Page 122 - AIH-2-1

Artificial Intelligence in Health Does improving diagnostic accuracy increase AI adoption?

Figure 1. Example of a randomized scenario

and saliva test as the benchmark for further analysis. were significantly different at a 1% threshold). Figure 2
Then, descriptive analyses were conducted to compare the shows the proportion of AI adoption (vs. saliva test)
proportions of acceptance of the AI and/or AI+ option across different levels of accuracy. Upon stratifying the
with those of the saliva test option. The differences in answers by sensitivity level, we found that 35.04% of the
acceptance rates between the AI-versus-saliva-test option answers were in favor of AI or AI+ when AI’s sensitivity
and the AI+-versus-saliva-test option were analyzed using was 60% and 31.63% when AI’s sensitivity was 70%, and
pairwise z-tests, whereas the differences in proportions this proportion increased to 48.68% when AI’s sensitivity
between three levels of sensitivity and specificity per type was 95% (Chi-squared test showed significant difference
of pairs offered (i.e., AI vs. saliva and AI+ vs. saliva) were at a 1% threshold). With respect to specificity, 29.68%,
compared using Chi-squared tests. 13,14 28.18%, and 44.24% of the answers favored AI/AI+
test over saliva test when specificity levels were 60%,
4. Results 70%, and 95%, respectively (Chi-squared test showed
Contingency tables describing proportions of agreement significant differences at a 1% threshold). Finally, when
between the AI and/or AI+ alternative with the saliva we compared AI-versus-saliva-test option and AI+-
test alternative are presented in Table 1. From the 3225 versus-saliva-test option per sensitivity or specificity
pairs of AI/AI+ against saliva tests that were offered, levels, we found significantly higher acceptance rates
1482 are linked to a choice of AI test versus saliva test, in the AI group than in the AI+ group from the same
whereas 1743 are associated with a choice of AI+ test sensitivity or specificity level (except when sensitivity is
versus saliva test. Only 36.68% of the answers were 60%, where we found no significant differences between
pro-AI/AI+ (45.82% for the AI-vs.-saliva-test pairs and rates of acceptance when AI-vs.-saliva-test or AI+-vs.-
28.92% for the AI+-vs.-saliva-test pairs; proportions saliva-test options were offered).

Volume 2 Issue 1 (2025) 116 doi: 10.36922/aih.3561