Catheter-related bloodstream infections (CR-BSIs) increase health costs and patient morbidity 1, and their prevention has been the target of national initiatives to create safer and more efficient healthcare systems 23. These healthcare-acquired infections are among the group for which the US Centers for Medicare and Medicaid Services are now able to withhold payments 4, thereby shifting the cost onto the hospitals rather than healthcare payers who reimburse the clinical facilities. Given this change in the economic context for infection control, decision makers are likely to pay more attention to the cost effectiveness of interventions they employ to reduce rates of CR-BSI 5.

The use of specific types of antimicrobial-coated central venous catheter (A-CVC) to prevent CR-BSI has been shown in earlier economic evaluations to be cost-saving and generate health benefits within the wider healthcare system 67. However, not all have been evaluated and there are concerns over the quality of these evaluations and the usefulness of their findings for real-world decision making 8.

Problems with the existing economic evidence contribute to the ongoing uncertainty about the use of A-CVCs. First, the relative cost effectiveness of the different types of A-CVC is unknown as none of the previous evaluations compared all available types. Second, recent epidemiological evidence 1 suggests earlier evaluations may have overestimated the attributable mortality and length of stay associated with CR-BSI, and these were key drivers of the results 8. Third, the excess length of stay due to infection is a major source of cost savings and the dollar value given to each bed day released will depend on the preferences of the decision maker. They cannot be directly observed and require careful elicitation, and the valuation may change depending on who is making the decision. To date there has been no discussion as to how these value judgments are derived, creating another subtle source of uncertainty in the results of the earlier evaluations.

There is continued uncertainty among clinicians over which, if any, A-CVC to use. Clinical guidelines recommend their use only in specific circumstances 9, and evidence suggests that the uptake of these technologies remains patchy 1011. The purpose of this study is to evaluate the cost effectiveness of adopting A-CVCs to prevent CR-BSI in Australian intensive care units (ICUs). We considered all available catheter types, used updated estimates of the consequences of infection, and explored how uncertainty can impact the adoption decision. By doing so, we provide a deeper analysis of this infection control decision that will support those working in this clinical area.

We undertook an economic evaluation to identify the cost effectiveness of triple-lumen A-CVCs for standard use in Australian adult ICUs. We considered all commercially manufactured A-CVCs sold in Australia: minocycline and rifampicin (MR)-coated catheters; silver, platinum and carbon (SPC)-impregnated catheters; and two chlorhexidine and silver sulfadiazine-coated catheters; one coated on the external surface (CH/SSD (ext)) and the other coated on both catheter surfaces (CH/SSD (int/ext)). The baseline comparator was uncoated polyurethane catheters.

The results of the first analysis, without uncertainty, showed all four types of A-CVC were cost-saving relative to uncoated catheters (Figure 2). The antibiotic-coated catheter (MR) achieved the greatest health benefits and lowest costs and dominated the use of uncoated and the three antiseptic-coated catheter types (SPC, CH/SSD (ext) and CH/SSD (int/ext)). Compared to uncoated catheters, the use of MR catheters avoided 15 infections and generated 1.6 QALYs per 1,000 catheters placed. The MR catheters also released 32 ICU bed days and 95 general ward bed days and achieved cost savings of AUD $130,000 per 1,000 catheters (Table 3).

The second analysis based on PSA to incorporate uncertainty indicated that at a willingness-to-pay threshold of AUD $40,000 per QALY the average monetary net benefits estimated for each catheter type were very similar with substantial overlap in the 95% CIs (Table 3). Figure 3 shows the distribution of monetary net benefits for the MR, CH/SSD (ext) and uncoated catheters, for clarity the other catheter types have been omitted as their distributions lie over those presented. The MR catheters returned the highest monetary net benefits and represented the optimal choice given current information. They are associated with expected incremental monetary net benefits of AUD $948 per catheter relative to retaining the uncoated type (that is, the average monetary net benefits for a MR catheter, AUD $391,612, minus those for an uncoated catheter, AUD $390,664); however the probability of error in this decision is 62% (Table 4).

For the third analysis, using alternate scenarios for key parameters, the MR catheters maximized monetary net benefits in all cases. However, the probability of error in this decision was consistently high and the 95% CIs for the estimated monetary net benefits associated with these catheters were large (Table 4). The lowest estimate of monetary net benefits arose when bed days were valued at variable cost savings only (that is, the narrow hospital perspective). In this scenario, monetary net benefits associated with MR catheters would be just under AUD $200 per catheter relative to uncoated catheters because the catheters were no longer cost saving instead requiring an increase in expenditure to produce health benefits (Table 4). The highest estimate of monetary net benefits was obtained when an attributable mortality of 15% was assumed. In this scenario, MR catheters were cost-saving and generated high monetary net benefits of nearly AUD $2,441 per catheter relative to retaining uncoated catheters. Scenarios that returned higher estimates of expected monetary net benefits were associated with less uncertainty. Under the high mortality scenario the probability that choosing MR catheters as optimal was incorrect fell to 46%, whilst in the low bed days scenario, where expected monetary net benefits were low, the error probability in this decision was 76%.

Our evaluation suggests any decision regarding the use of A-CVCs in ICU patients is uncertain. The findings from the first analysis, which do not consider uncertainty, concur with existing economic evidence 67. This shows that, for all four types of antimicrobial catheter, health gains will be accompanied by cost savings. Given the assumption of a low attributable mortality and a low rate of infection, expected health gains are minimal and the decision is driven by the change in costs. Most of these costs represent the value of obtaining increased capacity within the ICU, rather than cash savings. Nevertheless, the results of the first analysis imply a decision not to adopt these catheters will harm patients by reducing their health status and increasing their risk of mortality and, simultaneously, waste resources within the healthcare system.

Our second analysis, using PSA, introduces the uncertainty associated with the decision. Based on current information, the MR catheters are the optimal decision because they return the highest net monetary benefits relative to all other catheter types. However, the probability of error in this conclusion is high, at 62%. Our third analysis shows that MR catheters remain the optimal decision across a range of scenarios and quantifies how uncertainty in this decision varies. Uncertainty is lower for scenarios where decision makers believe that attributable mortality is high, where they value bed days highly, or where the starting infection rate is high. This finding fits with conclusions from a recent meta-analysis that suggests that antimicrobial catheters will return a higher treatment benefit when infection rates are high 40, and provides support for current guidelines which recommend reserving their use for settings with high infection rates 9. However, even in these scenarios the probability that this conclusion is wrong, and the MR catheters are not optimal, does not reduce below 46%.

Interpreting the results of cost effectiveness analyses under uncertainty requires decision makers to think beyond conventional error rates as used in statistical analysis. Decision makers looking to maximize health returns from their budget should choose between these catheters by selecting the option with the highest monetary net benefits. Given the current evidence, MR catheters should be chosen even if the probability of error in this conclusion exceeds the standard level of 5% used to define statistical significance. The justification is that a decision not to use them in favor of uncoated catheters would impose economic costs, arising from average monetary net benefits foregone, of AUD $948 per catheter 41 (AUD $391,612 minus AUD $390,664). This conclusion should lead to rapid and sustained uptake of the technology 5, yet their use appears to be limited despite earlier estimates of these catheters being cost effective 67. We suggest that uncertainty over this cost effectiveness evidence may be partly responsible.

Studies have shown that decision makers are heavily influenced by uncertainty 3542. Presenting decision makers with an estimate of uncertainty in the results of an economic evaluation is important for the following reasons: it makes the current state of knowledge about the decision explicit and quantifies confidence (or lack of) in conclusions; it allows them to weigh the cost effectiveness results against other relevant considerations in the adoption decision including their own attitude to risk; and it provides an indication of the value of conducting further research to reduce uncertainty.

Two aspects to uncertainty are important: the probability of making the wrong choice and the potential consequences of getting it wrong. Both elements are required; a decision with a 5% probability of being wrong may still be perceived as uncertain if the consequences are very large. Decision makers tend to be risk averse. Rather than being focused solely on maximizing health returns, they are also concerned with interventions that have the potential to result in harmful outcomes. If there is no potential for harm then decision makers may be happy to accept a new intervention with a high but uncertain benefit. But where the potential for harm is perceived to be high, an existing intervention with a lower benefit may be preferred. Antimicrobial catheters are perceived to carry a risk of a number of negative outcomes that are likely to deter from their introduction, including the potential for a loss of focus on hygiene procedures. There has also been discussion 43 that MR catheters may select for resistant organisms, with higher morbidity and costs 44. This negative could outweigh potential short-term benefits from these catheters 45. An absence of clear evidence 46 makes it difficult to quantitatively incorporate this risk into an economic evaluation 47 but it is an important consideration in the adoption decision.

Decision makers deciding whether to use antimicrobial catheters also have a second choice: whether to collect more information to reduce uncertainty in their choice 48. Value of information analysis 41 can be used to estimate the expected monetary net benefits arising from collecting new information and compare this to the anticipated research costs to indicate whether the research is justifiable. It has been suggested further trials of antimicrobial catheters should be undertaken 43. Due to the relative rarity of infection these will require a large sample size and the involvement of multiple institutions 43, making them an expensive proposition. Estimating the expected monetary net benefits from a trial would indicate if this is the best way to spend research dollars.

Some important sources of uncertainty have been explored in these analyses, but there are other uncertain elements in this decision that have not been explicitly examined. There is evidence the relative effectiveness of A-CVCs, as compared to uncoated catheters, varies according to duration of catheterization 49 and causative organism 50, and there have been reports of toxicity associated with use of particular types of catheter 13. However, a lack of data about these concerns both generally, and in relation to each specific coating, meant we were unable to model their impact. If these aspects reduce the effectiveness of any of the catheter types then its cost effectiveness would also be reduced. Alternatively there may be specific subgroups of patients for whom the cost effectiveness of these catheters can be determined with greater certainty. We did not test assumptions about life expectancy and quality of life in ICU survivors, although these will not alter conclusions about which catheter is optimal as all types will be affected equally.

This evaluation, like those reported in earlier studies, is based on a simplified version of a complex decision. It did not include intangible benefits to reduced infection rates, including the increases to clinical morale and public confidence in the healthcare system demonstrated by the national campaigns to reduce rates of CR-BSI 2 and forming part of the rationale for the introduction of the Deficit Reduction Act 4. Decision makers often consider a wider range of outcomes when deciding on the adoption of a new technology 3542 and clearly the economic value in reducing infection rates goes beyond the capacity released within hospitals. Valuing these intangible outcomes may improve the representation of the economics of preventing infection, but it would be difficult to achieve. It has been suggested that MR-coated catheters are difficult to insert 6, making them unpopular amongst clinicians, but data comparing failure rates for insertion are not available in order to incorporate this cost. Finally, recent research has shown that improving catheter care by intervention 'bundles' is a highly effective way to reduce rates of CR-BSI 51. In an evaluation comparing 'bundles' with antimicrobial catheters, it may be that the former would dominate. This is not evaluated here and deserves rigorous exploration rather than hypothesizing.

Antimicrobial catheters have been available as a means of preventing CR-BSI in the ICU for two decades. Although earlier studies have indicated these devices are cost saving, the findings of this evaluation represent a deeper analysis of the decision than previously available that will help decision makers in any setting considering adopting A-CVCs judge the cost effectiveness of these devices. We have shown that the cost effectiveness of these catheters is uncertain, and are not surprised that infection control decision makers are reticent about using antimicrobial catheters despite the economic evidence. Failure to consider uncertainty generates overly simplistic results and creates skepticism amongst decision makers using them to guide infection control policy. Value of information analyses may suggest where research to reduce this uncertainty should focus, but in the meantime, legislation based on the economics of infection control would do well to consider the complexity of producing good quality evidence in this area.

• The cost effectiveness of antimicrobial catheters to prevent catheter-related bloodstream infection in the intensive care unit is highly uncertain.

• Estimates of cost effectiveness which do not consider uncertainty indicate that minocycline and rifampicin catheters generate the greatest health benefits (1.6 quality-adjusted life years) and opportunity cost savings (AUD $130,289 per 1,000 catheters placed) relative to uncoated catheters.

• When uncertainty in information currently available to inform this decision is considered, MR-coated catheters return the highest monetary net benefits relative to all other catheter types but there is a 62% probability that this is incorrect.

• Value of information analyses may suggest where research to reduce uncertainty in this decision should focus to achieve maximum benefit, but in the meantime, decision makers would do well to consider the complexity of producing good quality evidence in this area.

A-CVC: antimicrobial central venous catheter; CR-BSI: catheter-related bloodstream infection; CH/SSD (ext): chlorhexidine/silver sulfadiazine (external coating); CH/SSD (int/ext): chlorhexidine/silver sulfadiazine (internal/external coating); MR: minocycline and rifampicin; QALY: quality-adjusted life year; PSA: probabilistic sensitivity analysis; SPC: silver, platinum and carbon.

DP has received grant support from AstraZeneca and is a consultant to Three Rivers Pharmaceuticals, Merck, Pfizer, AstraZeneca, Johnson and Johnson, and Sanofi Aventis.

KH coordinated the overall design of the study, collected and analyzed the data and drafted the manuscript. DC aided in defining the clinical context, design of the decision model and collection of data. MW helped conceive of the study and obtain funding. DP aided in defining the clinical context and helped to draft the manuscript. NG conceived of the study and obtained funding, participated in its design and helped to draft the manuscript. All authors read and approved the final manuscript.