Burden of Hospital Acquired Infections and Antimicrobial Use in Vietnamese Adult Intensive Care Units

Background Vietnam is a lower middle-income country with no national surveillance system for hospital-acquired infections (HAIs). We assessed the prevalence of hospital-acquired infections and antimicrobial use in adult intensive care units (ICUs) across Vietnam. Methods Monthly repeated point prevalence surveys were systematically conducted to assess HAI prevalence and antimicrobial use in 15 adult ICUs across Vietnam. Adults admitted to participating ICUs before 08:00 a.m. on the survey day were included. Results Among 3287 patients enrolled, the HAI prevalence was 29.5% (965/3266 patients, 21 missing). Pneumonia accounted for 79.4% (804/1012) of HAIs Most HAIs (84.5% [855/1012]) were acquired in the survey hospital with 42.5% (363/855) acquired prior to ICU admission and 57.5% (492/855) developed during ICU admission. In multivariate analysis, the strongest risk factors for HAI acquired in ICU were: intubation (OR 2.76), urinary catheter (OR 2.12), no involvement of a family member in patient care (OR 1.94), and surgery after admission (OR 1.66). 726 bacterial isolates were cultured from 622/1012 HAIs, most frequently Acinetobacter baumannii (177/726 [24.4%]), Pseudomonas aeruginosa (100/726 [13.8%]), and Klebsiella pneumoniae (84/726 [11.6%]), with carbapenem resistance rates of 89.2%, 55.7%, and 14.9% respectively. Antimicrobials were prescribed for 84.8% (2787/3287) patients, with 73.7% of patients receiving two or more. The most common antimicrobial groups were third generation cephalosporins, fluoroquinolones, and carbapenems (20.1%, 19.4%, and 14.1% of total antimicrobials, respectively). Conclusion A high prevalence of HAIs was observed, mainly caused by Gram-negative bacteria with high carbapenem resistance rates. This in combination with a high rate of antimicrobial use illustrates the urgent need to improve rational antimicrobial use and infection control efforts.


Introduction
Hospital-acquired infections (HAIs) and antimicrobial resistance are growing global public health problems [1,2]. The incidence of HAIs is substantially higher in Low and Middle Income Countries (LMICs), with an average prevalence of 15.5%, compared to prevalence of 7.1% and 4.5% in Europe and USA, respectively [3]. This problem is more serious in intensive care units (ICUs). The HAI prevalence in ICUs ranges from 9.1% in the United States to about 23.0%-23.5% in Europe and England [4][5][6][7], and even higher in LMICs with a pooled prevalence of 35.2% [1]. A recent report of the International Nosocomial Infection Control Consortium 2007-2012 from 503 ICUs shows that ventilator-associated pneumonia is fifteen times and catheter-associated urinary tract infection four times higher in LMICs than in better resourced settings [8]. Due to economic development in LMICs, the healthcare systems are changing rapidly, with increasing ICU capacities. However resource constraints often result in high occupancy rates, crowding, a lack of isolation facilities, and insufficient resources for adequate infection control all of which may contribute to the reported high incidence of HAIs and drug-resistant infections at ICU's in these settings [1, 9,10].
Vietnam is a LMIC with a population of 90.796 million [11] and an increasingly sophisticated health care system, typical of countries in the region. Health expenditure per capita in Vietnam was around 100$ per annum in 2012, approximately a seventh of the regional average [11]. Up to now, there is no national surveillance system for HAIs and limited data about HAIs in ICUs. The few studies performed are small and only some include ICUs, but reported that the HAI prevalence in those ICUs ranged from 19.3% to 31.3% [12][13][14][15][16][17]. Only one of these studies is from the international peer reviewed literature [12], the others are published in the Vietnamese medical literature. Antimicrobial resistance levels are high in Vietnam; up to 70% of Enterobacteriaceae were resistant to 3 rd generation cephalosporins and > 40.0% of Acinetobacter spp. resistant to carbapenems in 2009 [9,18].
In order to provide up-to-date, systematic data and to demonstrate the feasibility of developing a national surveillance network for ICUs in a LMIC, we performed a prospective study on the prevalence of HAI in ICUs across Vietnam, exploring risk factors, antimicrobial use, and antimicrobial resistance [19].

Study design, hospital and patient selection
We conducted a repeated point prevalence survey (PPS) to determine the prevalence of HAIs, and to assess antimicrobial use and antimicrobial resistance using the methodology developed by the European Center for Disease Prevention and Control (ECDC) [20]. The survey was conducted on one day each month from October 2012 through September 2013 in 15 adult ICUs in 14 acute care hospitals, of which 7 were tertiary hospitals and 7 provincial hospitals, throughout Vietnam (Fig 1). Patients aged 18 years, admitted to participating ICUs before 8 a.m. on the survey day, and remaining there at the survey time were included regardless after that time patient was discharged or remain in that ICU.

Data collection
The following patient data were collected: reason for admission, location of patient at admission to ICU, comorbidity, current interventions, involvement of patient's family in patient care (participating in bathing, cleaning, changing position, and feeding patients), antimicrobial agent use for any indication, presence of HAI according to ECDC definitions [20], and results of routine microbiological investigations.
All participating hospitals provided data on basic infrastructure and infection control indicators at the beginning of the study, including total number of beds, rooms, single bed rooms, number of doctors and nurses at the ICUs, admissions per year, patient days per year, alcohol hand rub consumption, and availability of alcohol hand rub at ICU bed.
All participating hospital laboratories were trained to follow the Clinical and Laboratory Standards Institute guidelines (CLSI) for antimicrobial susceptibility testing and were enrolled in an external quality assurance program (The United Kingdom National External Quality Assessment Service (UK NEQAS) for Microbiology) [19]. All participating doctors received training on the protocol, HAI definitions, how to complete case record forms, and data entry before the study. A database adapted from HELICSWin.Net developed by the ECDC was used for the study [21]. Each ICU received a laptop with the software installed and Vietnamese instructions. The anonymised data were submitted to the project coordinators electronically once a month during the implementation period. Data were then checked for missing or inconsistent data, with regular queries sent to the hospitals and visits by study investigators to assist in data reconciliation.

Statistical analysis
For descriptive statistics we calculated percentage, frequency, mean, and median values and 95% confidence interval (95% CI) or interquartile ranges (IQR), as appropriate. Where patients were included in more than one PPS, we only used the data from the first survey and had no HAI at ICU admission to calculate the odds ratios of risk factors for developing HAIs in ICU. After univariate analysis, we included all HAI risk factors in the multivariate logistic regression models. We used IBM SPSS Statistics software (version 22 IBM, California, USA) for data analysis. P-values < 0.05 (two-sided) were considered statistically significant.

Ethical considerations
The Ethical Committee of the National Hospital for Tropical Diseases (27/HDDD-NHTD) had approved of the study and confirmed that the need for informed consent was waived due to data were anonymous and collected by surveillance and no intervention was conducted; and the study was also approved by the Vietnamese Ministry of Health (4921/QD-BYT).

Patient characteristics
In total, 3,401 patients were screened, of whom 3287 (114 were excluded due to age < 18 years) were enrolled from 15 ICUs of 14 participating hospitals between October 2012 and October 2013. Due to prolonged ICU stay, 162 patients were enrolled in more than one survey leaving 3125 unique patients.
The  Table B in S1 File.

Discussion
The HAI prevalence in this study (29.5%, 965/3266 patients) is higher than that reported from adult ICUs in European hospitals using the same protocol (23.0%, 1750/7613 patients) [5] and in ICUs in Southern Europe, Turkey and Iran (23.5%, 176/749 patients) [7] and much higher than rate of 9.1% (156/1707) in ICUs in the United States in 2011 [4]. These differences can be explained by several factors: There were few single rooms in ICUs and common rooms usually contain 4-5 beds and even up to 30 beds; In addition with low healthcare staff ratios [11] and high bed occupancy rates [22], often over 110%, makes successful infection control very challenging. The lack of nursing staff for patient care allows little time for proper infection control measures, which may lead to an increased HAI rate [23][24][25]. This is supported by the smaller amount of alcohol hand rub used: a median of 66.4 ml (IQR: 23.7-100.5 ml)/patient day, compared with a median of 83 ml (IQR: 64-105)/patient day in ICUs of Germany in 2010 [26]. However, the HAI prevalence in Vietnamese ICUs was lower than the pooled prevalence of 35.2% in other LMICs [3].
Hospital-acquired pneumonia (HAP) was the most common type of HAI as in other studies [1, 3,7]. HAP accounted for 79.4% of all HAIs in our study, nearly double that reported (40.0% -45.3%) in ICUs of developed countries [5,6]. This is due to a large proportion of HAP (37.4%, 301/804) was acquired in the same hospital before admission to ICU. This figure raises the need for effective infection control program outside ICU for HAP prevention. Blood stream infections (BSI) accounted for 4.4% of all HAI, considerable lower than the reported 18.0% in European ICUs [5]. Potentially this is due to the high HAP prevalence, but there might also be an underestimation of BSI due to underutilization of blood cultures; as an indication there were a total of only 243 blood culture samples taken from the 3287 patients in the 24 hour preceding the survey day, we are not aware of any comparable data from other settings.
Multivariate analysis identified intubation as an important risk factor, which combined with the high prevalence of HAP suggests that future interventions should target this risk factor. Potential interventions to assess include ventilation and sedation strategies, cuff and tracheal tube design, cuff pressure management [27] Strategies like selective gut decontamination using polymyxins may not be appropriate in this setting with high antimicrobial resistance background rates with colistin as last resort drug [28]. Multivariate analysis revealed that the involvement of the patient's family in patient care was a protective factor from HAI. As ICUs had departmentwide policies to either allow or forbid family from taking care of their admitted family member this practice was highly clustered at the ICU level, the result needs to be interpreted with caution. Further research is needed to understand the risks for HAI regarding family members involved in patient care. National surveillance systems for HAI are scarce in LMIC settings yet ongoing surveillance is crucial to informing policymakers of the needs of the population and high quality data is critical to the development of potential interventions applicable at the national level.
The proportion of HAIs related to medical devices were higher than that reported by a European survey in 2011-2012, where device related HAIs were 59.5% for UTI, 57.3% for primary BSI, and 33.2% for HAP [5]. This may be partly explained by less compliance with hand hygiene [25] and low alcohol hand rub consumption, but probably also sub-optimal routines for insertion and care of urinary and blood catheters.
Nearly 85% of the ICU patients were on antimicrobials in this study, higher than the 56.5% and 77.3% reported in European and American ICUs, respectively [5,29]. Broad-spectrum beta-lactam antibiotics (3 rd , 4 th generation cephalosporins and carbapenems) and fluoroquinolones accounted for 39.7% and 19.4% of total antimicrobials use in Vietnamese ICUs, respectively. These proportions are higher than those reported by European ICUs of around 30% and 10% respectively [5]. Also combination therapy in our study is more common than in European ICUs: 62.5% versus 47.7% [5].
Gram-negative bacteria were the most common cause of HAI, similar to studies in mainly low and middle income countries [3,7,23]. Furthermore, the proportion of infection caused by Acinetobacter in ICUs was highest in Asia (19.2%), which was more than five times in compared with North America (3.7%) [30]. The two most frequently isolated bacteria were A. baumannii and P. aeruginosa (accounted for 38.2% of total isolates), similar to a study in Turkey in 2007-2008 (34.1% of isolates) [31]. Carbapenem resistance in A. baumannii, P. aeruginosa, and K. pneumoniae in this study was higher than rates of 69.1%, 44.4% and 2.9%, respectively, in Vietnam in 2007 and 2008 [25]. High prevalence of carbapenem resistance leads to an increased colistin use [9] and emergence and spread of colistin resistance will follow. Antimicrobial stewardship programmes and the deployment of alternatives to carbapenems may help to slow the development of further resistance to the remaining active antibiotics [32] and need to be evaluated in this context. The high rates of MRSA and glycopeptide resistant enterococci are also a major concern. However, the burden of Gram-positive infections as a cause of HAI is relatively low as compared to Gram-negative infections.
A potential limitation of this study is the paucity of microbiology data for many HAIs due to the few cultures taken in routine clinical care. However the microbiology results available represent valuable data for assessing the microorganisms causing HAIs and its susceptibility to antimicrobials in our setting. As HAIs had not previously been routinely surveyed in Vietnamese hospitals, criteria for the diagnosis of specific HAI may not have been fully understood, potentially leading to misclassification of specific HAIs. To limit this problem, we organized several workshops and site visits to train doctors. The enrolment of patients who were just admitted to ICU before 8.00 A.M. then were discharged from ICU soon after survey time could lead to underestimate the HAI prevalence. However, the mean ICU stays were much longer (4.8 to 16 days, Table A in S1 File) means that number of these cases was not many and its impact on the HAI prevalence was not significant. HAIs also are a sensitive issue in Vietnam; therefore ICU doctors may not have reported all HAIs, leading to an underestimation. To minimize this effect data, hospitals were anonymized using codes. When data were uploaded we ran data checks and returned queries to doctors that needed to be resolved. We therefore think that the hands-on supervision and input from the study team ensured that data quality was optimal.

Conclusions
High prevalence of HAIs in Vietnamese ICUs, mainly caused by Gram-negative bacteria with high rates of carbapenem resistance, and high levels of antimicrobial use illustrate the urgent need for capacity strengthening in both rational antimicrobial use and infection control efforts at national, regional and local levels.
Supporting Information S1 File. Supplementary Tables. Hospital and ICU Characteristics (Table A), Patient Characteristics (Table B), HAI Prevalence per Month (Table C), Antimicrobials Combinations Used (Table D), The Common Antimicrobial Agents Used (Table E) (DOCX)