Excess mortality and long-term disability from healthcare-associated carbapenem-resistant Acinetobacter baumannii infections: A nationwide population-based matched cohort study

Background Carbapenem resistance is perceived as a clinical challenge in the management of debilitated and immunocompromised patients who eventually will die from underlying diseases. We aimed to examine whether carbapenem resistance per se, rather than the underlying diseases, negatively affect outcomes, by comparing the excess mortality and morbidity from healthcare-associated infections (HAIs) caused by carbapenem-resistant Acinetobacter baumannii (CRAB) and carbapenem-susceptible A. baumannii (CSAB). Methods This was a nationwide retrospective matched cohort study of hospitalized patients in 96 hospitals which participated in Taiwan Nosocomial Infection Surveillance (TNIS). A total of 2,213 patients with A. baumannii HAIs were individually matched to 4,426 patients without HAIs. Main outcomes were excess risks for one-year all-cause mortality and one-year new-onset chronic ventilator dependence or dialysis-dependent end-stage renal disease. Results Excess one-year mortality was 27.2% in CRAB patients, compared with their matched uninfected inpatients, as well as 15.4% in CSAB patients (also compared with their matched uninfected inpatients), resulting in an attributable mortality of 11.8% (P <0.001) associated with carbapenem resistance. The excess risk associated with carbapenem resistance for new-onset chronic ventilator dependence was 5.2% (P <0.001). Carbapenem resistance was also associated with an extra cost of $2,511 per case of A. baumannii HAIs (P <0.001). Conclusion Carbapenem resistance is associated with a significant disease burden in terms of excess mortality, long-term ventilator dependence, and medical cost. Further studies on effects of antimicrobial stewardship programs in decreasing this burden are warranted.


Introduction
Healthcare-associated infections (HAIs), caused by the Acinetobacter baumannii, have emerged as a major global health problem posing significant threat to vulnerable hospitalized patients [1,2].This gram-negative bacterium is not only ubiquitous in nature but also capable of surviving for prolonged periods in healthcare environment [3,4].A. baumannii can causes pneumonia, bloodstream infections, urinary tract infection, and surgical site infections, with wide variations in the excess mortality rate estimates, from 7.8% to 34.0% in general patients [5,6] and 2.6% to 43.0% in critically ill patients [7,8], compared with patients without HAIs.Carbapenem antibiotics (e.g.imipenem and meropenem) were traditionally the most effective antimicrobials for treating A. baumannii infections [9].Since 2000s, carbapenem resistance became a clinical therapeutic challenge [10].Compared with carbapenem-susceptible A. baumannii (CSAB) HAIs, carbapenem-resistant A. baumannii (CRAB) HAIs were more likely to occur in debilitated or immunocompromised patients who eventually will die from underlying diseases [11].Whether carbapenem resistance per se negatively affect the outcomes of A. baumannii HAIs remained uncertain.The question is important for assessing potential impact of antimicrobial stewardship programs that aim to reduce carbapenem resistance.
Existing literature includes five small matched cohort studies, which reported an increased short-term mortality by 1.3-6.9folds in patients with CRAB HAIs compared to their matched patients with CSAB HAIs [10,[12][13][14][15].Of the five studies, three studies focused on bloodstream A. baumannii infections (40-63matched-pairs) [10,12,15] and the other two studies examined all-type A. baumannii HAIs (42-91 matched-pairs) [13,14].None of these studies looked at long-term mortality or survivors' functional status, such as chronic ventilator and dialysis dependence.Furthermore, the small sample size did not allow the researchers to control important confounding factors, particularly baseline patient characteristics.
A nationwide surveillance system, Taiwan Nosocomial Infection Surveillance (TNIS) collects HAI data throughout Taiwan.The TNIS data indicated that A. baumannii was responsible for 5.6% of all-type HAIs and 7.1% in bloodstream infections in Taiwan [16].By 2021, up to 75.6% of those A. baumannii isolates in the intensive care units of medical centers were resistant to carbapenems [16].This study aimed to investigate whether carbapenem resistance per se negatively affect the long-term outcomes of patients, including excess in mortality, newonset chronic ventilator dependence, and new-onset dialysis-dependent end-stage renal disease, using the national health databases of the Taiwan Ministry of Health and Welfare.

Study design
This was a nationwide retrospective matched cohort study comparing the excess mortality and new-onset irreversible long-term morbidities (including ventilator dependence and dialysisdependent end-stage renal disease) from HAIs caused by CRAB and that from HAIs caused by CSAB.

Source of data
This work is part of the Taiwan Nosocomial Infection Surveillance (TNIS) Study [17].To mitigate the potential confounding effects of the emergence of extensively drug-resistant A. baumannii (XDRAB) in Taiwan after 2008 [18] (TNIS data cannot differentiate between XARAB isolates and non-XDR CRAB isolates as tigecycline susceptibility was not routinely performed in most hospitals), this study exclusively included isolates collected from 2006 to 2008.The methodological details of TNIS Study, which applied matched cohort design to quantify excess mortality and long-term morbidity attributable to healthcare-associated infections, had been published [17].S1 Appendix described the methods in cross-linking national health databases when applied to the present study in detail.

Ethical statement
The study protocol was reviewed by the Research Ethic Committee of National Taiwan University Hospital (Taipei, Taiwan) and certified for exempt review in accordance with law and regulations (#201609005W), which does not require informed consent as all data were fully anonymized before we accessed them.

Settings and HAI surveillance
The TNIS was established by Taiwan Centers for Diseases Control (Taipei, Taiwan).The participating hospitals used the United States Centers for Disease Control and Prevention (Atlanta, GA, USA) HAIs surveillance definitions [19].

Statistical analysis
The excess risk of healthcare-associated A. baumannii HAIs was defined as the calculated the difference in mortality and new-onset organ failure risk between the A. baumannii HAI group and the uninfected group.The impact of carbapenem resistance was estimated by subtracting the excess mortality and morbidity risk, length of stay, and hospital cost of CRAB group from that of CSAB.All statistical analyses were performed using SAS, ver 9.2 (SAS Institute Inc., Cary, NC, USA).All P values were two-sided.Statistical significance was set at P < 0.05.The level of statistical significance for P values under multiple comparisons was set using Bonferroni's correction.

Results
Of the 32,026 HAI patients in 96 TNIS participating hospitals during the TNIS Study from 2006 to 2008, 7.8% (2,503/32,026) of isolated pathogen was A. baumannii, of which 95.7% (2,396/2,503) met the study inclusion criteria.Of them, 92.4% (2,213/2,396) were successfully matched to patients without HAIs (Fig 1).The A. baumannii HAI patients with unsuccessful matching (n = 183) had a significantly higher proportion of severe illness at baseline (on the date of admission) (13.7% vs. 2.8% for dialysis-dependent end-stage renal disease, P <0.001) and longer average length of stay before onset of the A. baumannii HAIs (31.7 days vs.17.1 days, P <0.001), than A. baumannii HAI patients with successful matching.The subsequent data analysis excluded the 183 patients who were unsuccessfully matched.
Comparison of baseline characteristics between 2,213 patients with A. baumannii HAIs and 4,426 patients without HAIs are shown in Table 1.The matched uninfected patients had a significantly higher proportion of ischemic heart disease (7.2% vs. 4.5%, P <0.001), diabetes mellitus (21.5% vs. 18.0%,P = 0.001), hypertension (23.7% vs. 13.6%,P <0.001), and lower average number of diagnoses (4.3 vs. 4.7, P <0.001), than A. baumannii HAI patients.Except the only 3 variables with significant differences, the remaining matching variables and validation variables had no significant differences between the two groups.
Patients with A. baumannii HAIs had an attributable in-hospital mortality, mortality within 30 days after discharge, and one-year mortality of 21.6%, 23.2%, and of 20.9%, respectively, compared with the matched uninfected patients.The excess risk of new-onset chronic ventilator dependence during hospitalization, within 30 days after discharge, and within one-year was 8.6%, 10.6%, and 10.2%, respectively.The excess mortality and excess risk of new-onset chronic ventilator dependence was highly statistically significant after Bonferroni correction for multiple comparisons and adjusting for the presence of ischemic heart disease, diabetes mellitus, and hypertension (all Ps <0.001) (Table 2).The excess one-year mortality by site of infection was highest for nosocomial pneumonia (28.7%), followed by surgical site infections (21.3%), and bloodstream infection (17.9%).The excess one-year mortality by the type of antimicrobial resistance was 27.2% for CRAB and 15.4% for CSAB (all Ps <0.001) (Table 3 and Fig 2).The excess hospital stay was 9.9 days and extra hospital cost was $6,096.The excess hospital  stay and extra hospital cost were also significant in the subgroup analysis by site of infection, the type of antimicrobial resistance, and the presence of severe illnesses at admission (all Ps <0.001) (Table 4).
Of the 2,213 A. baumannii HAI cases, the causal A. baumannii strains were CRAB in 1,036 cases (46.8%).Patients with CRAB HAIs (n = 1,036) had a longer length of stay in hospital before the onset of the A. baumannii HAI (18.8 days vs. 15.6 days), older age (70.2 years vs. 68.0years), and higher proportion of occurring in intensive care units (54.7% vs. 34.2%),compared with patients with CSAB HAIs (n = 1,177) (all Ps <0.001).These data implied the important differences in severity of underlying diseases between patients with CRAB HAIs and CSAB HAIs.
Table 5 summarizes the main outcomes of carbapenem resistance in A. baumannii HAIs.The excess one-year mortality was 27.2% among CRAB patients compared with their matched uninfected inpatients, and 15.4% among CSAB patients compared with their matched uninfected inpatients, resulting in an attributable mortality of 11.8% (all Ps <0.001).The excess one-year mortality of cabapenem resistance was highest for bloodstream (27.6%), followed by urinary tract infections (12.4%), and pneumonia (4.4%).Carbapenem resistance had an excess risk for new-onset chronic ventilator dependence of 5.2% (Ps <0.001) and had an extra hospital cost of $2,511 (P <0.001) (Table 6).

Discussion
Healthcare-associated A. baummannii infection is a significant threat to vulnerable hospitalized patients with severe outcome and deaths.To the best of our knowledge, this study is the first nationwide population-based large cohort study to investigate the negative impact of carbapenem resistance on the outcomes of patients with A. baumannii HAIs.Our study included 1,177 CSAB and 1,036 CRAB patients to demonstrate the burden of carbapenem resistance.We showed that carbapenem resistance in patients with A. baumannii HAIs increased the risks for long-term mortality of 11.8%, disability (new-onset chronic ventilator dependence) of 5.2% and extra hospital cost of $2,511.Apart from nationwide cohort study, our study had explored the clinical long-term outcomes, adjusted the confounding factors by using individual matching to enhance comparability between CRAB and CSAB HAI patients.The testing results of the matching variables and the comparability-validation variables did show no significance between-group difference in most baseline characteristics.
Our study is consistent with the previous studies on the excess mortality associated with carbapenem resistance in A.baumannii HAI patients [10,[12][13][14][15]. Existing literature revealed that an excess mortality ranged from 1.6-30.0%for carbapenem resistance in A. baumannii HAI patients.The wide range of excess mortality is likely due to the smaller sample size and different types of infections [10,[12][13][14][15]. Furthermore, our findings validated the existing results and found an excess one-year mortality of 11.8% for carbapenem resistance in all-type A.baumannii HAI patients, in which bloodstream infections and urinary tract infections suffered the largest excess one-year mortality of 27.6% and 12.4% for carbapenem resistance, respectively.A bloodstream infection represents a systemic dissemination of the infection rather than a localized infection such as pneumonia, urinary tract infection, and surgical site infection.The limited availability of effective treatment options for carbapenem resistance in A.baumannii bloodstream infections can increase the risk of treatment failure and result in severe sepsis or septic shock, both of which were associated with a higher excess mortality [1,20].
Apart from excess long-term mortality, our study also revealed that carbapenem resistance in A. baumannii HAI patients had an excess risk of long-term morbidity of new-onset chronic ventilator dependence. A. baumannii infections can cause severe outcome of acute dysfunction [21][22][23].The pathogen frequently causes respiratory infections in mechanically ventilated patients and thus worsen the respiratory function after A. baumannii HAIs [24,25].In the present study, the most common A. baumannii HAIs was pneumonia, represented 38% of all-type A. baumannii HAIs.Our findings provided the evidence on irreversibility of CRAB HAI-associated ventilator dependence and showed the carbapenem resistance increased the risks for new-onset chronic ventilator dependence by 5.2% after A. baumannii HAIs.
The attributable risks of long-term mortality and morbidity emphasized the importance of hospital infection control for CRAB.The major risk factors of CRAB acquisition included prior exposure to antibiotics (especially carbapenems), longer hospital stay, invasive procedures, and admission to a ward with CRAB colonization pressure were documented as the risk factors [26][27][28][29].A nationwide study demonstrated a strong positive association between hospital carbapenem consumption and CRAB prevalence, thus suggested that prudent use of carbapenems antibiotics is essential to control the spread of CRAB [30].Preventing and controlling of multidrug-resistant organisms (MDROs) are not only a national priority but also are assumed responsibility for all hospitals [1,31].Otherwise, the burden of antimicrobial resistance will result in increased morbidity, mortality, and costs of health care.Our results indicated that controlling the spread of CRAB within hospitals may translate to a decrease in incidence of irreversible severe outcomes and substantial savings in the National Health Insurance medical expenditure, particularly when the cost of long-term respiratory care for those who depend on ventilator is considered.This study presents limitations.First, our data were retrieved from the TNIS Study data which did not cover every hospital in Taiwan.However, we minimized the potential bias on estimating excess risk by individually matching the A. baumannii HAI patients to uninfected patients in the same hospital.Second, it is possible that part of the 4,426 matched uninfected patients might indeed have HAIs due to voluntary nature of TNIS participation, which would have caused the underestimation of excess risks for long-term mortality and morbidity.Third, the diversity in antibiotic treatment precluded meaningful comparison between different treatments in outcomes of CRAB.Fourth, this study exclusively included isolates collected from 2006 to 2008 to minimize potential bias resulting from treatment given for XDRAB infections after 2008.As a result, antibiotic treatment and patients' outcomes would not confound by XDRAB infections.Fifth, underlying disease might play an important role in the outcomes of these patients.There remains a possibility of residual confounding in our study despite conducting a matched cohort study design and adjusting for presence of underlying diseases.In conclusion, carbapenem resistance in patients with A. baumannii HAIs has significant negative effect on the long-term clinical outcomes on vulnerable hospitalized patients.Future study on the performance of antimicrobial stewardship program on the patients' outcome should be take into consideration the long-term negative outcomes.

Table 2 . Excess risks for mortality and new-onset organ failure in patients with A. baumannii HAIs.
a Follow-up duration from index date to endpoint of observation.b Number at risk: the number of patients who have not yet developed the outcomes at admission.c Adjusted for ischemic heart disease, diabetes mellitus and hypertension.d Statistically significant using Bonferroni's correction (P<0.05/18= 0.0028).Abbreviations: HAI, healthcare-associated infection; OR, odds ratio.https://doi.org/10.1371/journal.pone.0291059.t002