High prevalence of multidrug-resistant Gram-negative bacteria carriage in children screened prospectively for multidrug resistant organisms at admission to a paediatric hospital, Hamburg, Germany, September 2018 to May 2019

Background Increasing resistance to antibiotics poses medical challenges worldwide. Prospective data on carriage prevalence of multidrug resistant organisms (MDRO) in children at hospital admission are limited and associated risk factors are poorly defined. Aim To determine prevalence of MDRO carriage in children at admission to our paediatric hospital in Hamburg and to identify MDRO carriage risk factors. Methods We prospectively obtained and cultured nasal/throat and inguinal/anal swabs from children (≤ 18 years) at admission between September 2018 and May 2019 to determine prevalence of meticillin-resistant Staphylococcus aureus (MRSA), multidrug-resistant Gram-negative bacteria (MRGN) and vancomycin-resistant enterococcus (VRE) and associated species. We collected medical histories using a questionnaire and evaluated 31 risk factors using logistic regression models. Results MDRO carriage prevalence of 3,964 children was 4.31% (95% confidence interval (CI): 3.69–5.00). MRSA carriage prevalence was 0.68% (95% CI: 0.44–0.99), MRGN prevalence was 3.64% (95% CI: 3.07–4.28) and VRE prevalence 0.08% (95% CI: 0.02–0.22). MDRO carriage was associated with MRGN history (odds ratio (OR): 6.53; 95% CI: 2.58–16.13), chronic condition requiring permanent care (OR: 2.67; 95% CI: 1.07–6.13), antibiotic therapy (OR: 1.92, 95% CI: 1.24–2.94), living in a care facility (OR: 3.34; 95% CI: 0.72–12.44) and refugee status in previous 12 months (OR: 1.91; 95% CI: 0.27–8.02). Compared to established practice, screening using risk-factors had better diagnostic sensitivity (86.13%; 95% CI: 80.89–91.40) and specificity (73.54%; 95% CI: 72.12–74.97). Conclusion MRGN carriage was higher than MRSA and VRE. Extended risk-factor-based admission screening system seems warranted.


Introduction
Worldwide, multidrug-resistant organisms (MDRO) are an increasing medical and economic burden [1]. In particular, the increasing antibiotic resistance in Gramnegative bacteria poses major challenges to medical care [2,3]. Although meticillin-resistant Staphylococcus aureus (MRSA) colonisation shows a decreasing prevalence, the occurrence of multidrug resistant Gramnegative bacteria (MRGN), such as extended spectrum beta-lactamase (ESBL) Escherichia coli and Klebsiella pneumoniae, is increasing [2]. Different international prevalence estimates and risk factors for MDRO result in different strategies for MDRO management [4][5][6][7][8]. For example, our hospital isolates patients only after receiving a positive test result for MDRO, whereas other countries isolate newly admitted patients until a negative test result is received. While standardised MRSA screening is established in most European hospitals, the focus has changed to MRGN screening in adult patients in recent years [2,6,[9][10][11]. Although recent studies published between 2013 and 2018, from Europe and the United States (US) have reported transmission routes for MRGN colonisation, such as direct person-to-person transmission, indirect transmission through contaminated areas or materials (e.g., kitchen towels), antibiotic use, contact with healthcare facilities in high MDRO prevalence countries and travel history to Africa and Asia, most risk factors seem unknown [2,6,[12][13][14]. Because of belated or incomplete access to professional medical history at admission, MDRO detection is often delayed, which can increase morbidity and mortality [1].
With respect to MDRO prevalence, prospective data are scarce for children and adolescents [20]. In addition, changes in prevalence of MRSA and MRGN have not been adequately investigated. Furthermore, no uniform screening standards for MDRO detection in children have been established [6].
This study aimed to investigate the prevalence of MDRO carriage (MRSA, MRGN and VRE) at admission to the Altona Children's Hospital (AKK) in Hamburg, Germany for scheduled or emergency hospital admissions as well as to determine clinical, epidemiological and microbiological risk factors associated with MDRO carriage. This information was used to critically evaluate the current MDRO screening at admission.

Study design
This prospective observational study was performed at one hospital. After written informed consent was obtained, participants were recruited over 9 months, between September 2018 and May 2019, at the AKK. With about 12,000 in-patient visits per year, the AKK is one of the largest paediatric care facilities in Germany, attracting patients from Hamburg and surrounding areas as well as from across Germany. The study included all in-patients (≤ 18 years, hereafter referred to as children) with scheduled or emergency admission from every department: general paediatrics, intensive care unit (ICU), neonatology, neurosurgery, otorhinolaryngology, orthopaedics, surgery, trauma surgery, urology, and Lufthafen, a special unit for children requiring permanent ventilation. Patients were excluded from the study if they were re-admitted to our clinic after initial recruitment.

Data and clinical sample collection
Pooled smear tests were taken from typical regions (nasal/throat and inguinal/anal) to determine MDRO occurrence (MRSA, MRGN and VRE) and associated species. In general, the samples were taken at admission. In special emergency situations, samples were taken after the patient was in a stable condition. Additionally, we collected clinical, epidemiological and microbiological data using a questionnaire for parents and attending physicians, which was completed during admission. The questionnaire was available in Arabic, English, French, German, Persian (Farsi), Russian and Turkish to ensure inclusion of different ethnicities. Based on recommendations from the German Commission for Hospital Hygiene and Infection Prevention at the Robert Koch Institute (KRINKO) and on questionnaires from previous European studies [6,[13][14][15][16]21], we evaluated six general demographic factors: (i) sex, (ii) age, (iii) department, (iv) emergency, (v) transfer (patient transferred from another hospital) and (vi) country of birth of the child and parents, and 25 specific factors (see Supplementary Tables S1-3).
The patient questionnaire already in use and evaluating 17 risk factors was modified by adding eight additional risk factors: (i) history of VRE, (ii) contact with a family member treated at ICU for more than 7 days the previous 12 months, (iii) having a pet, (iv) being breastfed within the previous 12 months, (v) major surgery the previous 12 months, (vi) urogenital anomalies or recurrent urological infections, (vii) family member working in healthcare facilities with regular patient contact and (viii) stay at a neonatology unit for more than 7 days within the previous 12 months. For some questions (11 of 25), if the answer was affirmative, one to four sub-questions needed to be answered (a total of 18 sub-questions) -e.g., duration of treatment (see Supplementary Tables S1-3  and attending physicians needed approximately 3 to 5 minutes to complete the whole questionnaire. Some information was derived from the medical history recorded during admission, including information from the physical examination.

Laboratory procedures
Pooled smear tests were taken using the eSwabÔ (Copan, Brescia, Italy) collection and transport device consisting of Nylon Flocked Swabs and 1ml of Liquid Amies. Most of the smears were taken by the nursing staff and rarely by doctors. One swab was used for sampling the throat and then the nasal cavity, and another swab was used for sampling the inguinal and then the anal region. The swabs were immersed in 1mL of Liquid Amies (eSwabÔ Copan, Brescia, Italy) until processed. Smears were analysed in the same diagnostic laboratory for the entire study period. In the laboratory, nasal/throat swabs were incubated on culture media for MRSA (MRSA Chromagar, Biomerieux, Nuertingen, Germany) and ESBL (ESBL Chromagar, Biomerieux (resistant against piperacillin, cefotaxim and/or ceftazidim and ciprofloxacin), or 4MRGN (3MRGN plus resistant against imipenem and/or meropenem) [22].

Statistical analysis
In 2013, Wegner et al. found an MDRO (MRSA, MRGN and VRE) prevalence of 3.04% in adults [23], and a slightly lower prevalence was expected in children [5,8,15]. Sample size estimation showed that based on 4,000 patients, a prevalence of at least 2.34% could be estimated -based on a 5% significance level and the conservative assumption that precision is at one-fifth of the expected prevalence [24,25]. Hence more than 4,000 patients were recruited (4,092).
The primary endpoint was MDRO carriage prevalence at admission to the AKK, and the secondary endpoint was the prevalence of individual pathogens (MRSA, MRGN and VRE). The 95% confidence intervals (CI) for these prevalence estimates were derived based on Clopper and Pearson [26]. To visualise the trend in prevalence over age groups, a locally weighted scatterplot smoothing (LOESS) estimate was used. Associations between individual risk factors and the primary endpoint, MDRO, as well as with the secondary endpoint, MRGN, were assessed with two multivariable logistic regression models. From these, odds ratio (OR) estimates along with 95% CIs were derived. Including all potential risk factors in the model resulted in the estimation of the direct effects of each of them, independent of the other factors. As analyses were based on all available cases, 3,851 (97.1%) observations contributed to the prevalence estimate of the primary and secondary endpoint. Sensitivity and specificity of the original screening survey were assessed as in the extended version and were reported along with 95% Wald CIs.

Characteristics of the study population
During the study period, 4,092 children were recruited, but 124 patients had to be excluded as they were recruited twice during the study period and four additional children were excluded as data from their questionnaires could not be evaluated. As a result, 3,964 patients were included in the analyses (Table 1) Tables S1-3).
Of the 2.8% of participants with a previous positive history of MDRO, more than half had been tested positive for MRGN (1.8%) before the current study. In addition, a refugee status within the previous 12 months (n = 27) and living in a care facility (n = 31) resulted in a similar positive response rate. A higher proportion of participants indicated that they had a disability requiring permanent care (5.4%) or had received antibiotic therapy within the previous 6 months (16.6%). Other characteristics are shown in Table 2 (Supplementary  Tables S1-3).

Comparison of screening algorithms at admission
Based on the AKK established admission screening using 17 risk factors, 109 (65.7%) of 166 MDRO cases would have been detected, as they reported at least one risk factor resulting in MDRO screening (

Sex-, age-and department-specific MDRO prevalence
The prevalence of MDRO was similar in females (4.01%) and males (4.33%). Estimates of age-related prevalence revealed that MDRO and MRGN prevalence tended to be higher in the first years of life but also in adolescence ( Figure 1B). MRSA carriage prevalence was constant over all ages ( Figure 1B). The highest prevalence of MDRO carriage was observed in patients in the ICU (21.74%), the Lufthafen (the special unit for children on permanent ventilation) (11.11%) and the neurosurgical ward (9.38%) ( Table 1).   Odds ratio estimates were derived from logistic regression models as were 95% confidence intervals for MDRO and MRGN colonisation.
Results based on 3,017 complete observations.
In the figure, the risk factors are briefly summarized as keywords.

Associations between risk factors and MDRO occurrence
The most important risk factor for MDRO was MRGN history. An OR of 6.53 (95% CI: 2.58-16.13) indicated that the odds for MDRO occurrence in patients with previously reported MRGN was more than six times higher than for a patient without a previous MRGN history (Figure 2 Figure 2).

Discussion
In this study, we investigated the prevalence of MDRO carriage at admission to AKK, a major paediatric hospital in Hamburg, for scheduled or emergency admissions over 9 months.
The MRGN prevalence of 3.64% (95% CI: 3.07-4.28) in this study was higher than the observed prevalence in another German study examining adults (1.65%) [9] and similar to a recent Dutch study examining adults (5.0%, 95% CI: 3.4-6.6) [2]. Recent European studies have found an increasing ESBL carriage prevalence in preschool children, from 3.5% (95% CI: 2.5-4.8%) [13] to 16.8% [30], much greater than the MRSA carriage prevalence in children, from 1.2 [13]  . As screening all in-patients is not reasonable because of low prevalence and high screening costs, the best option might be an extended risk-factorbased admission screening system, as this increases the detection rate of MDRO carriers, secures adequate therapies and reduces health costs [12,34]. Moreover, it is also possible to use the extended risk-factorbased admission screening routinely at emergency units as the majority of the questions are relevant to the patient's medical history, which would need to be reported in any case. As a consequence of this study's findings, the extended screening system is now being established at the AKK.
When classifying MRGN cases according to antibiotic resistance, we primarily detected 2MRGN, followed by 3MRGN and 4MRGN. The detected carbapenemase of the OXA-48 group in 4MRGN K. pneumoniae and of the OXA-23 group in 4MRGN A. baumannii corresponds to the predominance in Europe [18] and Germany. According to KRINKO guidelines, 2MRGN colonisation is relevant in paediatric ICUs and other risk zones, such as neonatology units because of mandatory isolation and their same therapeutic status as 3MRGN [22]. Therefore, we also had the highest MDRO prevalence in our clinic risk zones, the ICU and the Lufthafen (Table  1).  [38], whereas sepsis caused by Enterobacter spp. and Escherichia spp. is associated with a high risk of mortality in the very low birth weight paediatric population [39].
Overall, this study had a high response rate: of the 4,092 patients recruited, 3,851 (97.1%) were included in our analyses. Although we generated high quality data, some questionnaires were incomplete, especially on days with a high patient influx at the clinic and thus some important risk factors, including patient/family member with treatment in an ICU, antibiotic treatments, parents working in healthcare facilities and country of birth, may have been missed. Because of the low number of cases or missing answers, five factors (country of birth of the child and parents, department, VRE history, dialysis and treatment of a family member in an ICU) and the 18 sub-questions could not be included in any of the logistic regression models. Data obtained from adolescent patients were generally more incomplete than from their younger counterparts as some adolescents refused anal swabs. This situation generated higher numbers of unknown anal and inguinal colonisation events, probably resulting in several missed MDRO cases. As previous studies have shown, taking a rectal swab has a higher certitude of detection than anal swabs. Because we already had a lower response for adolescence anal swabs, we would expect an even lower compliance if rectal swabs had been required. In addition, our results could have been influenced by the low number of participants who had refugee status (n = 27) [1] or who lived in a care facility (n = 31) [2], but it seems realistic to assume that the refugees are a minority among children being admitted at AKK. If the proportion of these had been higher, we would probably have had higher MDRO prevalence rates as other studies in Germany have shown among refugees and residents of care facilities [9,21,23,37,40].
Unfortunately, due to financial limitations, it was not possible to perform MDRO subtyping or to recruit participants over the entire year. Therefore, with regard to the described seasonal accumulations of, for example, ESBL occurrence in August and September, a year-round study would probably have shown an even higher MRGN prevalence [36].
Screening of parents was not performed and their medical history did not explicitly include questions about colonisation even though, as reported by van den Bunt et al. in 2016, parents of MDRO-positive children might also be carriers [31]. In any case, the source of parental colonisation would most likely be unknown. However, MDRO and especially ESBL transmission via contaminated items such as food and/or cloth (e.g., kitchen towels) has also been described as an MDRO risk factor [2,12].

Conclusion
Children and adolescents admitted to the AKK most likely represent the paediatric populations of Germany. This study has shown that the prevalence of colonisation of children with MRGN bacteria is higher than MRSA and VRE occurrence and that existing screening procedures can be optimised to decrease undetected MDRO colonisation events. Currently, the best option to detect MDRO in children from northern and western parts in Europe (or comparable low prevalence settings) might be an extended risk-factor-based admission screening system that considers healthcarerelated risk factors. Therefore, there is a high need for further European studies on the modification of admission screening algorithms with a focus on MRGN in preschool children as this was the age group with the highest prevalence in our study.