Skip to main content
. 2022 Feb 4;238(4):649–656. doi: 10.1159/000520474

Table 2.

Overall sensitivity and specificity of the app in detecting skin premalignancy and malignancy, including subgroup analyses between the iOS and Android devices, melanocytic versus nonmelanocytic skin lesions, skin fold lesion areas versus smooth skin lesion areas, suspicious skin lesions versus benign control lesions, skin lesions identified after GP and nondermatology referrals versus follow-up consultations of known patients at the outpatient clinics

Assessment type N (%) Sensitivity (95% CI), % p value Specificity (95% CI), % p value
Overall app accuracy 785 (100) 86.9 (82.3–90.7) 70.4 (66.2–74.3)

Android device iOS device 425 (54.1) 360 (45.9) 83.0 (75.7–88.8)* 91.0 (84.9–95.3)* 0.02 71.5 (65.9–76.7)
69.0 (62.6–75.0)
0.27

Melanocytic skin lesions Nonmelanocytic skin lesions 179 (22.8)
606 (77.2)
81.8 (59.7–94.8)
87.4 (82.6–91.2)
0.26 73.3 (65.6–80.0)
69.1 (64.0–73.9)
0.17

Skin fold lesion areas Smooth skin lesion areas 138 (17.6)
647 (82.4)
92.9 (85.3–97.4)*
84.2 (78.2–89.1)*
0.01 56.6 (42.3–70.2)*
72.0 (67.6–76.1)*
0.02

Suspicious skin lesions Benign control lesions 418 (53.3)
367 (46.7)
86.9 (82.3–90.7) 45.5 (37.1–54.0)***
80.1 (75.7–84.1)***
<0.001

GP and nondermatology referrals Follow-up consultations 213 (27.1)
205 (26.1)
89.6 (83.4–94.1)
84.0 (76.6–89.8)
0.09 39.1 (27.6–51.6)
51.4 (39.4–63.2)
0.07
*

p < 0.05

***

p < 0.001.