Table 2.
Performance characteristics of ventilator-associated pneumonia diagnostic algorithms
| Criteria Studied | Year, (citation) | Population | Comparator | Sample Size | Sensitivity | Specificity | PPV | NPV | ( +) LR | (-) LR | ROC AUC | Kappa (ĸ) index, agreement level a |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CDC/NHSN | 2015, (60) | Mixedb | CPIS | 38 | 0.37 | 1.0 | 1.0 | 0.84 | ĸ = 0.47, moderate | |||
| CPIS | 1999, (43) | Mixedb | Pathology | 23 | 0.77 | 0.42 | ||||||
| CPISc | 2004, (66) | Mixedb | Quantitative Cultures | 69 | 0.41 | 0.77 | 0.8 | 0.36 | 0.64 | |||
| CPIS | 2004, (67) | Mixedb | Quantitative Cultures | 88 | 0.89 | 0.47 | 0.57 | 0.84 | ĸ = 0.33, fair | |||
| CPIS | 2007, (68) | Burn | Quantitative Cultures | 28 | 0.30 | 0.80 | 0.70 | 0.50 | ||||
| CPIS | 2010, (40) | Mixedb | Pathology | 142 | 0.46 | 0.60 | 1.13 | 0.96 | ||||
| CPIS | 2015, (69) | Surgical (mixed) | Quantitative Cultures | 497 | 0.633 | 0.644 | 0.61 | 0.674 | 0.60 | |||
| CPIS | 2018, (70) | Surgical (acute care) | Quantitative Cultures | 198 | 0.611 | 0.781 | 0.64 | 0.759 | ||||
| HELICS | 2013, (36) | Mixedb | Not clearly specified | 57d | 0.86 | 0.99 | 0.77 | 0.995 | ĸ = 0.80, substantial | |||
| Johansene | 1999, (43) | Mixedb | Pathology | 23 | 0.69 | 0.75 | ||||||
| Johansen | 2018, (70) | Surgical (acute care) | Quantitative Cultures | 198 | 0.828 | 0.59 | 0.564 | 0.843 | ||||
| NTDB/NTR | 2015, (71) | Trauma | CDC/NHSN | 279 | 0.864 | 0.578 | 0.74 | 0.74 | ĸ = 0.47, moderate |
CDC/NHSN: centers for disease control and prevention national health safety network; ICU: intensive care unit; MV: mechanical ventilation; NR: not reported; NTDB/NTR: national trauma data bank / national trauma registry; NPV: negative predictive value; PPV: positive predictive value; ROC AUC: receiver operating curve area under curve; ( +) LR: positive likelihood ratio; (-) LR: negative likelihood ratio
aAgreement based on score: ≤ 0 (no agreement); 0.01–0.20 (slight); 0.21–0.40 (fair); 0.41– 0.60 (moderate); 0.61–0.80 (substantial); and 0.81–1.00 (almost perfect agreement)
bA mixed population containing both medical and surgical patients. Studies that did not specify ICU type were by default classified as mixed
cFor CPIS threshold of > 7, rather than current standard of > 6. The AUC using threshold CPIS > 6 was 0.54, other values not reported
dData from sub-population of a larger study assessing various types of ICU-acquired infections
eThe presence of all three criteria increased the specificity to 92% at the cost of a high beta error (sensitivity 23%)