Background
Infections with Highly Resistant Gram Negative Rods (HR-GNRs) are associated with higher (hospital) costs, morbidity and mortality in comparison to susceptible micro-organisms [
1‐
4]. Increasingly, studies report on the (colonization) prevalence of HR-GNRs, including ESBL (Extended Spectrum Beta Lactamase) producing bacteria isolated from hospitalized patients, general practitioner patients and nursing home residents [
5‐
9]. Knowledge about regional prevalence rates is important since HR-GNR colonized patients constitute a potential reservoir for patients at risk for nosocomial infections, such as immune compromised patients and/or patients with open wounds [
10‐
13]. Several studies showed that foreign travel is an important risk factor for HR-GNR colonization [
14‐
16]. Therefore, in Dutch hospitals, patients who have a recent history of foreign hospital admission are actively screened and pre-emptively isolated until test results are known [
17]. In addition, known HR-GNR positive patients are flagged in the Hospital Information System (HIS) and isolated when readmitted. Despite screening and isolation of high risk patients, numerous patients are found unexpectedly HR-GNR positive in clinical cultures to diagnose a possible infection [
18]. Before detection these unexpected positive patients were not nursed in isolation so that transmission to other patients may have occurred since no specific infection control measures were taken except standard hygiene procedures.
Willemsen et al. showed that the nosocomial transmission rate of HR-GNRs in Dutch hospitals was 7.0%, using AFLP (Amplification Fragment Length Polymorphism) to determine the genetic relation between clinical isolates [
18]. Tschudin-Sutter et al. showed that nosocomial transmission from unexpected ESBL positive patients to contact patients rarely occurs, with a transmission rate of 2.2% over a total study period of 11 years [
19]. Based on these studies it could be questioned if contact tracing within hospitals with standard hygiene precautions is required as contact tracing is considered time consuming and expensive. For the development of future health policies the results of such studies are of major importance.
In the present study, the nosocomial transmission rate from unexpected HR-GNR positive patients to contact patients was studied within three regional hospitals in the Dutch region Kennemerland. In addition, we estimated the overall HR-GNR incidence including patients who were screened and pre-emptively isolated at admission.
Discussion
During a study period of 1 year, a nosocomial transmission rate of 0% from unexpected HR-GNR positive patients to contact patients was found. Out of 152 unexpected HR-GNR positive patients, 35 patients met our inclusion criteria for index patients. Around these 35 index patients, 69 contact patients were sampled, accounting for a total of 178 contact days. Although no nosocomial transmission had occurred, five contact patients were HR-GNR positive (7.2%) and four of these were ESBL
E. coli positive (5.8%), which corresponds with earlier reported prevalence rates in Dutch hospitals [
6,
27]. As expected, ESBL positive patients were found most frequently among all HR-GNR positive patients (68.6%) and index patients (68.6%). From a micro-organism perspective, 85.7% of the index patients were positive for an HR-GNR
E. coli. MRSA, VRE (Vancomycin-resistant
Enterococcus) and PRSP (Penicillin-resistant
Streptococcus pneumoniae) were not included in the present study.
Other studies with comparable study designs that estimated the transmission rate to contact patients are scarce, limiting the comparison with other settings. Willemsen et al. showed that the nosocomial transmission rate of HR-GNR in Dutch hospitals was 7.0% [
18]. This is probably a worst-case scenario since only epidemiologically linked clinical isolates within a time window of 4 weeks were analysed using AFLP genotyping, which is considered less discriminatory. In 2012, Tschudin-Sutter et al. studied the transmission rate from unexpected ESBL positive patients to contact patients [
19]. Their results showed that during a period of 11 years two contact patients related to 93 index patients (2.2%) were ESBL positive by transmission, suggesting that nosocomial transmission rarely occurs. A study performed at the ICU in a French hospital showed an ESBL acquisition rate of 6.5% [
28]. However, only one patient (out of 19) appeared to be positive by nosocomial transmission. A complicating factor for these studies (and also for our study) is the relatively high ESBL colonization prevalence in the community. We only detected 172 of these patients during our study period instead of an expected amount of 4450 admissions with HR-GNR positive patients. Consequently, expensive high resolution genotyping is needed to exclude transmission since phenotypic results, MLST, AFLP or ESBL gene are not able to discriminate enough between closely related isolates [
25,
29]. Based on MLST and ESBL group alone we would have concluded that transmission had occurred between one index and a contact patient. Additional cgMLST or wgMLST analyses, as performed in the present study minimizes the chance on this kind of false conclusions. Another interesting study within a German hospital showed a nosocomial transmission rate of 2.3% for multidrug resistant
E. coli based on clinical (infection) isolates using cgMLST [
30]. When isolation measures of positive patients were ceased the transmission rate increased non-significantly to 5.0% and decreased on high risk wards (ICU). However, as these results were based on clinical infection cultures only, colonized (not infected) patients were missed, underestimating the real transmission rate.
Our results and the previously mentioned studies clues that routine contact tracing in case of an unexpected HR-GNR positive patient might be replaced by appropriate surveillance in a local setting within hospitals with standard hygiene precautions. Also, since we found no nosocomial transmission, these results advocate a more flexible isolation strategy. However, (cluster) randomized controlled trials are needed to compare nosocomial transmission rates between different isolation strategies. Preferably such studies must be accompanied by adverse events that are associated with isolation (such as patient well-being) so that a balanced conclusion could be made. Because we mainly isolated HR-GNR E. coli, our results should be interpreted with caution and cannot simply be generalized for less frequently isolated HR-GNRs such as CPE or other micro-organisms than E. coli such as K. pneumoniae.
Some studies have suggested that certain sequence types of
E. coli (ST 131) and
K. pneumoniae (ST 258) are hyperendemic, causing outbreaks and infections [
31,
32]. A recent review found evidence that
E. coli ST 131 is more pathogenic than non-ST131, but the increased transmissibility or prolonged carriage could not be confirmed [
33]. For
K. pneumoniae ST 258, this study could not confirm or reject the increased pathogenicity, transmissibility or prolonged carriage of this sequence type [
33]. As certain HR-GNR types or micro-organisms are potentially more dangerous in terms of transmissibility or pathogenicity, contact tracing can only be replaced in a local setting within hospitals where adequate standard hygiene precautions with sufficient surveillance or prevalence measurements are performed. Prevalence measurements will provide insight into local HR-GNR epidemiology and possible ongoing transmission within hospitals [
27]. Appropriate surveillance could be performed by reviewing (1) clinical HR-GNR isolates, (2) patient admission data and (3) genotyping of HR-GNR isolates when transmission is suspected as performed by Mellmann et al. [
30].
Comparing our overall cumulative HR-GNR incidence rate with the study of Willemsen et al. showed a lower cumulative incidence rate per 10,000 admissions (28 vs. 39) [
18]. An explanation for this difference could be the large variation between hospitals, hospital types and patient populations that were included in both studies. Comparing the incidence density per 100,000 patient-days between both studies showed a higher incidence density in our study (77 vs. 55) [
18]. The mean length of stay in our study was 3.6 days compared to 6.6 days, resulting in a lower denominator of patient-days. This decreasing trend of mean length of stay within Dutch hospitals was also noticed in a Dutch report published in 2013 [
34].
The present study has several limitations. First, the sample size (35 index and 69 contact patients) was relatively small which is reflected by the large confidence interval of the calculated transmission rate. Future studies are necessary to confirm our results. Second, for VRE (Vancomycin-resistant
Enterococcus) it is known that the inoculum size is related to the detection probability with culturing [
35]. For HR-GNR detection it is largely unknown how much time between colonization and sampling (using culturing) is sufficient. This may have resulted in a possible underestimation of the nosocomial transmission rate in our study, as some patients could have been marked as false negative. However, the median time between start of contact and sampling in our study was 3 days (median contact time plus time between end of contact and sampling), and we therefore do not think that this has markedly influenced our results. Future studies must incorporate repeated culturing after the end of contact in order to determine the optimal culturing strategy. Third, our results cannot be solely attributed to the transmission capacity of HR-GNR type or micro-organism alone. In a setting, with other prevalence rates or infection control policies other transmission rates could be found. Fourth, we have possibly missed cases of transmission, since only admitted index and contact patients were included in our study.