Confounders in Predictive Medical Models: Roles of Nationality and Immigrant Status

Abstract

Aim: The aim of this study was to assess the opinion of natural science specialists on the latest recommendations of official regulators regarding the prevention of causes of social disparities in artificial intelligence (AI) and machine learning (ML) models. Materials and Methods: An anonymous online survey was conducted using the Telegram platform, where participants were asked a single question: "Is the inclusion of predictors such as “nationality” and “immigrant status” in AI and ML medical models ethical and consistent with contemporary scientific standards?" Respondents were provided with two response options: "Yes" or "No." The survey was specifically targeted at international groups, focusing primarily on English and Russian-speaking clinicians and scientific researchers. Results: 180 unique individuals participated in the survey. The results revealed that one-third of the respondents (60 individuals) agreed that including predictors such as nationality and immigration status is inappropriate in the current ML and AI models. Conclusion: In conclusion, the fact that only one-third of respondents disagree with categorizing patients based on nationality is at odds with the standards set by official regulators. This discrepancy underscores the need for educational programs aimed at sensitizing the scientific community to prioritize biological predictors over data documented in passports or identity cards.