Table 2.

Subgroup analysis based on different standards.

Studies	Sensitivity (95%CI)	Specificity (95%CI)	DOR (95%CI)	AUC (95%CI)
All (23)	0.91(0.89–0.93)	0.65(0.55–0.75)	20(13–29)	0.91(0.89–0.94)
Study Design
Prospective (12)	0.91(0.87–0.94)	0.48(0.34–0.62)	9(4–20)	0.85(0.82–0.88)
Nonprospective (11)	0.87(0.78–0.93)	0.75(0.53–0.89)	20(5–84)	0.90(0.87–0.92)
Software
CAD4TB (18)	0.89(0.82–0.94)	0.57(0.42–0.70)	11(4–30)	0.83(0.80–0.86)
qXR (8)	0.79(0.61–0.90)	0.55(0.24–0.83)	5(1–38)	0.77(0.73–0.80)
Lunit INSIGHT CXR (8)	0.88(0.75–0.94)	0.78(0.40–0.95)	25(3–211)	0.91(0.88–0.93)
Reference standard
Human reader (5)	0.90(0.84–0.94)	0.90(0.80–0.95)	77(22–269)	0.95(0.93–0.97)
Xpert MTB/RIF (9)	0.90(0.85–0.93)	0.36(0.24–0.50)	5(2–12)	0.79(0.75–0.82)
AI type
Deep learning (13)	0.91(0.89–0.92)	0.62(0.48–0.74)	16(10–23)	0.91(0.88–0.93)
Machine learning (9)	0.93(0.85–0.97)	0.61(0.46–0.75)	21(11–42)	0.87(0.83–0.89)

Abbreviation: DOR, diagnostic odds ratio; AUC, area under curve.