The accuracy of fully-automated algorithms for the surveillance of central venous catheter-related bloodstream infection in hospitalised patients

Table 4 Rates central venous catheter-related bloodstream infection based on annotation (reference) and algorithm results

	Validation dataset with admissions with at least one positive BCx
	All CVC-BSI cases			HA CVC-BSI cases
	Reference^a (n = 51)	Algorithm 1^b (n = 47)	Algorithm 2^c (n = 49)	Reference^a (n = 32)	Algorithm 1^b (n = 26)	Algorithm 2^c (n = 27)
Potential episodes^d (n = 4326)	1.18%	1.09%	1.13%	0.74%	0.60%	0.62%
Positive BCx (n = 1766)	2.89%	2.66%	2.77%	1.81%	1.47%	1.53%
Admissions (n = 1000)	5.1%	4.7%	4.9%	3.2%	2.6%	2.7%
1000 in-patient hospital-days (n = 17,782)	2.87	2.64	2.76	1.80	1.46	1.52

	Extrapolated results to validation period cohort
	All CVC-BSI cases			HA CVC-BSI cases
	Reference^a (n = 267)	Algorithm 1^b (n = 246)	Algorithm 2^c (n = 256)	Reference^a (n = 167)	Algorithm 1^b (n = 137)	Algorithm 2^c (n = 142)
Potential episodes^d (n = 181,354)	0.15%	0.14%	0.14%	0.09%	0.08%	0.08%
Admissions (n = 132,850)	0.20%	0.19%	0.19%	0.13%	0.10%	0.11%
1000 in-patient hospital-days (n = 817,058)	0.33	0.30	0.31	0.20	0.17	0.17

BCx: blood culture; CVC-BSI: central venous catheter-related bloodstream infection; HA: healthcare-associated
^aReference: True rate based on manual record review in validation dataset and expected rate based on extrapolation in validation period cohort
^bAlgorithm 1: Relevant microbiological findings in BCx and CVC cultures combined with the presence of BSI symptoms
^cAlgorithm 2: Only relevant microbiological findings in BCx and CVC cultures
^dCVC-BSI-episode: all performed BCx were regarded as potential CVC-BSI-episodes during an admission, and admissions with no BCx counted as one potential CVC-BSI-episode

ISSN: 2047-2994