Acinetobacter baumannii is a gram-negative bacterium that can cause infections in both the community and hospital settings. Mortality is common in patients with Acinetobacter infections but may be due to the fact that individuals with these infections are often critically ill and/or immunocompromised rather than the virulence of the organism.Reference Karageorgopoulos and Falagas 1 Although the virulence is typically thought to be low, Acinetobacter can acquire antimicrobial-resistance genes quickly. This resistance has become an important public health concern owing to difficulty in treating these infections.Reference Bergogne-Berezin and Towner 2 , Reference Landman, Quale and Mayorga 3 Advanced age and having an invasive procedure that involves mechanical ventilation, catheters, shunts, or open wounds are risk factors for healthcare-associated infections (HAIs) due to Acinetobacter.Reference Fournier and Richet 4
HAIs in general are a major threat to patient safety and have been targeted for prevention by several recent pieces of federal legislation in the United States. 5 , Reference Brown, Doloresco Iii and Mylotte 6 Although 100% prevention may be unattainable, a recent study estimated that more than half of HAIs may be preventable.Reference Umscheid, Mitchell, Doshi, Agarwal, Williams and Brennan 7 Estimates of the burden of HAIs could help justify resource investments for prevention efforts and provide estimates of the value of healthcare resources that could be used for other purposes when HAIs are prevented. Recent studies have reported the economic and mortality burden associated with HAIs, including those caused by Clostridium difficile and methicillin-resistant Staphylococcus aureus (MRSA).Reference Zimlichman, Henderson and Tamir 8 – Reference Nanwa, Kendzerska and Krahn 10 However, little is known about the per-infection or cumulative burden of multidrug-resistant (MDR) Acinetobacter HAIs in the United States.
The aim of this analysis was to estimate the healthcare cost and mortality burden associated with MDR Acinetobacter HAIs using inputs from the published literature. We calculated this burden by first estimating the cost and mortality risk at the patient level and then using incidence estimates to generate estimates of the aggregate burden to the US population.
METHODS
Before our literature search, we identified 2 ways of calculating the healthcare cost associated with an MDR Acinetobacter HAI. The first is to compare the pre-discharge inpatient healthcare costs in patients with and without an MDR Acinetobacter HAI. A simplistic way of doing this would be to compare the average costs in the 2 groups with the hypothesis that costs are higher in patients with MDR Acinetobacter HAI. Of course, the costs may be higher in this group owing to the infection or other factors such as age, comorbidities, and other events that occurred in the hospital. Ideally, one would use a multivariable regression framework or matching methods in which these additional observable characteristics are controlled for or balanced across the 2 groups.
A health sciences librarian performed literature searches in PubMed, Cumulative Index to Nursing and Allied Health Literature, Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, National Health Services Economic Evaluation Database, Web of Science, and EMBASE. We reviewed the reference lists of retrieved articles to identify studies that were not identified from the preliminary literature searches. The searches were limited to January 2000 through May 2014. We searched MeSH terms and text words related to “Acinetobacter,” “Acinetobacter infections,” “mortality,” “survival,” “length of stay,” “cost,” “economics,” and “incidence.” Data were abstracted by 2 reviewers and included information on cost, incidence, length of stay (LOS), and mortality. The parameters for our model that were identified through this literature search are presented in Table 1.
Input Parameters in Study of Costs and Mortality Associated With MDR Acinetobacter HAIs
NOTE. HAI, healthcare-associated infection; HCUP, Healthcare Cost and Utilization Project; LOS, length of stay; MDR, multidrug-resistant; RR, relative risk; VA, Veterans Affairs.
a Indicates that the parameter is Acinetobacter-specific.
Our literature search yielded just one study that reported the direct medical cost associated with MDR Acinetobacter HAI, a 2004 study by Wilson and coauthors that focused on patients in the burn unit of a public teaching hospital.Reference Wilson, Knipe and Zieger 11 Converting to 2014 US dollars, the study found an unadjusted difference in mean cost between patients with an MDR Acinetobacter HAI and randomly selected control patients of $129,665.
The second method for calculating the healthcare cost associated with an MDR Acinetobacter HAI would be to estimate the impact of these infections on the excess LOS in the hospital and then multiply these additional days by an estimate of the average cost per inpatient day. As with the cost outcome, the excess LOS associated with an infection could be generated by a simple unadjusted difference or through more rigorous regression or matching methods. We used an estimate for the cost per inpatient-day from both the hospital ($2,030) 12 and the third party payer ($4,350) 13 perspective. Our search yielded 6 studies that generated estimates of the excess LOS attributable to MDR Acinetobacter HAI.Reference Wilson, Knipe and Zieger 11 , Reference Grupper, Sprecher, Mashiach and Finkelstein 14 – Reference Wong, Tan, Ling and Song 18 In each study, this estimate was calculated on the basis of the unadjusted difference in mean or median LOS. These 6 studies had an average effect size of 16.49 days with a range between 5.00 and 32.33.
Another complication with estimates of the excess cost and LOS associated with HAIs is time-dependent bias. In generating these estimates, researchers will often use the costs or LOS over the patients’ entire inpatient stay whereas it is only the cost or LOS that occurs after the HAI that could be attributed to the infection. We used 2 methods to reduce time-dependent bias. The first is by estimating a multistate model in which HAIs are treated as one of several mutually exclusive states through which a hospitalized patient can move. The second method entails matching patients with HAIs to patients without HAIs with a requirement that the non-HAI patients’ LOS be at least as long as the time until the HAI in the infected case patient. We recently estimated the magnitude of this time-dependent bias by comparing the estimates of LOS generated using conventional methods as well as either a multistate model or matching on the timing of infection in the same data.Reference Nelson, Nelson and Khader 19 We found that, on average, using a multistate model yielded an estimate of the excess LOS that was 29.6% of the estimate generated using conventional methods. This value was 52.6% for matching on the timing of infection. We used these 2 estimates to adjust both the excess cost and LOS estimates for MDR Acinetobacter HAIs that we identified in the published literature.
Our systematic literature review identified estimates of the relative risk of mortality attributable to MDR Acinetobacter HAI.Reference Ababneh, Harpe, Oinonen and Polk 20 – Reference Reddy, Chopra and Marchaim 22 We multiplied this estimate by the probability of inpatient mortality (0.02) in the general inpatient populationReference Hall, Levant and DeFrances 23 in order to generate estimates of the probability of death associated with resistant Acinetobacter HAI.
Our economic model generated estimates of the cost and mortality burden per MDR Acinetobacter HAI. In order to generate estimates of the annual aggregate cost and mortality burden associated with these infections in the United States, we multiplied the per-infection estimates by the number of infections, which was calculated using the incidence of MDR Acinetobacter HAI and the number of hospitalizations in the United States per year between 2005 and 2009. 24 We estimated the incidence of MDR Acinetobacter HAI using Veterans Affairs electronic medical record data. Positive cultures for Acinetobacter were considered healthcare-associated if they were identified more than 48 hours after admission and MDR was defined as resistant to 3 or more antibiotic drug class.
We ran our economic model using 1,000 first order Monte Carlo simulations and 10,000 second order Monte Carlo simulations for a total of 10,000,000 simulated individuals using TreeAge Pro 2013 (TreeAge Software). We used beta distributions for probability input parameters, gamma distributions for cost input parameters, and log-normal distributions for relative risks input parameters.
RESULTS
Table 2 contains the estimates of the cost and mortality burden per MDR Acinetobacter HAI. Using the total cost calculation method, our model estimates that the cost (95% CI) of an MDR Acinetobacter HAI would range from $129,917 ($94,899–$170,062) unadjusted for time-dependent bias to $38,423 ($24,273–$55,941) when adjusted for this bias using the adjustment from multistate models. When calculating cost based on LOS and an estimate of the cost per day in the hospital, these estimates ranged from $72,025 ($24,384–$149,035) to $21,294 ($6,875–$45,590) from the payer perspective and from $33,510 ($10,176–$72,748) to $9,906 ($2,882–$22,208) from the hospital perspective. In addition, our model estimated that the mortality rate attributable to MDR Acinetobacter HAI was 10.6% (2.5%–29.4%).
Estimates of the Cost and Mortality Burden per MDR Acinetobacter HAI
NOTE. HAI, healthcare-associated infection; LOS, length of stay; MDR, multidrug-resistant.
From our Veterans Affairs data, we estimated the incidence (95% CI) of MDR Acinetobacter HAI as 0.148 (0.136–0.161) per 1,000 patient-days. Our model estimated an annual cumulative incidence (95% CI) of 12,524 (11,509–13,625) MDR Acinetobacter HAIs in the United States (Table 3). These infections result in an estimated $1.627 billion ($1.495 billion–$1.770 billion) in attributable annual cost when calculating these costs on the basis of the total cost method. When adjusting these estimates for time-dependent bias, they ranged from $481 million ($442 million–$524 million) to $856 million ($787 million–$931 million). The attributable costs calculated using the excess LOS and an estimate of the excess hospital stay from the payer perspective ranged from $902 million ($829 million–$981 million) to $267 million ($245 million–$290 million). From the hospital perspective, these estimates ranged from $420 million ($386 million–$457 million) to $124 million ($114 million–$135 million). Finally, our model estimated 1,330 (1,222–1,447) annual deaths due to MDR Acinetobacter HAIs in the United States.
Annual Incidence and Aggregate Cost and Mortality Burden of MDR Acinetobacter HAI in the United States
NOTE. HAI, healthcare-associated infection; LOS, length of stay; MDR, multidrug-resistant.
DISCUSSION
To our knowledge, this is the first analysis to estimate the per-infection and aggregate economic and mortality burden associated with MDR Acinetobacter HAIs. Our results suggest that more than 10,000 of these infections may occur each year in the United States, resulting in more than 1,000 deaths and costs that could be as high as $1.6 billion. While it was beyond the scope of this study, one explanation for the high mortality rate that we identified may be a delay in appropriate therapy. Further studies are needed to investigate this area. We present a wide range of estimates based on explicitly stated assumptions. It is clear that more evidence is needed in order to generate more accurate estimates in the future. For example, we found only 1 study with an estimate of the attributable cost and only 3 with estimates of the attributable mortality of MDR Acinetobacter HAIs. However, all 4 of these studies were limited to just the time prior to discharge from the hospital. Several recent articles focusing on MRSA HAIs have identified a significant increase in both cost and mortality in the postdischarge period as well.Reference Nelson, Jones and Liu 25 , Reference Nelson, Stevens, Jones, Samore and Rubin 26 Future studies should estimate the effects of MDR Acinetobacter HAIs in patients both prior to and after discharge from the hospital. In addition, none of the 4 published studies with estimates of attributable cost or mortality adjusted for confounding in these estimations. Future studies should control for patient characteristics that influence both MDR Acinetobacter HAIs and cost or mortality outcomes in order to generate more accurate estimates of the true cost and mortality attributable to the HAI.
It is important to place these results in the context of other organisms that often cause HAIs. Zimlichman and colleaguesReference Zimlichman, Henderson and Tamir 8 conducted a systematic literature review and meta-analysis to identify the attributable cost, incidence, and cumulative incidence associated with healthcare-associated MRSA surgical site infections (SSIs), healthcare-associated MRSA central line–associated bloodstream infections, and healthcare-associated Clostridium difficile infections (HA-CDIs). Their estimates for the attributable cost for these infections were lower, per case, compared with most of our estimates ($42,300 for healthcare-associated MRSA SSIs, $58,614 for healthcare-associated MRSA SSIs, and $11,285 for HA-CDIs in 2012 US dollars). However, the aggregate costs were similar to our results for healthcare-associated MRSA SSIs ($990 million) and HA-CDIs ($1,508 million) owing to cumulative incidence estimates that were substantially higher (twice as high and 10 times as high for healthcare-associated MRSA SSIs and HA-CDIs, respectively). The US Centers for Disease Control and Prevention (CDC) reports an even larger estimate of the aggregate cost of HA-CDIs ($4.8 billion). 27
Although A. baumannii is most commonly linked to HAIs,Reference Manchanda, Sanchaita and Singh 28 there are a number of additional species within the Acinetobacter genus. The majority of the published studies from which we drew input parameters for our models focused solely on A. baumannii species.Reference Wilson, Knipe and Zieger 11 , Reference Playford, Craig and Iredell 15 – Reference Wong, Tan, Ling and Song 18 , Reference Ababneh, Harpe, Oinonen and Polk 20 – Reference Reddy, Chopra and Marchaim 22 The remaining studies indicated that they identified infections due to “Acinetobacter species” without specifying the type of species.Reference Grupper, Sprecher, Mashiach and Finkelstein 14 , Reference Sunenshine, Wright and Maragakis 16 In addition, because of the small number of relevant studies identified in our systematic review, the input parameters from our model were drawn from disparate and heterogeneous studies. Several studies focused solely on patients in an intensive care unitReference Playford, Craig and Iredell 15 , Reference Weingarten, Rybak, Jahns, Stevenson, Brown and Levine 17 whereas others included only patients in a burn unit.Reference Wilson, Knipe and Zieger 11 , Reference Wong, Tan, Ling and Song 18 The remaining studies included patients regardless of their location in the hospital.Reference Grupper, Sprecher, Mashiach and Finkelstein 14 , Reference Sunenshine, Wright and Maragakis 16 , Reference Ababneh, Harpe, Oinonen and Polk 20 – Reference Reddy, Chopra and Marchaim 22
Healthcare cost and mortality are intertwined. For example, healthcare costs could potentially be lower in patients who die because of a HAI compared with those who survive because they have fewer opportunities for healthcare services and resources to be applied. On the other hand, a number of studies have documented the increase in healthcare costs associated with end-of-life treatment.Reference Chastek, Harley, Kallich, Newcomer, Paoli and Teitelbaum 29 – Reference Calfo, Smith and Zezza 33 So, mortality may, in fact, lead to an overall increase in cost. Because of these relationships, it is important to report both cost and mortality in order to get a complete picture of the burden associated with a particular condition.
Estimates of the burden of disease can be important for a number of reasons. For example, these estimates are one way of comparing the relative impact of a wide range of conditions using common metrics.Reference Rice 34 In addition, decision makers can use burden of disease estimates as a way of allocating scarce healthcare resources toward treatment or prevention.Reference Clabaugh and Ward 35 – Reference Rice 37 However, in order for burden of disease to be useful in these ways, it is important that the estimate be as accurate as possible. Excessively high burden estimates, although attractive to disease advocates, may simply be not believable and therefore not useful for decision makers or may set unreasonably high expectations that will not be fulfilled when prevention efforts are undertaken.Reference Larg and Moss 38 For this reason, we chose to adjust our estimates for time-dependent bias.Reference Graves, Harbarth, Beyersmann, Barnett, Halton and Cooper 39
Although Acinetobacter HAIs have historically occurred in immune-compromised patients with many other serious health conditions, the bacteria itself has not been considered a high-virulence pathogen.Reference Paterson and Harris 40 However, a recent outbreak investigation identified a novel strain of Acinetobacter that caused mortality among relatively healthy patients, indicating that the current understanding of this organism may need to be reevaluated.Reference Paterson and Harris 40 , Reference Jones, Clancy and Honnold 41 As future studies evaluate the evolving virulence of Acinetobacter, estimates of the mortality risk and resources required to treat individuals with these infections will need to be updated.
In a 2013 report, the US CDC classified MDR Acinetobacter as a serious threat requiring prompt and sustained action to ensure that the problem does not grow. In this report, the estimated annual number of MDR Acinetobacter HAIs was 7,300 and the number of infection-related deaths was 500, both slightly lower than our estimates. 42 The CDC estimates were calculated by multiplying the proportion of Acinetobacter HAIs amongst a survey of 452 patients hospitalized in 10 different states by the estimated number of total HAIs in the United States during 2011 based on the Nationwide Inpatient Sample data from the Agency for Healthcare Research and Quality.
Our study was subject to a number of limitations. The most important limitation is that our economic model was parameterized using inputs from the published literature and supplemented with estimates generated from national Veterans Affairs inpatient data. To the extent that these parameter estimates are biased or otherwise inaccurately measured, the burden estimates produced by our model will be similarly flawed. We were able to find only one paper that estimated the impact of MDR Acinetobacter HAIs on attributable cost.Reference Wilson, Knipe and Zieger 11 Besides not adjusting for confounders in the analysis, the data from this study, which were from patient encounters that occurred from July 2000 to August 2001 in the burn unit of a hospital, are more than a decade old at this point and may not be generalizable to less severely ill patients. It is possible that using other MeSH terms in our search strategy would have yielded other studies. In addition, the estimates of excess cost and LOS that we identified from the literature were all subject to time-dependent bias, which has been shown to inflate the effects of HAIs on outcomes, as well as confounding bias. We attempted to overcome time-dependent bias using adjustment factors from a recently published paper but confounding bias may still exist. Although the studies used to estimate these adjustment factors focused on HAIs, none were specific to Acinetobacter. Finally, we did not find any published studies that estimated the effect of MDR Acinetobacter HAIs on postdischarge mortality or healthcare costs. Therefore, our results may underestimate the true burden of these infections.
In conclusion, we find that MDR Acinetobacter HAIs are associated with substantial costs and mortality. Efforts to prevent these infections could save lives and preserve healthcare resources that could be used for other purposes.
ACKNOWLEDGMENTS
Financial support. CDC; Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Health Services Research and Development Service (grant CDA 11-210 to R.E.N.). This material is the result, in part, of work supported with resources and the use of facilities of the George E. Wahlen Department of Veterans Affairs Medical Center, Salt Lake City, Utah.
Potential conflicts of interest. All authors report no conflicts of interest relevant to this article.
Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government.
