ICU infection surveillance can be based on electronic routine data: results of a case study | BMC Infectious Diseases

Sample characteristics

Six thousand, four hundred sixty patients were admitted to the ICU between April 2017 and March 2018. Four hundred ninety-nine patients had more than one ICU treatment episode (maximum: 6 ICU treatment episodes). The total number of ICU treatment episodes was 7350, of which 3327 (45%) had a duration of > 48 h. While the median (IQR) treatment duration of all patients was 1.8 (0.9, 3.9) days, the median (IQR) treatment duration of patients at risk for nosocomial events was 4.3 (3.0, 8.6) days.

The microbiological data contains 41,719 unique laboratory numbers. Overall, the median (IQR) number of laboratory numbers per patient was 3.0 (2.0, 7.0). In 176 patients (2.7%), no laboratory number was documented. As for patients with an ICU treatment duration of > 48 h, the median (IQR) number of laboratory numbers per patient was 6.0 (2.0, 12.0) with a total of 40 patients (1.3%) for whom no laboratory number was documented.

Approximately 35%, 25%, and 15% of the tests can be attributed to nose/throat swabs, rectal/anal swabs and blood samples respectively.

In total, 5822 positive bacterial swabs were detected during the period under observation, of which 3277 were gram-negative (56.3%), 1309 were gram-positive (22.5%), 1157 were common commensals (19.9%) and 79 were “exceptional” (1.4%) [11]. The three most prevalent bacteria were Escherichia coli (n = 1289), Staphylococcus epidermidis (n = 539) and Staphylococcus aureus (n = 497). A total of 192 distinct organisms were found in the data.

Multidrug-resistant organisms

MRSA

A total of 57 Staphylococcus aureus were considered MRSA by either EFFECT or manual documentation. While 38 were mutually considered MRSA (67%), 19 (33%)/0 were considered MRSA by EFFECT/manual documentation only (see Fig. 1). Of the 19 MRSA detected by EFFECT only, 17 (89%) were detected before ICU admission.

1. 1.

   Total number of MRSA identified as possibly ICU-acquired by EITHER EFFECT or manually on all 5 ICUs between April 2017 and March 2018 (n = 57).

2. 2.

   How many of these 57 MRSA were identified by BOTH EFFECT and manually? (n = 38)

   1. a.

      19 MRSA were found by EFFECT only; this is due to detection prior to ICU admission (n = 17) and detection between ICU admission and day two after ICU admission (n = 2)

3. 3.

   How many of these 38 MRSA were considered…

   1. a.

      ICU-acquired? (n = 14, right-hand side of the diagram)

      1. i

         ….by EFFECT only (n = 1, due to human error in manual documentation)

      2. ii

         …manually only (n = 1, due to human error in manual documentation)

      3. iii

         …by both EFFECT and manually (n = 12)

   2. b.

      Non-ICU-acquired? (n = 24, left-hand side of the diagram)

Of the 38 MRSA detected by both EFFECT and manual documentation, 12 (32%)/24 (63%) were mutually considered ICU-acquired/not ICU-acquired and 1 (3%)/1 (3%) were considered ICU-acquired by EFFECT/manual documentation only. In 36 out of 38 total cases (95%), EFFECT and manual documentation were concordant.

In the one case where EFFECT and manual documentation were discordant, the MRSA was detected between days one and two after ICU admission and is therefore not considered ICU-acquired according to EFFECT. In the other case, the MRSA was incorrectly classified as non-ICU-acquired by manual documentation.

VRE

A total of 124 E. faecium/E. faecalis isolates were considered VRE by either EFFECT or manual documentation. While 67 were mutually considered VRE (54%), 57 (46%)/0 were considered VRE by EFFECT/manual documentation only (see Fig. 2). Of the 57 VRE only detected by EFFECT, 56 (98%) were detected either before ICU admission or up to two days after ICU discharge.

Of the 67 VRE detected by both EFFECT and manual documentation, 35 (52%)/19 (28%) were mutually considered ICU-acquired/not ICU-acquired, and 7 (10%)/6 (9%) were considered ICU-acquired by EFFECT/manual documentation only. In 54 out of 67 total cases (81%), EFFECT and manual documentation were concordant.

In three out of six cases where the VRE were considered ICU-acquired via manual documentation only, it was due to differences in definition of the time at risk for nosocomial events (see Additional file 1: 5.3). In two cases, the VRE was detected before ICU admission and should not have been considered ICU-acquired. And in the remaining case, the VRE was detected one month earlier by EFFECT when compared to manual documentation.

In five out of seven cases when the VRE were considered ICU-acquired by EFFECT only, ULMC received information about prior findings from external hospitals where the patient was treated prior to ICU admission. This information was not part of the microbiological data set and could therefore not be taken into account by EFFECT. In two further cases, the different results can be attributed to significant differences in the day the VRE was detected.

MDRGN

A total of 340 gram-negative microorganisms were considered MDRGN by either EFFECT or manual documentation. While 198 were mutually considered MDRGN (58%), 118 (35%)/24 (7%) were considered MDRGN by EFFECT/manual documentation only (see Fig. 3). Of the 118 MDRGN detected by EFFECT only, 94 (80%) were detected either before ICU admission or up to 2 days after ICU discharge. For 16 out of 24 MDRGN detected by manual documentation only, no antibiogram was available; in seven cases, the antibiogram did not indicate multidrug-resistance, and in one case the MDRGN was not documented in the microbiological data.

Of the 198 MDRGN detected by both EFFECT and manual documentation, 73 (37%)/103 (52%) were mutually considered ICU-acquired/not ICU-acquired and 10 (5%)/12 (6%) were considered ICU-acquired by EFFECT/manual documentation only. In a total of 176/198 cases (89%), EFFECT and manual documentation were concordant.

In six out of ten cases where the MDRGN was considered ICU-acquired by EFFECT only, ULMC received information about prior findings from external hospitals where the patient was treated prior to ICU admission. This information was not part of the microbiological data set and could therefore not be taken into account by EFFECT. In the four remaining cases, the MDRGN should have been classified as ICU-acquired via manual documentation.

In two out of twelve cases where the MDRGN was considered ICU-acquired via manual documentation only, EFFECT and manual documentation came to different conclusions due to significant differences in the date the organism was considered multi-resistant. In three further cases, EFFECT did not consider the MDRGN ICU-acquired, because the organisms were detected on day two after ICU admission. In the remaining seven cases, the MDRGN should not have been classified ICU-acquired via manual documentation, as they were detected before the time at risk (up to day one after ICU admission).

Primary bacteremia/sepsis

During the period under observation, 94 ICU-acquired primary sepses were documented using ICU-KISS. The following analyses show how primary sepses considered ICU-acquired by ICU-KISS were classified by EFFECT (using definitions a and b for common commensals; see Additional file 1: 5.2).

Depending on the definition of bacteremia with common commensal organisms, EFFECT found 202 (definition a) and 117 (definition b) cases of ICU-acquired primary bacteremia.

In total, 94 primary sepses were considered ICU-acquired by ICU-KISS. Of these, 80 (85.1%) were considered primary bacteremia, 8 (8.5%) were considered secondary bacteremia and 6 (6.4%) were considered no bacteremia by EFFECT.

In five out of six bacteremia cases considered primary sepsis by ICU-KISS and no bacteremia by EFFECT, a negative/non-positive blood test was found by EFFECT up to two days after the positive blood culture. In the one remaining case, the microorganism identified as causing primary sepsis by manual documentation via ICU-KISS was not documented in the microbiological data set used by EFFECT.

Eight cases of primary sepsis (according to ICU-KISS) were considered secondary bacteremia by EFFECT. In all cases, the same organism found in a blood culture was also found in other relevant material (therefore making a case for secondary bacteremia as opposed to primary).

Pathogenic organisms

A total of 67 ICU-acquired cases of primary bacteremia with pathogenic microorganisms were found by EFFECT, and 49 ICU-acquired primary sepses with pathogenic microorganisms were documented by ICU-KISS. The number of distinct pathogenic microorganisms found by EFFECT (n = 23) was greater than the number documented by ICU-KISS (n = 15). The three most prevalent pathogenic organisms found to be the cause of ICU-acquired primary bacteremia (EFFECT)/sepsis (manual documentation/ICU-KISS) were Staphylococcus aureus (n = 14/11), Enterococcus faecium (n = 13/10) and Enterococcus faecalis (n = 9/8).

Common commensals

As specified in Additional file 1: 5.2, EFFECT uses two different definitions for primary bacteremia with common commensal organisms.

While 135 and 52 ICU-acquired primary bacteremia with common commensal organisms were found by EFFECT by applying definitions (a) and (b) respectively, 45 ICU-acquired primary sepses with common commensal organisms were documented in ICU-KISS. The number of distinct common commensal organisms detected was 16 (EFFECT, definition a), 10 (EFFECT, definition b) and 8 (ICU-KISS). Table 1 shows the three most prevalent common commensal organisms found to be the cause of ICU-acquired primary bacteremia/sepsis.