Clinical and Economic Burden of Carbapenem-Resistant Infection or Colonization Caused by Klebsiella pneumoniae, Pseudomonas aeruginosa, Acinetobacter baumannii: A Multicenter Study in China

Background: Carbapenem resistant Klebsiella pneumoniae (CRKP), Pseudomonas aeruginosa (CRPA), and Acinetobacter baumannii (CRAB) pose significant threats to public health. However, the clinical and economic impacts of CRKP, CRPA, and CRAB remain largely uninvestigated in China. This study aimed to examine the clinical and economic burden of CRKP, CRPA, and CRAB compared with carbapenem susceptible cases in China. Method: We conducted a retrospective and multicenter study among inpatients hospitalized at four tertiary hospitals between 2013 and 2015 who had K. pneumoniae, P. aeruginosa, and A. baumannii positive clinical samples. Propensity score matching (PSM) was used to balance the impact of potential confounding variables, including age, sex, insurance, number of diagnosis, comorbidities (disease diagnosis, and Charlson comorbidity index), admission to intensive care unit, and surgeries. The main indicators included economic costs, length of stay (LOS), and mortality rate. Results: We included 12,022 inpatients infected or colonized with K. pneumoniae, P. aeruginosa, and A. baumannii between 2013 and 2015, including 831 with CRKP and 4328 with carbapenem susceptible K. pneumoniae (CSKP), 1244 with CRPA and 2674 with carbapenem susceptible P. aeruginosa (CSPA), 1665 with CRAB and 1280 with carbapenem susceptible A. baumannii (CSAB). After PSM, 822 pairs, 1155 pairs, and 682 pairs, respectively were generated. Compared with carbapenem-susceptible cases, those with CRKP, CRPA, and CRAB were associated with statistically significantly increased total hospital cost ($14,252, p < 0.0001; $4605, p < 0.0001; $7277, p < 0.0001) and excess LOS (13.2 days, p < 0.0001; 5.4 days, p = 0.0003; 15.8 days, p = 0.0004). In addition, there were statistically significantly differences in hospital mortality rate between CRKP and CSKP, and CRAB and CSAB group (2.94%, p = 0.024; 4.03%, p = 0.03); however, the difference between CRPA and CSPA group was marginal significant (2.03%, p = 0.052). Conclusion: It highlights the clinical and economic impact of CRKP, CRPA, and CRAB to justify more resources for implementing antibiotic stewardship practices to improve clinical outcomes and to reduce economic costs.


Introduction
Carbapenem resistance poses a significant threat to public health globally. It occurs mainly among gram-negative bacteria such as Klebsiella pneumoniae, Pseudomonas aeruginosa, and Acinetobacter baumannii, which are the most concerning bacteria with a propensity toward multi-drug resistance and cause the life-threatening healthcare associated infections amongst critically ill and immunocompromised individuals [1]. Carbapenem resistant K. pneumoniae (CRKP), P. aeruginosa (CRPA), and A. baumannii (CRAB) were recognized as critical-priority bacteria, by the World Health Organization (WHO), which was mainly based on the associated increasing clinical and economic burden, unavailability of effective therapy, and antibiotic resistant characteristics [2].
Recently published studies mainly focus on gram-positive bacteria [9], while only few studies estimated the burden of CRKP, CRPA, and CRAB in terms of mortality, length of stay (LOS), and cost [10][11][12][13][14]. It was reported that patients with CRKP, CRPA, and CRAB infections were associated with higher mortality, longer hospital stay, and higher hospital costs compared with carbapenem susceptible cases [9,[15][16][17]. In addition, hospital costs incurred by colonized patients with CRKP were also high [18]. However, it was also found that there were no significant differences in hospital mortality between CRPA and carbapenem susceptible P. aeruginosa (CSPA) cases [10,15], in total hospital cost between CRAB and carbapenem susceptible A. baumannii (CSAB) cases [14].
In China, there were some studies exploring the mortality and LOS associated with CRKP [19][20][21][22][23], CRPA [24][25][26], or CRAB [27][28][29][30], and exploring the economic burden attributed by CRKP [19], CRPA [24], or CRAB [27,31]. Although hospitalized patients with carbapenem resistance appeared to have increased hospitality mortality, hospital stay, and hospital costs, no significant differences in hospital mortality and hospital stay between CRKP and carbapenem susceptible K. pneumoniae (CSKP) cases were showed as well [19,21]. In addition, majority of studies were conducted in a single hospital setting, and using univariate analyses regardless of confounding risk factors [32]. Therefore, there are limited published data and the clinical and economic impacts of CRKP, CRPA, and CRAB remain largely uninvestigated in China. In this study, we aimed to examine the clinical and economic burden of carbapenem resistance in K. pneumoniae, P. aeruginosa, and A. baumannii compared with carbapenem susceptible cases in China.

Study Site
This study was conducted in four tertiary hospitals in China. Hospital site 1 and site 2 are general provincial hospitals in Zhejiang province and Shandong province, respectively. Hospital site 3 and site 4 are general county hospital and combined traditional Chinese and Western medicine provincial hospital in Zhejiang province, respectively. Hospital site 1-4 are with 3200, 3500, 1727, 2100 of hospital beds, and with 170,000, 160,000, 80,000, 50,000 of inpatients per year, respectively.

Study Design and Patients
It was a retrospective and multicenter study. Patients were identified from the records of the clinical microbiology laboratory. 100% of inpatients in hospital site 2-4, and 60% of inpatients in hospital site 1 between 2013-2015 who had clinical samples positive for CRKP, CRPA, CRAB infection or colonization were included. The control group comprised of patients with CSKP, carbapenem susceptible P. aeruginosa (CSPA), and CSAB-positive clinical samples with the same percentages who were hospitalized during the same study period. Only 60% of inpatients were randomly selected from hospital site 1 due to the large inpatient population [31]. The interpretation of carbapenem susceptibility was based on the Clinical and Laboratory Standards Institute (CLSI) definitions and reported as resistant (R), susceptible (S), and intermediate (I). K. pneumoniae, P. aeruginosa, and A. baumannii isolates, were considered resistant to carbapenems if they were resistant or intermediate to any carbapenem (imipenem, meropenem, panipenem or ertapenem) [31], and the control group was defined as patients infected or colonized by K. pneumoniae, P. aeruginosa, and A. baumannii susceptible to all carbapenems. To avoid duplication, cases with the first episode detected in any clinical specimen (e.g., blood, stool, cervical, urethral sources) were included. If multiple samples were taken from a patient during the study period, only the first sample was included in this study.

Data Collection
Data were obtained from patients' medical records. Patients' characteristics included demographics (age, sex, and insurance), comorbidities (disease diagnosis, and Charlson comorbidity index (CCI), hospital events (admitting service, surgical services, and dates of hospital and intensive care unit (ICU) admission/discharge), and clinical outcomes (discharged alive or death during hospitalization). We also noted any related laboratory results when isolation of K. pneumoniae, P. aeruginosa, and A. baumannii was recorded in the inspection system, and recorded the antibiotic susceptibility test results obtained from the microbiology laboratory. In addition, economic costs associated with these patients were obtained from the financial system. The economic costs included total hospital cost, medication cost (antibiotic cost), diagnostic cost, treatment cost, material cost, and other costs, including out-of pocket payment and payment covered by health insurers. Values of the economic costs were converted from Chinese Yuan to United States dollars in 2015 [33,34].

Propensity Score Matching
To eliminate the impact of potential confounding variables, propensity score matching (PSM) with 1:1 nearest-neighbor matching using STATA was conducted. For logistic regression modeling with carbapenem resistant and carbapenem susceptible as dependent variables, independent variables included age, sex, insurance, number of diagnosis, CCI, admission to ICU, surgery, and comorbidities, but not clinical and economic outcomes (economic costs, LOS, and in-hospital mortality). The generated pairs, who were matched for all included variables, were subjected to further analyses of economic costs, LOS, and in-hospital mortality.

Indicators and Statistical Analyses
In this study, the main indicators included economic costs, LOS, and in-hospital mortality. We performed 1000 iterations of Monte Carlo simulations to calculate the 95% uncertainty interval for each indicator with normal distribution using SAS [35]. We compared the main indicators between CRKP and CSKP group, between CRPA and CSPA group, between CRAB and CSAB group using t test and χ 2 test for quantitative and qualitative variables, respectively. All p-values were two-tailed, <0.05 were deemed statistically significant, and ≥0.05 and <0.10 were considered marginally significant.

Ethical Approval and Informed Consent
The study was approved by the institutional review board of Zhejiang University School of Public Health, who waived the need for informed consent. All inpatients data were anonymized prior to analysis.

Results
We included 12,022 inpatients infected or colonized with K. pneumoniae, P. aeruginosa, and A. baumannii between 2013 and 2015, including 5159 K. pneumoniae, 3918 P. aeruginosa, and 2945 A. baumannii. Of these, a total of 831 with CRKP and 4328 with CSKP, 1244 with CRPA and 2674 with CSPA, 1665 with CRAB and 1280 with CSAB were included. Significant differences were found in insurance, admission to ICU, surgery between CRKP and CSKP group, in age, sex, number of diagnosis, CCI, admission to ICU between CRPA and CSPA group, in age, sex, insurance, number of diagnosis, CCI, admission to ICU, surgery between CRAB and CSAB group. Some comorbidities between the three groups were significantly different as well. After PSM, there were no differences in patients' characteristics between the three groups, and generated 822 pairs, 1155 pairs, and 682 pairs, respectively (Tables 1-3).
Inpatients with carbapenem resistance were significantly associated with higher economic costs than carbapenem susceptible cases. For patients with K. pneumoniae, the mean differences (95% UI) in total hospital cost, antibiotic cost, medication cost including antibiotics, diagnostic cost, treatment cost and material cost were  (Table 6).      Compared with inpatients with CSKP, CSPA, and CSAB, those with CRKP, CRPA, and CRAB were significantly associated with longer LOS, with mean differences (95% UI) of 13.2 days (12.7-13.7days), 5.4 days (4.4-6.5 days), and 15.8 days (13.9-17.7 days), respectively (Table 7).

Discussion
To the best of our knowledge, this is the first large and multicenter study quantifying the clinical and economic impact of CRKP, CRPA, and CRAB in mainland China using the PSM method. We found that after PSM, compared with carbapenem susceptible cases, those with CRKP, CRPA, and CRAB were associated with significantly increased economic costs, excess LOS, and attributable in-hospital mortality rate. A marginal difference in hospital mortality rate existed between CRPA and CSPA group.
It was clearly demonstrated that economic costs for infection or colonization caused by carbapenem resistance were higher than in case of carbapenem susceptible bacteria, suggesting that carbapenem resistance indeed incurs excessive economic costs on patients infected or colonized with K. pneumoniae, P. aeruginosa, and A. baumannii. It also shows that the impact of carbapenem resistance on economic costs might depend on the type of gram-negative bacteria, with resistance in K. pneumoniae having a larger impact, followed by A. baumannii and P. aeruginosa. This finding is consistent with several previous studies, in which carbapenem resistance was associated with higher economic costs for infection or colonization caused by gram-negative bacteria, including K. pneumonia [19], P. aeruginosa [10,12,24], and A. baumannii [11,13,27,31]. However, one study did not find a statistically significant association in total hospital cost among infants with ventilator associated pneumonia in the ICU between CRAB and CSAB group [14], which might be because that critical illness can attenuate the effect of carbapenem resistance [36]. Effective control of carbapenem resistance would result in cost savings and could be useful for assessing the cost-effectiveness of interventions to reduce the development and spread of carbapenem resistance in hospital settings [19].
We indeed found that carbapenem resistance was associated with longer LOS, which is similar with other investigations that patients with resistant bacteria requires increases in the number of hospitalization days [19,24,37]. Prolonged LOS associated with carbapenem resistance might be independent of the type of gram-negative bacteria, with resistance in A. baumannii having a larger influence, followed by K. pneumoniae, and P. aeruginosa. Hospital stay is the major contributor to the additional economic costs in carbapenem resistant infection or colonization, as patients with longer hospital stay were associated with more antibiotic therapy and more surgeries [19,24,37]; therefore, LOS was not included in PSM analyses [38].
The poorest clinical outcomes were observed for CRAB, followed by CRKP and CRPA. In-hospital mortality rate was significantly higher for patients with CRKP and CRAB than for carbapenem susceptible cases, which is similar with other studies [16,39,40]. A 2.03% increase in mortality rate for the CRPA patients is clinically meaningful, and the non-significant p-value (p = 0.052) is too low to confidently rule out an effect of CRPA on mortality rate, which is different with other findings [16,39,40], however, several studies reported non-significant association between mortality and CRPA as well [10,15]. Spending on medical services was associated with a reduction in the mortality rate, therefore, it is critical to consider the clinical and economic burden posed by CRKP, CRPA, and CRAB when shirting resource allocation [41].
The proportions of CRKP, CRPA, and CRAB in the four sampled hospitals were 10.29%, 31.75%, and 56.54%, respectively, which were approximate the national levels reported in China Antimicrobial Resistance Surveillance System (CARSS) (8.03%, 22.61%, and 58.05%) [42]. In addition, the mean total hospital cost, length of hospital stay, and hospital mortality were $3042 and $2470, 10.1 days and 9.4 days, 0.3% and 0.4% in Zhejiang and Shandong province in 2015, respectively, which were similar to the national levels ($2378, 9.6 days, and 0.4%). Therefore, we assumed that the antibiotic resistant level, and clinical and economic burden due to CRKP, CRPA, and CRAB from four sampled hospitals in China were approximate results representing the national level.
Our study has some limitations. First, due to the retrospective nature of our study, it is difficult to distinguish infection or colonization, which may underestimate the clinical and economic outcome of carbapenem resistant infection. However, colonization is an important reservoir for bacteria causing infection, therefore, it is important to identify the burden of patients with CRKP, CRPA and CRAB, either infected or colonized, in order to contain the spread of these bacteria. In this study, we explored the clinical and economic outcomes for CRKP, CRPA and CRAB isolation and not specifically infection. Studies on clinical infection or colonization separate should be considered in the future as well. Secondly, this is a retrospective study, inherently resulting in bias and confounding, thus, PSM was conducted to balance potential confounding factors in order to minimize the risk of bias. However, PSM is not without its own limitations, which ignore unmeasured confounders that could potentially impact outcomes. Moreover, although it is a multicenter study that was conducted in four tertiary hospitals, the antibiotic resistant levels and the main indicators approximate the national levels, it is necessary to expand this study to different types of hospitals in different areas in the future. Finally, as we had data between 2013 and 2015 only, Monte Carlo simulations with 1000 iterations were conducted and mean values during the study period were reported. Although study period did not influence the conclusions, future studies with update data are warranted.

Conclusions
This study underscores the substantial clinical and economic burden associated with CRKP, CRPA, and CRAB in hospitalized patients. The input of more resources to control carbapenem resistance in K. pneumoniae, P. aeruginosa, and A. baumannii can be justified both for improving clinical outcomes, and for reducing economic costs, thus maximized benefit with available resources. In addition, urgent need for implementing antibiotic stewardship practices across the continuum of hospital settings will hopefully help to curb the emergency and spread of carbapenem resistance and K. pneumoniae, P. aeruginosa, and A. baumannii.