Skip to main content

Table 2 Prevalence of pneumococcal pneumonia (PPN) in case-patients and performance of real-time polymerase chain reaction (qPCR) and urine antigen test (UAT): conventional analysis with UAT as a gold standard vs. Bayesian latent class models (BLCM) (n = 353)

From: Pneumococcal pneumonia prevalence among adults with severe acute respiratory illness in Thailand - comparison of Bayesian latent class modeling and conventional analysis

 

Conventional analyses

Mean (95% CI)

BLCM

Median (95% CrI)

Primary

Adjusted

Binary qPCR model

Continuous qPCR model

27 UAT positives considered true positives

UAT sensitivity is 75% and specificity is 100%

qPCR and UAT

qPCR and UAT

Prevalence

8% (5–11%)

10% (7–15%)

10% (7–16%)

11% (7–17%)

Sensitivity of qPCR

96% (81–100%)

 

96% (83–100%)

 

Specificity of qPCR

74% (69–79%)

 

75% (71–79%)

 

PPV of qPCR

23% (16–32%)

 

31% (20–44%)

 

Mean qPCR Ct among PPN

25.6 (24.2–26.9)

  

25.7 (24.6–27)

Mean qPCR among Non-PPN

31 (30–32)

  

32.4 (31.4–33.4)

Sensitivity of UAT

100% (assumed)

75% (published)

72% (53–87%)

67% (51–83%)

Specificity of UAT

100% (assumed)

100% (assumed)

99.5% (98.5–99.9%)

99.5% (98.6–99.9%)

PPV of UAT

100%

100%

94% (83–99%)

95% (83–99%)

Probability of PPV of UAT < 90%

0

0

18%

17%