Risk factors of multidrug-resistant bacterial infections among patients admitted in a tertiary care hospital of north India

Introduction: Infections with drug-resistant organisms (DRO) have been associated with poor patient outcomes. To tackle this global problem, it is necessary to understand the risk factors that predispose to infections with DRO. Methodology: This was a prospective observational study conducted over a three-year period at a tertiary-care hospital. Bacterial culture isolates from patients admitted in medicine wards with community or hospital-acquired infections were included. Logistic regression analysis was used to determine risk factors for drug-resistant infections. Results: Of the 295 patients with 323 isolates included, 40 (12.3%) had non-MDR (N-MDR) infections, 86 (26.6 %) had MDR infections and 197 (61%) had possible extensively drug-resistant (P-XDR) infections. History of previous admission in the preceding three months (Odds Ratio, OR = 4.53, 95% Confidence interval, CI = 1.8 – 11.42, p = -0.01), high SOFA score at admission (OR = 1.14, 95% CI = 1.0 – 1.290, p = -0.039) and prolonged duration of ventilation (OR = 1.25, 95% CI = 1.05 – 1.41, p = -0.012) were independently associated P-XDR infections when compared to patients with N-MDR. Conclusions: High rate of multidrug-resistant infections in the studied area is alarming. In this single-centre study, we elicited various risk factors for drug-resistant bacterial infections ranging from patient characteristics to iatrogenic risk factors during the hospital stay. Infections with P-XDR and MDR isolates independently increased hospital and ICU stay duration and were associated with increased mortality.


Introduction
The alarming rise in anti-microbial resistance (AMR) incidence is a paramount global health concern. Infections with drug-resistant organisms (DRO) have been consistently associated with poor patient outcomes. They also place a significant economic burden on the healthcare system [1,2]. To preserve the efficacy of currently available antibiotics without compromising patient care, it is necessary to understand the risk factors that predispose to infections with DRO.
Several studies from developed nations have tried to elucidate various risk factors for DRO acquisition [3,4]. Studies have shown that the risk of AMR depends on the inherent virulent potential of the microorganism, inappropriate antibiotic abuse, local environmental factors, and various patient-related factors [5]. However, due to regional differences in epidemiology, these results may not hold for developing countries like India [6]. Thus, regional and local data is essential for developing countries to understand the risk factors associated with infections due to DRO. Such data will facilitate the design and implementation of an effective anti-microbial stewardship program (ASP). It will also help select appropriate empirical antibiotic therapy in patients presenting with risk factors for infections with DRO.
This study has assessed the locally prevalent risk factors predisposing to drug-resistant infections and the clinical outcomes of patients admitted with these infections. The results of this study are intended to aid developing nations that are grappling with the problem of rising AMR by designing and implementing effective and targeted ASPs.

Study Design
This is a prospective observational study conducted in the department of medicine at a tertiary care apex teaching institute in New Delhi, India. This study was conducted on patients admitted in the medicine wards and intensive care unit (ICU) for 18 months after obtaining approval from the institute's ethics review board. Informed consent was taken from all participants before enrollment.
All patients admitted with a suspected infectious syndrome (community or hospital-acquired) whose body fluids were sent for microbiological cultures were assessed. Patients with any clinically significant culture growth were enrolled, and samples with at least one clinically significant culture isolate were included in the study. Clinically significant cultures were defined as microbial cultures with significant colony count as per infectious syndromes in question and without polymicrobial, contaminant, or commensal growth. Usually, a single sample was considered in each patient; however, multiple samples from the same patient were only considered from the same patient in the following scenario: 1. The samples were obtained from the same site (e.g., bronchoalveolar lavage) but with different clinical infectious episodes; 2. When samples were obtained from the same patients from two different sites (i.e. two different clinical infectious syndromes).
Cultures that showed fungal or mycobacterial growth were excluded from the study. Baseline patient data were collected from clinical records on the day of culture positivity and used for risk factor analysis. Recruited patients were followed up until discharge or death to evaluate clinical outcomes ( Figure 1).

Sample collection and anti-microbial susceptibility testing
Clinically relevant samples included sputum, endotracheal aspirate, bronchoalveolar lavage, blood, urine, and other body fluids. Samples were sent to the microbiology laboratory in sterile containers, except blood, which was inoculated manually in culture bottles at the bedside. Sample processing, direct demonstration of organisms, inoculation on culture media, and identification were made per standard microbiological practices. Anti-microbial susceptibility testing was done using the disc diffusion method (Kirby-Bauer test), and interpretation was made following recent Clinical and Laboratory Standards Institute (CLSI) guidelines [7].

Definitions
Various infectious syndromes were defined as per the CDC National Healthcare Safety Network surveillance guidelines, including ventilator-associated and hospital-acquired pneumonia (VAP, HAP), symptomatic and catheter-associated urinary tract infection (S-UTI, CA-UTI), primary and secondary bloodstream infections (BSI), skin/soft tissue infections (SSTI), etc. [8].
Culture isolates were then categorized based on the degree of drug resistance. A bacterial isolate was considered non-susceptible to an anti-microbial agent if tested resistant or intermediate when using clinical breakpoints as interpretive criteria, as provided by the CLSI guidelines [7]. Multi-drug resistance (MDR) was defined as non-susceptibility to at least one agent in three or more anti-microbial categories. Pan drug resistance (PDR) was defined as non-susceptibility to all agents in all anti-microbial categories. Extensive drug resistance (XDR) was defined as nonsusceptibility to at least one agent in all but two or fewer anti-microbial categories. However, in most microbiological laboratories in resource-limited settings, resistance testing is done only for commonly prescribed antibiotics. Hence, the terminology of "possible-XDR" (P-XDR) was coined. Bacterial isolates were defined as P-XDR when found resistant to most of the routinely tested classes of anti-microbials and susceptible to only one or two available categories of anti-microbials. P-XDR, however, should still be regarded as a marker of extensive resistance. All isolates which did not satisfy the criteria for MDR and P-XDR were classified as non-multidrug resistant (N-MDR) isolates (resistance to at least one agent in one or two different anti-microbial classes since the number of pan-susceptible isolates were few they were clubbed with N-MDR isolates [9].

Statistical analysis
The analysis was performed using STATA Version 12.1. Categorical variables were analyzed using the χ 2 test or Fisher's exact test and expressed as percentages. Continuous variables were analyzed using Student's ttest or Mann-Whitney U test and have been generally presented as means and standard deviation. Multivariate logistic regression analysis using the forward likelihood ratio selection method was done to identify independent risk factors of MDR/P-XDR infections and presented as odds ratios (95% confidence intervals, CI). All p-values were two-tailed, and p < 0.05 was considered statistically significant.

Results
During the study period, 295 patients with suspected infectious syndromes with 686 positive cultures were identified. After excluding cultures from the same clinical event as well as clinically insignificant cultures, 323 positive cultures from clinical samples belonging to 229 patients were included in the study ( Figure 2).
These 323 culture isolates were further classified as per the spectrum of infections they caused and according to resistance pattern as N-MDR (40, 12.3%), MDR (86, 26.7%) P-XDR (197, 61%) ( Table 1). The most common infection site with MDR and P-XDR organisms was the respiratory tract and urinary tract, respectively.

Patient characteristics
There was a preponderance of male patients (174, 54%) as compared to female patients (149, 46 %) included in this study. There was no significant difference in gender distribution between N-MDR,  MDR, and P-XDR groups. Most patients were found to be middle-aged with a mean age of 46 ± 17 years, whereas patients with P-XDR (48. 18 ± 18.18) infections were slightly older than other groups, but this was not statistically significant. The baseline laboratory parameters of patients belonging to the three groups corresponding to the day of culture isolation have been shown in Table 2. Patients who had significant anaemia, leukocytosis, renal/hepatic dysfunction, and hypoalbuminemia had a higher risk of getting P-XDR than patients in the MDR and N-MDR groups.

Previous hospital stay
Univariate analysis showed that previous hospital admission within the three months, previous ICU admission, and previous intravenous antibiotic usage were significantly associated with P-XDR and MDR isolates compared with N-MDR isolates, whereas the duration of PHS did not ( Table 3). The presence of comorbidities and as well composite co-morbidity index like Charlson's co-morbidity index (CCI) did not significantly correlate with the occurrence of MDR, P-XDR compared to NMDR. Sick patients at admission with high clinical severity scores (APACHE II, SOFA)  and presence of shock at admission were more likely to have drug-resistant infections (Table 3).

Current hospital stay
Patients with acute kidney injury, undergoing renal replacement therapy, prolonged hospital stay/ICU stay, prolonged mechanical ventilation and prolonged duration of in-situ central venous catheters and urinary catheters had a significantly increased risk of infection with an MDR or P-XDR isolate (p < 0.05, Table 3).

Multivariate analysis of risk factors
The multi-nominal regression analysis showed that those who had P-XDR infections were more likely to have a previous admission in another hospital in the preceding three months (Odds ratio, OR = 4.53, 95% CI = 1.8 -11.42, p = -0.01) and prolonged ventilator days duration (OR = 1.25, 95% CI = 1.05 -1.41, p = 0.012) ( Table 4).

Outcome analysis
Patients with P-XDR isolates require prolonged mechanical ventilation and ICU stay compared to patients with N-MDR and MDR isolates, as shown in Table 5. Also, patients with P-XDR isolates were more likely to die early, when compared with N-MDR and MDR patients. Logistic regression analysis with adjustment for significant confounders showed that duration of ICU stay and hospital stay was increased by 1.15 times (95% CI = -3.34 -5.65, p > 0.05) and 8.09 times (95% CI = 0.1 -16.07, p < 0.05) respectively in patients with MDR isolates as compared to those with N-MDR isolates, as shown in Table 6. Among patients with P-XDR infections, the total duration of ICU stay and hospital stay was increased by 6.87 times (95% CI = -2.79 -10.9, p = 0.01) and 8.4 times (95% CI = 1.17 -15.64, p = 0.023), respectively in patients with P-XDR isolates as compared to those with N-MDR isolates. Coefficient regression analysis showed mortality is 1.31 (p < 0.05) times higher with MDR and  Parametric data has been summarized as mean ± standard deviation and non-parametric data as median (interquartile range). N-MDR: Non-multidrug resistant; MDR: Multidrug resistant; P-XDR: Possibly extensively drug resistant.

Discussion
This large, prospective observational study of hospitalized patients with culture-proven infections was designed to identify risk factors for drug resistance in a developing country where such data has been historically sparse. In this present study, drug-resistant bacterial infections were classified as N-MDR, MDR, P-XDR, of which N-MDR culture isolates constituted only 12% of all culture isolates. Such a small proportion of isolates that were drug-sensitive/ resistant to less than two anti-microbial isolates are reflective of the current burden of AMR being faced in tertiary care hospitals [10,11].
Previous publications have implicated male gender and advanced age as risk factors for drug-resistant bacterial infections [12,13]. Although there is a similar trend, this association was not significant in our study. Similarly, various comorbidities are risk factors in other studies. Co-morbidities prolong the duration of hospital stay, thus increasing the period of exposure to drugresistant flora. They also impair patients' defences and render them more susceptible to infections [3]. However, neither individual's co-morbid illness nor the Charlson Comorbidity Index was found to increase risk in this study [14]. This could be because the severity of acute illness and adequate treatment of comorbid conditions during the hospital stay may have nullified any association with the risk of drug-resistant infections.
Various guidelines recommend considering the history of previous hospital stay (PH) and prior antibiotic exposure when selecting appropriate empirical anti-microbial therapy [4,15]. Commensal floras in hospitals face constant selection pressure due to the use of broad-spectrum antibiotics. These selected drug-resistant organisms colonize and ultimately cause infections in susceptible hosts with impaired defences [16][17][18]. Thus, patients referred from other hospitals or have had recent hospital or ICU stay are more likely to harbour drug-resistant organisms requiring higher antibiotics at the outset. A similar association was apparent in this current study between PHS history for more than 48 hours within the preceding three months, the previous ICU stays, and previous intravenous antibiotic use with an increased risk of acquiring infections with MDR and P-XDR organisms. Among these, the history of PHS within the preceding three months was an independent risk factor for acquiring P-XDR infections compared to N-MDR infections (OR = 4.5, 95% CI = 1.8 -11.42, p = -0.001).
Patients who are critically ill are highly susceptible to infections because of exposure to invasive procedures that compromise the anatomical barriers/defences, impairment of protective mechanisms such as airway protective reflex or gastric acid by sedative drugs or stress-ulcer prophylaxis and the frequent impairment of the immune response induced by trauma, surgery, and sepsis [19]. Further, these procedures and invasive lines disrupt the local protective immunity and allow pathogens to form biofilms into which antibiotic penetration is poor, leading to organisms' exposure to sub-therapeutic concentrations of antibiotics and selection of drugresistant strains [20].
Our study found that patients who were sicker at admission were more likely to get infections with DRO, including those with shock, acute kidney injury, and higher clinical severity scores (APACHE II and SOFA). This is consistent with other published studies and supports higher antibiotics in sicker patients [21,22]. The risk of drug-resistant infections was found to increase in our study with invasive procedures during hospital stays like mechanical ventilation, urinary catheterization, and central venous catheter insertion. This association has been observed in previously published studies [23,24]. This association reiterates the need to prevent VAP, CLABSI, and CAUTI by applying appropriate care bundles. Our study showed that only mechanical ventilation was independently associated with both MDR (RR = 1.21, 95% CI = 1.01 -1.45, p = -0.032) and P-XDR (OR = 1.25, 95% CI = 1.05 -1.41, p = -0.012) infections compared to N-MDR infections.
Infections with DRO are associated with increased mortality and morbidity in hospitalized patients in other studies [25]. We showed that mortality was significantly higher among patients with P-XDR infections (72%) as compared to MDR (42.4%) and N-MDR infections (37.5%). This may explain the shorter hospital stay duration (9 days in PXDR vs 10 days in NMDR) and the longer duration of ICU stay seen in patients with P-XDR infections. Logistic regression analysis showed that P-XDR infections increased mortality by 4.3 times (OR = 4.3, 95 % CI = 2.11 -8.7, p = -0.0001) compared to N-MDR infections. However, the individual impact of isolation of a drug-resistant organism on mortality is difficult to accurately determine as patients with P-XDR isolates were critically ill, and their poor outcome may be attributable to other coexistent factors.
Our study has several significant limitations. Firstly, the risk factors selected for the study may not account for other unknown variables that may influence the risk of a drug-resistant infection. Secondly, antibiotic sensitivity testing in our laboratory was done by the disc diffusion method, whereas the broth dilution method is recommended. Thirdly, our study did not consider the choice and dose of antibiotics when considering the role of previous antibiotic exposure in the acquisition of drug-resistant infections. Finally, as per institutional policy, only antibiotics that are used in routine clinical practice were tested for sensitivity in culture isolates. Therefore, sensitivity testing for antibiotics from every class could not be done, thus preventing culture isolates from being classified as XDR or PDR. However, P-XDR was intended to describe culture isolates with drug resistance beyond what is considered MDR and should be considered a marker of extensive resistance.

Conclusions
This study has shown AMR's threat in patients with infectious syndromes in a developing country and its impact on patient outcomes. We identified the risk factors that could independently predict the acquisition of P-XDR infections that may help in early stratification/identification of patients and subsequent early initiation of antibiotics. The conclusions drawn from this robust dataset may provide the basis for future interventions to reduce AMR's burden by effective ASP.