Healthcare-associated Infections (HAIs) pose a serious health threat to hospitalized patients with an estimated 4% of hospitalized patients in the United States (US) infected at any given time [1]. To mitigate this threat, actions are being taken by a myriad of public health organizations including government agencies, professional associations, private industry, and consumer groups. These actions include mandatory public reporting of hospital HAI rates and the formation of prevention collaboratives composed of multiple hospitals working together to prevent HAIs [2]. The US Congress passed several acts to empower agencies to implement polices to combat HAIs including the funding of states to develop HAI prevention plans, hospital reporting of select HAIs rates to the Centers for Medicare and Medicaid Services (CMS), and a penalty system to reduce hospital Medicare reimbursements for high rates of HAIs [3–5]. Additionally, the Executive Branch has launched efforts to coordinate various voluntary activities of both public and private sector stakeholders to achieve national goals of reducing HAIs and antibiotic resistant infections [6, 7].

As policy actions to combat HAIs and antibiotic drug resistance increase, economic evaluations of how to efficiently achieve national goals are needed to help inform policy decisions. However, conducting economic evaluations of policies affecting HAIs have been challenging given (1) the quality of hospital cost data, (2) the divergent cost perspectives (i.e. of patients, providers, third party payers, or society) that the analysis can take, (3) the methodological difficulties of assessing both short term and long-term attributable morbidities and mortality, and (4) difficulties in identifying patients with HAIs having their onset post-discharge [8–10]. Historically, most HAI economic analyses have taken the cost perspective of the healthcare provider and /or administrator [9, 11]. These studies attempt to demonstrate the cost savings to hospital budgets resulting from prevented cases of HAIs so as to make a ‘business case’ that investments in infection control reduce treatment costs and improve outcomes [12–14]. Using peer-reviewed attributable cost estimates for select HAIs, a 2012 meta-analysis found that the range of total direct medical costs to the US healthcare system due to hospital-onset central-line associated bloodstream infections, catheter-associated urinary tract infections, ventilator-associated pneumonia, surgical site infections and Clostridioides difficile infection (CDI) was $8.3–$11.5 billion (2012 dollars) [15]. While this work is still important, relying on analyses based on the provider cost perspective ignores the cost impacts to patients (travel costs, lost wages, long term morbidities, insurance co-pays), third party payers (increased per-patient reimbursements), and to society (mortality).

As required by executive order, US federal regulatory agencies conducting economic analyses of regulations impacting human health must take the societal perspective, which includes measuring the economic benefit of mortality risk reductions [16]. The US Office of Management and Budget (OMB) Circular A-4, Regulatory Analysis (2003) also directs regulatory agencies on use of the value of statistical Life (VSL) which is a monetized measure of the additional cost that individuals would be willing to pay for a small reduction in the risk of mortality [17]. An example from Robinson (2007) illustrates the VSL concept [18]. Assuming a population of 100,000, if each individual indicates they would be willing to pay an average of $50 to prevent one death (a risk reduction of 1/100,000), the VSL would be $50 X 100,000 or $5 million.

VSL estimates are derived from survey methods in which respondents are asked what they would be willing to pay for small changes in the risk of premature death (referred to as state-preference studies) or from statistical models that evaluate wage differentials for occupations with varying job-related mortality risks (referred to as revealed-preference studies). Concerns have been raised about the accuracy of VSL estimates from stated-preference studies, particularly related to the potential for hypothetical bias. As respondents are responding to a hypothetical market for risk reductions described within a questionnaire, respondents may provide “spurious” responses that do not reflect how they would respond to an actual market, and thus overstate (or understate) what they would actually pay for a reduction in risk [19–21]. However, there have been advances in both survey design and the use of statistical methods to make adjustments to responses if needed to minimize potential bias [20, 22]. Also, comparison of VSL estimates between revealed-preference and stated-preference studies have found that estimates from revealed-preference studies tend to be higher, which may be a reflection of how the perception of risk may differ between those who actually face the risk (as in wage studies), and those who may discount their own perceived individual risk to a hazard described in a survey [21]. While further research is needed to understand the discrepancies between VSL estimates from the two types of studies, certain types of mortality risk cannot be assessed using wage data (i.e. cancer) and require the use of stated-preference methods [20, 21].

Until recently, only the US Environmental Protection Agency (2014) and the US Department of Transportation (2015) had issued their own guidelines to ensure that their use of VSL in internal analyses complied with OMB directives [23, 24]. The US Department of Health and Human Services (HHS) has now published its own guidelines (2017) to inform and advise department agencies on the methods used in conducting regulatory impact analysis that are consistent with standing executive orders and OMB recommendations [25].

Prior to the HHS guideline publication, CMS had proposed a new rule to require that all hospitals certified by Medicare and/or Medicaid (4900 hospitals and 1300 critical access hospitals) establish and maintain antibiotic stewardship (AS) programs by 2020 as called for in the National Action Plan for Combating Antibiotic-Resistant Bacteria [7, 26]. While the rule also proposed other more modest requirements related to patients right to services (regardless of race, color or national origin), employment of an infection preventionist/infection control professional, mandated review of current infection control programs, and other nursing and medical record services, the requirement associated with the highest costs and benefits in the corresponding regulatory impact analysis was for AS programs. Using results from an economic evaluation of a multifaceted intervention (including AS programs) for CDI in US hospitals based on a Medicare cost perspective, the regulatory impact analysis concluded that requiring AS programs would result in annual savings of over 1 billion dollars to hospitals from reduced incidence of CDIs and drug cost savings from reduce inappropriate antibiotic prescribing [26, 27]. However, it was noted in the regulatory impact analysis that the potential societal benefits of reduced non-fatal CDI illness and the societal benefits and costs of reduced fatal CDI illness had been ignored due to lack of information on these benefits. A request for this information to be considered in the analysis of the finalized rule was made.

Our objective is to provide estimates of both the morbidity and mortality risk reductions associated with an active hospital AS program and enhanced infection control by expanding the previously mentioned Medicare analysis to incorporate a societal cost perspective. The VSL estimates, the methods for assessing intervention costs, and the methods for deriving the value of morbidity risk reductions from the VSL are taken from the new HHS guidelines. The possible ramifications of incorporating the societal perspective in evaluations of HAI prevention programs, and their interpretation by stakeholders and policy makers, will be discussed.

This analysis relied on the results of two recent CDI studies to develop a net present value model to assess the social costs and benefits of a multifaceted CDI prevention program including the economic value of reduced mortality risks (see Appendix for model details) [27, 28]. Our economic model is partly based on a decision analytic (Markov) model developed by Slayton et al. to measure the net benefits of a multifaceted intervention (enhanced hospital infection control practices coupled with the implementation of an AS program) to prevent CDI in Medicare patients in acute care hospitals over a 5 year time horizon [27]. The simulation model incorporated information on projected hospitals discharges, infection incidence rates, intervention effectiveness, prevention program costs, and Medicare reimbursements saved per case averted, with each model representing a cohort of 1000 patients and outcomes assessed for 1000 trials. Using the cost perspective of the federal government (as a third party payer), the national financial savings to the Medicare program (under the base case scenario of 50% program effectiveness and a 3% discount rate) was estimated to be $2.5 billion (2011 dollars) with a credible range of $1.2 billion to $4.0 billion over the five year study period.

To expand this analysis from a Medicare cost perspective to reflect a societal cost perspective, we developed national estimates of the benefits of averted cases and reduced mortality risk for all ages using results from Lessa et al. which derived population-based estimates of the incidence and disease burden for (1) health care-associated Clostridioides difficile (HCA-CDI) (which included community-onset health care–associated, hospital-onset, and nursing home-onset infections) (2) recurrent HCA-CDI cases (within 14 to 56 days after the initial occurrence) stemming from these infections, and (3) the number of deaths occurring within 30 days after the diagnosis of HCA-CDI [28]. The net present value (NPV) model is defined as:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \mathrm{NPV}={\sum}_{t=0}^5\frac{Benefits_t+{Costs}_t}{{\left(1+r\right)}^t} $$\end{document}NPV=∑t=05Benefitst+Costst1+rtwhere:

Benefits_t = the total benefits arising in year t (t = 0,1,2,3,4,5),

Costs_t = the total costs arising in year t (t = 0,1,2,3,4,5), and.

r = the social discount rate (3 and 7%).

The analysis of the net present value model was conducted using Excel for Windows 2016.

The study horizon for the measurement of costs and benefits in our societal analysis is a 6 year period beginning in 2015 through 2020 to be consistent with the published CMS analysis [26]. Unlike the Slayton analysis, we assumed a 3 month lag before benefits start to accrue in 2015 to account for the time needed to implement the intervention. Also, in contrast to the regulatory impact analysis done by CMS, we calculated the cost of the intervention to cover expenses to all inpatient prospective payment system (IPPS) hospitals as opposed to just 60%. The CMS analysis took into account that 40% of IPPS hospitals had already established ongoing AS programs, but we would argue that without the proposed requirement, hospitals would be free to suspend these programs given changes in any clinical or financial conditions facing the hospital [29]. To avoid any potential biases by limiting the coverage of the AS requirement, we fully assessed all cost and benefits that would accrue to all hospitals subject to the rule.

In our high estimate scenario based on 50% program effectiveness and attributable mortality, the national intervention was projected to avert 1,017,953 total inpatient cases and prevent 40,176 deaths using a 7% discount rate, and 1,121,518 total inpatient cases and 44,280 deaths using a 3% discount rate over the study period (Table 2). Using the low estimates of program effectiveness and attributable mortality (10 and 35% respectively), the model projects that 203,591 total inpatient cases and 5625 deaths were averted at the 7% discount rate, and 224,304 total inpatient cases and 6199 deaths were averted at the 3% discount rate.

Also from Table 2, the projected cost of the prevention program ranged from $3.3 billion (7% discount rate) to $4.2 (3% discount rate) at 50% program effectiveness. At 10% program effectiveness, the program costs ranged from $3.7 billion (7% discount rate) to $4.7 billion (3% discount rate). The higher costs under the 10% program effectiveness scenario arise from the increased costs associated with implementing the enhanced infection control practices (to avoid transmission) around the remaining 90% of cases (as opposed to 50% of cases).

Without considering the value of morbidity and mortality risk reductions, the net benefits in reduced patient care costs and reduced antibiotic expenditures from the intervention ranged from $8.1 billion to $9.1 billion (subtracting direct medical cost savings of $13.3 billion and $11.4 billion from the intervention costs of $4.2 and $3.3 billion respectively) under the 50% intervention effectiveness scenarios, but these net benefits decreased in the 10% effectiveness scenarios to $1.3 billion (3% discount rate) to $1.6 billion (7% discount rate) (Table 3). However, when the values for morbidity and mortality risk reductions were included, net benefits from the intervention ranged from $24 billion to $626 billion across all scenarios (Table 3). The inclusion of the economic value of mortality risk reductions using VSL estimates overwhelmed the difference between the intervention costs and the benefits of both the direct medical cost savings and the value of morbidity risk reductions as the total net benefits are substantial. The proportion of the economic value of mortality risk reductions to total benefits ranged from 80% (lowest total benefit estimate) to 97% (highest total benefit estimate).Table 3

Benefits and Costs of a Comprehensive CDI Prevention Program 2015–2020 (2015 $)

Even with VSL estimates that have been age adjusted to reflect the older age distribution of HCA-CDI patients, the intervention still produced a range of total net benefits of $3.7 billion to $14.5 billion under our lowest effectiveness scenario (10% program effectiveness, 35% attributable mortality proportion) (Table 7). As this same intervention program achieved an 80% reduction in hospital CDI cases in England from 2008 to 2012, the credible range of net benefits is $21.2 billion to $625.8 billion which are associated with the 50% program effectiveness scenarios (Tables 3 and 7) [56–58]. Given the evidence on program effectiveness and the theoretical and empirical uncertainties associated with age-adjusted VSL estimates, we suggest that a likely, but conservative, scenario outcome from a public policy perspective is $121.4 billion in total net benefits (50% program effectiveness, 3% discount rate, 35% attributable mortality proportion, and the low VSL estimate). This translates to an annualized net benefit of approximately $25.5 billion.

The societal cost perspective has rarely been considered in economic evaluations of HAI prevention programs, but doing so in accordance with HHS guidelines for conducting regulatory impact analysis may provide stakeholders and policy makers a broader view on the benefits of such programs. As our intervention costs ranged from $3.3 billion to $4.2 billion, these costs would have to be at least 28 times larger to overlap with our lowest benefit estimate of $121.4 under the 50% program effectiveness scenario (Table 3) and would have to quintuple to overlap with our lowest benefit estimate of $21.2 when using age-adjusted VSL estimates and the 50% program effectiveness scenario (Table 7).

While the benefits of reducing mortality risks comprised a vast majority of total net social benefits in our model, a number of other relevant benefits (cost savings) associated with CDI disease are ignored in this analysis. The most important of these include non-hospital medical costs (e.g., outpatient treatment and pharmacy costs), medical costs and the value of morbidity risks due to severe or long-term morbidities, lost labor productivity, and the economic impacts on family/caregivers. More broadly, AS programs also produce spillover benefits in terms of reducing rates of antibiotic resistance, although such economic impacts are diffuse and difficult to quantify and require more research to understand their full effects. However, the addition of these potential benefits just provides more support for the proposed rule. As the range of credible costs and benefits did not overlap, we did not perform a Monte Carlo simulation in our sensitivity analysis.

Regardless of the type of interventions taken for HAI prevention, the inclusion of monetary valuations for morbidity and mortality risk reductions in cost-benefit analyses of HAI disease prevention potentially provide economic justification for interventions that might not otherwise be considered cost saving, as illustrated by this analysis. Our model shows that the total net benefits from having AS programs and enhanced infection control are significantly large enough to cover the total investment costs in AS programs (as opposed to just 25% in our model) as the total annualized intervention cost (totaling $14.6 billion for 2015–2020 at a 3% discount rate) is only $2.7 billion.

Along with the requirement to include AS programs, CMS also proposed additional changes to the conditions of participation including a requirement that hospitals, (1) identify a qualified infection preventionist or infection control professional as an officer responsible for their infection control program and (2) conduct a review of their infection control program [26]. Other changes affecting the writing of restraint and seclusion orders for violent/self-destructive patients by a licensed practitioner, the granting of dietary ordering privileges to qualified dieticians or nutritionist professionals (in critical access hospitals), and the implementation of quality assessment and improvement programs were also proposed. While lacking data on the potential cost savings associated with many of these changes and ignoring the economic value of morbidity and mortality risk reductions from averted infections, the accompanying regulatory impact analysis concluded that the overall proposal resulted in a net benefit to society of $284 million with the majority of costs coming from changes affecting infection control programs and the addition of AS programs. Regardless of the differences in the calculated cost of the proposed CMS rule and the cost of our CDI prevention program, the additional economic benefits of morbidity risk reductions alone can be readily applied to the CMS analysis and would significantly increase the magnitude of net benefits.

A potential limitation to VSL studies of HAI prevention is the variability in methodology and quality of studies that generate attributable mortality estimates given the range of published estimates currently available [33]. The dominance of VSL estimates in the calculation of benefits highlights the need for attributable mortality measures that accurately and consistently reflect the mortality impacts of HAIs. Additional research is needed to improve the measurement of attributable mortality associated not only with HAI but with any cause of disease-related death. Given the decades of research on the VSL, the measurement of the VSL and the estimates currently used are generally accepted for injury-related risk reduction while the evidence suggests that VSL estimates for illness-related risks are probably similar [59]. However, the quality of data on mortality associated with disease should also be scrutinized as the new HHS guidelines do not directly address this issue. Use of unadjusted-crude mortality estimates, as opposed to age-adjusted or risk-adjusted estimates can produce significantly different estimates of the benefits of mortality risk reductions in cases of HAI [55, 60].

In the case of CDI disease, an important factor that likely increased attributable mortality from 2000 through 2010 was the emergence and spread of the epidemic, hyper-virulent, North American Pulse-field type 1 (NAP1) or ribotype 027 strain [61, 62]. Although declines of the NAP1/027 strain in the United States have not been as dramatic as that seen in places like England, where declines have occurred, it is likely that declines in attributable mortality have followed [58, 63]. Ironically, however, it may be AS, alone or in combination with infection control, that has led to most dramatic declines in NAP1/027 [64]. In addition, the development of new therapies and recommendations for generally more aggressive treatment of HAIs, like CDI, may also result in declines in attributable mortality but also add their own costs that must be considered [64, 65]. However, in the case of AS there may be additional impacts on morbidity and mortality that are yet to be fully understood as growing evidence suggests that the effect of unnecessary antibiotics on the microbiome may result in other adverse outcomes among hospitalized patients such as sepsis [66]. Such impacts, even if only partially realized, could even more dramatically sway the cost benefits in favor of aggressive stewardship interventions.

While our analysis illustrates the potential impact on benefits measurement when the value of mortality risks are considered with regulations that impact health, this analysis also illustrates the role that cost-benefit analysis, as opposed to cost-effectiveness analysis, has in rule making that involves public health and safety by the US Federal Government. The cost-benefit analysis described here can easily be adopted by other countries as VSL estimates, especially for mortality risk reductions related to air quality improvements, have been developed for many locations, including Europe, Asia and Australia [67–69]. While the National Health Service in the United Kingdom relies on cost-effectiveness analysis in their decision-making for assessing adoption of new healthcare interventions, HM Treasury has had guidelines (referred to as “The Green Book”) for government ministries on how to incorporate the “Value of a Prevented Fatality” (another name for the VSL) in policy assessments and evaluations of government actions that involve risks to life and health [70]. To better understand the characteristics of VSL estimates from different countries, researchers at the Organization for Economic Co-operation and Development (OECD) have conducted a meta-analysis of VSL estimates related to environmental, health and transport polices (from stated-preference studies) that have been done around the globe [69, 71]. OECD has also published a VSL user’s guide to better inform policy makers on how to incorporate VSL in policy decisions [72]. The median and mean VSL values (using the full dataset) of $2.4 and $7.4 million (in 2005 US dollars), along with other evidence from this study, can potentially be adopted for use in a global economic assessments of antimicrobial resistance [69].

Although progress has been made, HAIs still pose a serious threat to patients across healthcare settings. A recent study on the prevalence of HAIs in US acute care hospitals estimated that the total number of HAIs occurring annually was 722,000, of which there were 75,000 HAI-associated deaths [1, 73]. Along with CDI, carbapenem-resistant Enterobacteriaceae (CRE) has been classified as an urgent threat to the public health in the National Action Plan for Combating Antibiotic-resistant Bacteria (White House 2015) [7]. The Plan has set targets for a reduction in incidence of both CRE and CDI, calling for 60% reduction in hospital-acquired CRE and 50% in overall CDI infections by 2020.

Our study accounted for the economic value of morbidity and mortality risk reductions, components of the total societal health benefits that have not previously been included in cost-benefit analyses of HAI prevention programs. As the US federal government intensifies its efforts to control antibiotic resistant infections, our results suggest that these ambitious goals can produce very large net societal benefits. As these benefits accrue mostly to patients, policy makers can address how the burden for the additional prevention costs should be shared among patients, third party payers and healthcare providers.