systematic Review and Meta-Analysis Protocol for Machine Learning Algorithms for Prediabetes Risk Calculation (Preprint)

Abstract

UNSTRUCTURED Identifying and delivering interventions to patients with prediabetes was one strategy for dealing with the rising prevalence of T2DM. Risk assessment tools help in disease detection by allowing screening of the high risk group. Machine learning was also used to support in the detection and diagnosis of prediabetes. The purpose of this review is to assess the diagnostic test accuracy of various machine learning algorithms for calculating prediabetes risk. This protocol was written in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis of Protocols (PRISMA-P) statement. The databases that will be used include PubMed, ProQuest, and EBSCO, with access limited to January 1999 and September 2022 in English only. Two reviewers will identify articles independently by reading the titles, abstracts, and full-text articles. Any disagreement will be resolved through consensus. To assess the quality and potential for bias, the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tool will be used. Data extraction and content analysis will be carried out in a systematic manner. A forest plot with 95% confidence intervals will be used to visualize quantitative data. The summary receiver operating characteristic curve will describe the diagnostic test outcome. The Review Manager 5.3 (Rev Man 5.3) software package will be used to analyze the data. Discussion: Using the proposed systematic review and meta-analysis, we will determine the diagnostic accuracy of various machine learning algorithms for estimating prediabetes risk. Machine learning classification is a form of artificial intelligence (AI) that allows computers to learn without being specifically programmed. It has been used to develop a scoring method for prediabetes identification and diagnosis. As far as we know, there is no systematic review and meta-analysis regarding machine learning utilization for prediabetes risk estimation. Therefore, we proposed this study to obtain the diagnostic accuracy of machine learning algorithms in estimating prediabetes risk. This protocol has been registered in the Prospective Registry of Systematic Review (PROSPERO) database. The registration number is CRD42021251242.