Abstract Body: Background:

Due to limited diagnostic accuracy in the symptom-based tools currently used by paramedics, triage nurses, and emergency physicians, up to 35% of strokes are missed at initial clinician contact. As a result, there has been a push to identify blood biomarkers that could aid in stroke recognition at triage. There is evidence to suggest stroke is more frequently mistriaged in minority patients than White patients; to avoid perpetuating these disparities as stroke biomarkers move towards clinical use, it is important that involved investigations meaningfully account for race and ethnicity. However, it is unclear how commonly this occurs in practice.

Purpose:

This systematic review and meta-analysis aimed to quantify the degree that race and ethnicity information was reported in prior acute stroke blood biomarker investigations, and assess the effect of racial and ethnic heterogeneity on reported diagnostic performance.

Methods:

An electronic database search to identify published human acute stroke blood biomarker investigations yielded 14,253 articles. After abstract and full-text screening, 189 articles reporting 705 unique diagnostic comparisons were analyzed. For each diagnostic comparison, the reported primary diagnostic statistics were manually extracted, along with the racial and ethnic composition of the study sample if available. Diagnostic statistics were converted to a single diagnostic odds ratio, and race and ethnicity information was used to calculate a Gini diversity index value to represent the demographic heterogeneity of the sample.

Results:

Subject race or ethnicity was only described for 21.8% of diagnostic comparisons. For diagnostic comparisons reporting this information, study populations showed high homogeneity with median Gini diversity indices of 0.33 for race and 0 for ethnicity. We found a negative correlation between Gini diversity index for race and diagnostic odds ratio (Spearman’s rho= -0.17, p=0.14), indicating to some degree that that more homogenous study samples are more likely to produce higher diagnostic performance estimates.

Conclusions:

Findings from acute stroke biomarker investigations are at risk for poor generalizability to the diverse clinical populations that stroke triage serves. To reduce this risk, future work needs to address race and ethnicity more meaningfully in terms of both sampling methods and reporting.