Unequal health risk alarms: How thresholds drive false-positive burdens across sex and race

Health risk stratification algorithms are increasingly deployed in clinical and public health settings. Despite their potential to predict future outcomes, improve efficiency and consistency, these tools can perpetuate disparities when error patterns fall disproportionately across demographic groups. This study evaluated a simple cardiovascular risk index using NHANES 2017-2018 to examine how a threshold-based classification distributes predictive errors among sex and race subgroups. At a cutoff labeling the top 70% of individuals as 'high risk,' sensitivity was perfect across all subgroups. Yet false positives were not evenly distributed: males were misclassified more often than females (59% vs. 42%), and differences across racial groups reached more than 10 percentage points. These results highlight how technical choices, particularly thresholds, translate into unequal burdens in real-world clinical practice. We discuss the implications for equity, operational trust, and accountability, and recommend fairness auditing, threshold calibration, and transparent reporting as essential components of responsible model deployment.