JMIR Publications

ABSTRACT

Background:

Background:

With an increase in the number of artificial intelligence (AI) and machine learning (ML) algorithms available for clinical settings, appropriate model updating and implementation of updates are imperative to ensure applicability, reproducibility, and patient safety. Keywords: Model updating; model calibration; artificial intelligence; machine learning; direct clinical care

Objective:

We aim to characterize updating practices of AI/ML models used to support direct patient-provider clinical decision-making.Additionally, we are interested in the extent to which published algorithms show the racial and gender demographic distribution of their training data

Methods:

We utilized the PRISMA checklist and the PRISMA-P protocol guidance in addition to a modified CHARMS checklist to conduct this systematic review. A comprehensive, medical literature search of databases including Embase, Medline, PsycINFO, Cochrane, Scopus, and Web of Science was conducted to identify AI/ML algorithms that would impact clinical decision making at the level of direct patient care. Our primary endpoint is the rate at which model updating is recommended by published algorithms; in addition, we will conduct an assessment of study quality and risk of bias in all publications reviewed. We will also evaluate the rate at which published algorithms include racial and gender demographic distribution information in their training data as a secondary endpoint.

Results:

Our initial literature search yielded approximately 7,000 articles. We hope to complete the review process and disseminate the results by the spring of 2022.

Conclusions:

Although AI/ML applications in healthcare have potential to improve patient care, there currently exists more hype than hope due to lack of proper external validation of these models. We anticipate that AI/ML model updating methods are proxies to model applicability and generalizability upon implementation. Our findings will add to the field by determining the degree to which published models meet criteria for clinical validity, real-life implementation, and best practices to optimize model development, and in so doing, reduce the “inflated expectations” of the contemporary model development process. Clinical Trial: our systematic review protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO); CRD42021245470