Sara E. Cosgrove
Eli N. Perencevich
Organized infection control (IC) programs have expanded in the past decade. However, hospital administrators require more economic justification for maintaining and expanding programs. This is a particular challenge because while it is clear that healthcare-associated infections (HAIs) are costly—$6.5 billion in 2004 dollars in the United States [1]—healthcare administrators do not generally view IC programs as cost-saving because they do not generate revenue for the institution. Thus, hospital epidemiologists and infection control professionals (ICPs) need the tools to prove the worth of the surveillance and interventions that they perform for their healthcare institution.
HAIs are a significant risk to patient safety. Unfortunately, this has not opened the door to improving the resources provided to prevent HAIs. While society would benefit from reduced HAIs, there is currently no direct reimbursement to hospitals for the purpose of IC. This has led to the current situation in which hospitals must make economic decisions about funding IC studies on an individual basis. This situation also has impacted the literature so that most studies describe the hospital perspective of the impact of HAIs (90% of studies) with only 3% taking a societal perspective [1,2]. As we emphasize here, it is important to complete a business cost analysis from a hospital perspective in order to inform local decisions; however, these such analyses are not useful on a public health level. It has become increasingly important to justify the importance of funding IC activities at a broader level through the completion of cost-effectiveness analyses from the societal perspective.
This chapter will detail important concepts in economic analysis, including types of economic analyses and their strengths, the different perspectives of analyses, and placing monetary values in constant dollar terms. Then it will describe approaches to assess the financial impact of specific HAIs and of control interventions and provide a methodology for developing a cost-effectiveness analysis at a societal level. After completing the necessary review of health economics, it will describe the basic steps needed to complete a business case analysis of a specific IC intervention for an individual institution. While we have attempted to outline important considerations regarding economic measurement of HAIs and related interventions, more detailed texts about design and analysis of economic research are available [3,4,5].
P.236
Basic Economic Concepts
This section defines important concepts regarding economic analyses including types of cost analyses, the perspective of the analysis, and discounting and inflating costs.
Types of Economic Analyses
There are four basic types of economic analysis used in health care: cost minimization analysis, cost-effective analysis, cost-utility analysis, and cost-benefit analysis (Table 16-1). Experts have noted that the distinctions between these various forms of analysis are often blurred, yet it is important to consider what is included and not included in each specific analysis so that informed decisions can be made [3]. A recent review of the IC literature found that of 30 publications reporting to be economic analyses, only 8 were cost-effectiveness or cost-consequences analyses. Of note, in the IC and quality improvement literature, there is increasing use of the term business case analysis, which is an extension of a simple cost analysis from a hospital or payer prospective that excludes dollar valuation of human life and morbidity [6,7].
A detailed discussion of performing a business case analysis appears later in this chapter.
|
TABLE 16-1 |
||||||||||||||||||
|
Cost-Minimization Analysis
In cost-minimization analyses, the effectiveness of two interventions or products are assumed to be the same (equal efficacy and side effects), and the analysis is aimed at determining which can be delivered least expensively [8]. An example of a cost-minimization analysis in IC is the choice between two brands of nonlatex gloves. In this example, most would just choose the less expensive brand. Note that this type of analysis does not apply to the choice between a brand of latex and a brand of nonlatex gloves, because these can be associated with different levels of health care worker (HCW) satisfaction and also allergic side effects.
Cost-Effectiveness Analysis
In contrast to cost minimization, cost-effectiveness analysis compares interventions or products that have different costs and different effectiveness. If a specific new intervention costs more and is less effective or alternatively costs less and is more effective than an existing intervention, the choice is easy. However, if a new intervention delivers more at increased cost, which occurs frequently in the setting of rapid technologic intervention, the choice often is difficult. In cost-effectiveness analysis, the benefits of an intervention are measured in the most natural unit of comparison, such as lives saved or infections prevented [3]. Programs then are compared in terms of dollars per life-year gained or dollars per infection prevented.
Cost-Utility Analysis
Cost-utility analysis is very similar to cost-effectiveness analysis except that benefits of a specific intervention are adjusted by health preference scores or are utility weighted [3]. Thus, programs are compared in terms of quality-adjusted life years gained (QALY). The rationale of this approach is that it allows the incorporation of disability or side effects associated with the condition being treated or the treatment side effects. For instance, a year spent in an intensive care unit (ICU) would be valued differently by a patient compared to a year spent at home with his or her family. Perhaps four years spent in an ICU would be equal to one year spent healthy, so four years spent in an ICU would equal only one healthy year or rather, one QALY. A good example of a cost-utility analysis (and a cost-effectiveness analysis) in the IC literature is one that studied the use of vancomycin as perioperative prophylaxis during coronary-artery bypass graft surgery [9].
Cost-Benefit Analysis
Cost-benefit analysis is a different form of economic evaluation in that all aspects of the analysis, including the consequences of the intervention, are valued in monetary or dollar terms. If an intervention's benefits measured in dollars exceed its costs, then it is considered worthwhile [4]. The major impediment to the use of cost-benefit analysis in healthcare is the requirement to value human life or health benefits in monetary units, such as a human life-year equaling $200,000. Of note, most economic analyses of IC interventions that claim to be cost-benefit analyses are mislabeled because they do not include a dollar value for the important outcomes of interest (e.g., they do not place a dollar value on a human life or quality of life and do not include dollars saved from saving a life or improving quality of life in the analysis).
P.237
Which Type of Analysis is Preferred?
Over the past 10 years, cost-effectiveness analysis and the closely related cost-utility analysis have emerged as the preferred methods for economic evaluation in healthcare [4,7]. Importantly, it is recommended to compare new interventions to a reference case whenever possible using standard units such as cost per lives-saved or QALYs-saved [4]. If an agency wanted to choose between funding a hand-hygiene promotion initiative and a cancer-screening program, it would be difficult to compare cost per infection prevented with cost per cancer detected. However, if the comparison was cost per life-years saved or cost per QALY saved with each program, an informed decision could be reached.
|
TABLE 16-2 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
What Is Considered Cost Effective?
A standard threshold for calling a program cost effective is for the intervention to cost less than $50,000/QALY saved; however, some suggest the threshold has increased to $100,000/QALY saved [10]. The World Health Organization recommends that a threshold for calling an intervention cost effective be three times the country's gross domestic product per capita, so this threshold is $94,431 in the United States [11]. Frequently, but incorrectly, researchers will state only that an IC intervention is cost effective or cost beneficial if it is cost saving from a hospital perspective. Most healthcare interventions are not cost saving. A review of all cost-effectiveness analyses published between 1976 and 2002 found only 130/1433 (9%) cost-effectiveness ratios actually saved costs, meaning that they saved lives and money at the same time [12].
Perspective
The economic impact of HAIs and interventions can be assessed from the perspective of society, the hospital, a third-party payer (e.g., a health maintenance organization or the Centers for Medicare and Medicaid Services [CMS]), a government agency (e.g., the Veterans Health Administration [VHA]), or the patient. Studies that examine one perspective can underestimate the full economic effect of an infection or intervention; thus, it is important to recognize the perspective of a study to appropriately interpret its results and to design a study from the perspective of interest (Table 16-2). For instance, outpatient physician visits to treat a surgical site infection (SSI) would be important to include in an analysis for the CMS but would not be included in a typical acute care hospital cost analysis.
The societal perspective is one that incorporates all costs and all health outcomes, regardless of who incurs the costs and who obtains the benefit [4]. Typically, except in instances when a specific organization is funding the analysis, the investigator should choose the societal perspective, which is the broadest and most useful when comparing disparate medical interventions. The U.S. Panel on Cost-Effectiveness in Health and Medicine states that even when a particular analysis is requested from a nonsocietal
P.238
perspective, a complete societal perspective analysis also should be completed [4]. Importantly, a societal perspective analysis will inform broader comparisons of programs and could lead to more equitable distribution of resources to improve public health. It is possible that an analysis from the societal perspective would suggest a different strategy than a more limited perspective [4].
For example, an economic analysis done from a hospital perspective would not include patient morbidity (e.g., reduced functional mobility) and outpatient drug costs (Table 16-2). Thus, a hospital might decide not to fund a SSI prevention program, because it will cost more in implementation and equipment costs than it could recoup through reduced SSI costs, such as in shortening length of stay or decreasing antibiotic costs. However, an insurance company that must pay for additional outpatient physician visits, medications, and home-health visits attributable to the preventable SSIs might want to fund the same SSI prevention program. Of course, neither the hospital nor the insurance company perspective includes patient morbidity, mortality, and other important factors, such as the opportunity cost of lost wages. A societal perspective would include all such factors. It might be that a proper cost-effectiveness analysis of the SSI prevention program showing large cost savings and lives saved from a societal perspective would inform CMS or VHA to fund the program to the entire society's benefit. It is possible that the current lack of societal perspective cost-effectiveness analyses of IC interventions has facilitated the current underfunding of IC programs and the continued incidence of preventable HAIs.
Placing Monetary Values in Constant Dollar Terms
Adjustment for Inflation
When data on costs used in economic analyses come from different years, they should be brought into current year values. For instance, if you wanted to include the cost of a methicillin-resistant Staphylococcus aureus (MRSA) bacteremia in a business case analysis for your hospital and you had only an estimate of the cost from 2002, you would need to inflate that amount to current year dollars. The typical method for handling these adjustments is to inflate the dollar amounts using a standard price index (e.g., the Medical Component of the Consumer Price Index) [4,13].
Discounting
It is widely accepted that in economic analyses, all future costs and future health consequences should be stated in terms of their present value [3,4]. The process of converting both future dollars and health outcomes to their present value is called discounting. The U.S. Panel on Cost-Effectiveness in Health and Medicine recommends using a discount rate of both 5% and 3% [4]. For example, if you assume that you will save $10,000 in preventing a MRSA infection next year if you decolonize a patient with intranasal MRSA colonization this year, using a 3% discount rate, the discounted savings would be $10,000/(1 + 0.03)n, or $9,709, where n is the number of years in the future the benefit is accrued.
Measuring the Economic Impact of an HAI or Infection Control Intervention
Measuring the economic impact of an HAI or an intervention to reduce HAIs is important for two reasons. First, these data can be valuable at the local institutional level. Obtaining data regarding the incidence and attributable cost of an HAI allows an individual institution to understand the financial burden of the HAI, and assessing the impact of an intervention is critical in determining whether it is successful and whether extensions of the interventions should be planned. Second, results regarding costs associated with infection and cost savings associated with interventions provide raw data for use in cost-effectiveness, cost-utility, and cost-benefit analyses. This section describes the design and analysis of studies that quantify the impact of HAIs and that measure the outcomes of IC interventions.
Measuring the Attributable Cost of HAIs
Many studies that aim to define the attributable cost of HAIs have been published. Generally, these studies involve a set of patients who develop the infection of interest and a reference group who do not develop the infection. Outcomes such as attributable mortality, length of hospitalization, and costs are compared for the two groups. These studies are by definition cohort studies because the outcomes of interest (e.g., morbidity, mortality, and cost) occur after the exposure of interest (HAI). Examples of these studies include examinations of the mortality and costs associated with central venous catheter–associated bloodstream infection (CVC-BSI) or MRSA-SSIs [14,15].
Defining Costs
Of critical importance is deciding which “costs” to measure. Potential approaches to evaluating the economic burden of HAIs within an institution include measurement of hospital costs, hospital charges, resources used, and actual reimbursed charges [16]. Hospital costs include daily operating costs (sometimes called fixed costs) that do not vary based on patient volume and the cost of drugs, tests, and other patient care–related activities (sometimes called variable costs), which depend on the number of patients admitted or their length of hospitalization [17]. A hospital must ensure that all of its costs are reimbursed; therefore, it assigns fees to hospital resources that are seen on a patient's bill as charges. Insurance companies, Medicare, and Medicaid will not pay the amount on the bill because they receive discounts; therefore, the
P.239
charge on the bill for all patients is greater than the actual hospital costs in order to cover these “losses” [18]. Hospital costs can be a useful outcome measure for an individual hospital because they best reflect the actual economic burden of the hospital. However, while some institutions have implemented complex cost accounting systems that track resources used and assign costs, in most institutions costs are difficult to retrieve [19]. In contrast, hospital charges are less reflective of actual cost but usually are easy to retrieve from administrative databases and are consistent from patient to patient in most settings. Because hospital charges typically overestimate actual cost by 25% to 67%, adjustment using cost-to-charge ratios can be performed [19,20]. Both hospital and departmental cost-to-charge ratios are determined annually based on data submitted to the CMS. Hospital cost-to charge ratios may be a more accurate measure of costs for a cohort of patients in multiple diagnostic related groups (DRG) while departmental cost-to-charge ratios may be more accurate for a cohort of patients in the same DRG [19,21,22].
Direct measurement of resource utilization through the use of microcosting assesses specifically what services or procedures are used by a patient. However, for comparative purposes, use of resources must be translated into monetary value by multiplying the number of tests by their cost or charge. It is important to note that physician professional fees and costs to the patient in the form of lost work are not captured when assessing only hospital costs or charges. In addition, economic measures of health care are not necessarily set by a market-based pricing system. The costs of care for a specific patient are artificial and arbitrary computations that may vary between sites and at different time periods.
Depending on the perspective of the study, the investigator must determine what proportion of hospital costs are reimbursed by payers. If a portion of the costs of an infection are reimbursed by an insurer, then only the portion not reimbursed should be included in a cost analysis from a hospital perspective [23]. This also is the proper cost estimate of a specific HAI that should end up in the business case analysis discussed later. Hospitals may use different ways to limit costs based on their method of reimbursement. For example, if reimbursement occurs per diem, the hospital will focus on reducing costly days of stay, such as ICU or surgery days, rather than the total length of stay; if reimbursement occurs based on the DRG or capitation, total expenses are the focus of cost reduction.
The ratio of the total costs or charges of patients with HAIs compared to those without HAIs within one institution over a relatively short period provides the most generalizable estimate of the magnitude of the economic impact of HAIs. In contrast, absolute values of cost or charge cited in studies should be interpreted with more caution because they may not be applicable beyond the institution in which they were collected. It is important to note that some administrators may view business case analyses with skepticism if the cost data used are not from the local institution. Multicenter studies must report measures that are standardized across institutions.
If costs of HAIs cannot be measured within your own institution, it may be necessary to use literature sources for the estimation of the economic impact of specific infections before completing a business case analysis of an intervention. A synthesis of the published literature on the cost of HAIs was published by Stone et al. both for the periods from 1990–2000 and from 2001–2004 [1,24]. Their results were limited by the different methods that costs are estimated in the studies that they evaluated but provide the most complete available data of the costs associated with the most common HAIs (Table 16-3).
|
TABLE 16-3 |
|||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||
Methodologic Issues in Cost Outcomes Studies
Several methodologic issues in the design of these cohort studies merit discussion, including controlling for length of hospitalization before infection, adjusting for underlying severity of illness, and selecting of the reference group.
Adjustment for Length of Hospitalization Before Infection
Adjustment for differences in length of hospital stay before the onset of infection in patients with HAIs and total length of hospital stay of the comparator group who did not develop an HAI is important because there is a direct correlation between length of hospitalization and risk of HAI, cost, length of stay after infection, and mortality. Studies in which no adjustment for the “time at risk” for the development of an HAI has been shown to overestimate the length of hospitalization and costs that are attributable to the HAI by up to twofold [25].
Several methods have been proposed for accurately estimating the extra days spent in the hospital as a result of HAIs and the associated increased costs. At a minimum,
P.240
patients in the reference group who do not develop an HAI should be hospitalized at least as long as the patients who develop an HAI were hospitalized before acquiring the HAI. This can be accomplished by matching case- and comparator-patients based on length of stay before the HAI or by performing more complicated statistical analyses [25].
Adjustment for Underlying Severity of Illness and Comorbidities
In addition, care must be taken in controlling for pre-HAI illness severity and comorbidity. In studies assessing the impact of HAI, adjustment for underlying illness severity and comorbidities is essential because patients who develop HAIs often have a more severe underlying disease than those who do not, which can independently result in adverse outcomes.
Various methods have been proposed and employed to grade severity of illness, including subjective scores, ICU-data driven measures, or administrative severity scores. However, there is currently no well-validated aggregate illness severity score for infectious disease outcomes. McCabe and Jackson used a simple three-category score to predict mortality in patients with gram-negative bacteremia [26]. This scoring system is widely used but subjective, based completely on the judgment of the individual reviewing the patient record. No objective physiologic data are included, limiting its generalizability from study to study. This system works better as predictor for mortality than as a predictor of morbidity and cost.
Other scores that have been proposed also have significant limitations. The Acute Physiology, Age, and Chronic Health Evaluation (APACHE) score relies heavily on physiologic parameters, the majority of which are collected only in ICU settings, and the score has been validated only to predict mortality in ICU patients [27]. Scoring systems such as the Medical Illness Severity Grouping System (MedisGroup) admission severity group score and the All Patient Refined Diagnosis Related Groups (APR-DRG) that were developed for administrative purposes for risk adjustment have questionable utility in predicting infectious disease outcomes and need further evaluation [28].
The timing of the assessment of underlying disease severity is of significant importance. Severity of illness is strongly influenced by the presence of infection and, therefore, may represent an intermediate variable in the chain of events between the exposure (i.e., the infection) and the outcome of interest if assessed when the patient is actively infected. Because adjustment for an intermediate variable usually causes an underestimation of the effect of the exposure of interest on the outcome, care must be taken to assess severity of illness before (e.g., >48 hours) the first signs of infection [29]. Results of studies that assign the illness severity score at the time of the infection should be interpreted with caution because they may underestimate the magnitude of the effect that resistance has on outcomes [30].
Aggregate comorbidity measures such as the Charlson Co-morbidity Index [31] or the Chronic Disease Score [32] have been used to summarize patients' underlying comorbidities for the purpose of adjustment in studies examining risk factors and outcomes of patients with HAIs [33,34,35,36]. In particular, these scores can be a useful method to summarize the degree of comorbidity when it is not feasible to include all individual comorbidities in an analysis with small numbers of subjects.
The Charlson Comorbidity Index was originally designed as a measure of the risk of one-year mortality attributable to comorbidity in a prospective study of hospitalized patients and has been adapted so that it can be calculated using International Classification of Diseases, Ninth Revision (ICD-9) codes obtained from administrative databases [37]. Commonly, it is used, although not well validated, in HAI risk factor and outcome studies.
The Chronic Disease Score is calculated based on current medication use. It was originally based on outpatient medications and used to predict physician-rated disease status, self-rated health status, hospitalization, and mortality. It has since been modified by investigators to be based on medications prescribed on the day of hospital admission to predict SSI risk and the economic impact of SSI [34,38]. In addition, other investigators have developed and validated new comorbidity risk measures based on the Chronic Disease Score for use in HAI risk-factor studies and infections due to MRSA or vancomycin-resistant enterococci (VRE), although these have not been validated to predict outcomes due to infections [36].
Selection of the Reference Group
The majority of studies assessing HAIs have compared outcomes in patients with the infection of interest to patients without the infection. This design assesses the independent impact of HAI acquisition. However, studies that aim to assess the impact of an HAI caused by a specific organism with a particular antibiotic resistance pattern may have two reference groups, one with infection due to the susceptible organism and one without infection. For example, the outcomes of patients with SSI caused by MRSA can be compared to those with SSI caused by methicillin susceptible S. aureus (MSSA) to determine the incremental cost associated with methicillin resistance or can be compared to patients without infection to determine the cost associated with MRSA SSI. The latter type of comparison results in a much higher estimate of adverse events attributable to resistance [15].
Measuring the Economic Impact of Interventions to Reduce HAIs
Optimal decisions concerning IC programs must incorporate the economic impact of the specific interventions. Most of the utility of economic analyses in the area of IC is in convincing hospital administration or public health authorities
P.241
to fund and support a specific intervention. Unfortunately, the current literature is lacking in high-quality studies, such as randomized controlled trials, that can be used to support the effectiveness and cost effectiveness of specific interventions.
Decision making around infection-control interventions requires the availability of proper cost-effectiveness analyses. Several important papers have outlined the optimal methodologies to use when measuring the economic impact of antimicrobial-resistant pathogens [16,20]. However, a 2005 survey of all infection-control intervention studies published found that 69% used the quasi-experimental design, and only 4% incorporated a cost analysis [39]. From the period January 2001 to June 2004, of the 30 studies claiming to be economic analysis of infection-control interventions, only 5 were proper cost-effectiveness analyses [1]. So few studies have been published that assess the cost effectiveness of interventions that there is a glaring need for proper economic evaluation of most infection-control interventions. Importantly, even in the few studies completed, many have inherent methodologic weaknesses that bias against reporting an infection-control intervention as “cost effective.” Following are the strengths and weakness of the basic study designs, which should be used when assessing the effectiveness and cost effectiveness of IC interventions.
Randomized Controlled Trials and Cluster-Randomized Control Trials
IC interventions can be broken down into two basic categories. The first is when the patient who is being intervened on is the same patient who directly benefits from the intervention. An example of this type of intervention is optimal timing of antibiotic prophylaxis to reduce SSI risk [40]. In this example, the person who receives the correct antibiotic(s) at the correct time would be at reduced risk of developing an SSI and no other patients in the hospital would directly benefit from this intervention. Therefore, it can be said that the “unit of analysis” is the individual patient if the purpose is to try to measure the benefit of appropriately timed antibiotic use. In this example, the gold standard study design to evaluate the efficacy and safety is the randomized controlled trial. Even though observational trials, such as cohort studies, can yield similar results to randomized controlled trials, a randomized controlled trial is considered the gold standard for evaluating the efficacy of interventions [41,42,43].
The second category of IC interventions is one in which the specific IC program is directed at either individual patients or a specific population of patients and a group of patients benefits from the program. An example of this type of intervention is active surveillance culturing for MRSA and isolation of colonized patients in a medical ICU setting. To study these types of programs, a cluster randomized trial may be most useful to adjust for the clustering effect that is inherent in control programs of transmissible infectious diseases [44,45]. Patients impacted by these types of IC programs represent a cluster (e.g., an ICU) exposed to a common environment, care practices, and other patients who are colonized with MRSA. Studies that fail to control for the nonindependence of patient outcomes may overestimate the effectiveness of the intervention. Thus, the “unit of analysis” in this case is the entire ICU if the purpose is to try to measure the benefit of active surveillance in reducing MRSA colonization and infection. Instead of randomizing individual patients, individual ICUs need to be randomized so that multiple hospitals will need to be involved at great economic and time costs. These types of trials are called cluster randomized trials or group randomized trials and are used increasingly by public health officials to study group interventions and individual interventions that have group-level effects [46]. Numerous articles have been written about the specific methodologic and ethical issues in cluster randomized trials [45,46,47,48,49].
Situations in which randomized trials cannot be ethically completed are frequent in hospital epidemiology, such evaluating the costs and effectiveness of an intervention to stop an active outbreak [50,51]. Quasiexperimental studies, also known as pre-post intervention studies, and decision analytic models can be used when it is not feasible to perform randomized controlled trials or cluster-randomized control trials.
Quasiexperimental Studies
Like randomized controlled trials, quasiexperimental studies aim to demonstrate causality between and intervention and an outcome [52]. They differ from randomized trials in that patients are put in the intervention arm and control arm without randomization. As a result, these studies have a potentially lower internal validity in that multiple confounders and biases can affect their quality [53]. Even with these limitations, the quasi-experimental study design has been used with increased frequency in IC research; a 2004 review of the published studies that assessed interventions to reduce HAIs found that 69% used a quasiexperimental design and 23% used a randomized trial [39].
The most basic quasiexperimental design is the one-group pretest and posttest design in which a preintervention period is compared to a postintervention period in the same population. An example of this type of study is one in which MRSA infection rates and associated treatment costs in a medical ICU are measured one month before and one month after an intervention when patients are bathed with chlorhexidine. One would expect MRSA rates and associated costs to fall after the intervention, but because there is only one measurement before and after and no control group, many alternative explanations for the fall in MRSA rates could exist.
There is a variety of suggested improvements in the basic quasiexperimental design. These include adding multiple preintervention and postintervention measurements of rates and costs (e.g., increasing the number of months in
P.242
which MRSA infection rates and associated treatment costs are measured before and after the intervention), including a nonequivalent control group population (e.g., comparing changes in MRSA rates and costs in the medical ICU to those in a surgical ICU when the intervention has not occurred during the same time period, and removing the intervention (e.g., comparing MRSA rates and costs before the intervention, during the intervention and after the intervention is stopped). A more detailed explanation of options for quasi-experimental study designs can be found elsewhere [52,53,54,55].
A recent systematic review of IC and antibiotic stewardship articles published during 2002 and 2003 found that 39/73 studies (53%) used the most basic quasiexperimental study design with single measurements before and after the intervention and no control group [54]. Importantly, studies that assess the cost effectiveness of specific IC interventions using the basic quasiexperimental design should be interpreted with caution.
Decision-Analytic Models and Mathematical Models
Mathematical models are useful tools for evaluating intervention strategies before implementing them in human populations [56,57,58,59,60,61,62]. Models allow the use of existing knowledge and data in a rigorous, efficient, and testable manner toward the goal of making medical decisions for the assessed population's greatest benefit. Importantly, clinical trials are expensive, labor intensive, and do not necessarily answer the question adequately for populations with all possible baseline characteristics. Creation and analysis of mathematical models can typically be done more quickly and allow investigation within populations with varying characteristics. Therefore, models can be an ideal way to determine which interventions are most cost effective and when they are most cost effective in preventing the spread of transmissible pathogens, including MRSA or VRE [62,63,64].
As an example, active surveillance and isolation of patients as a tool to control the spread of resistant organisms in hospitals has been available for years but is implemented only in a minority of hospital ICUs due to perceived costs and lack of definitive clinical trial or other data [65]. Many factors or variables that are related to the population (e.g., size of an ICU, discharge rate), individual patients (e.g., comorbidities, age), or infectious organism being evaluated (e.g., duration of colonization, likelihood of infection) can be individually evaluated with modeling strategies to assess their individual and combined importance in causing the observed outcome. This evaluation is called “sensitivity analysis” and is used in most mathematical and decision models [62,66]. Thus, mathematical models can focus future clinical trials, greatly benefit patients, and optimize the expenditures within the limited budgets of microbiology and IC departments. Given the number and great variety of hospitals and other healthcare institutions that exist, it would be nearly impossible to perform clinical trials to test the cost effectiveness of all potential IC interventions.
Performing a Cost-Effectiveness Analysis
Plenty of existing works describe the step-by-step completion of a cost-effectiveness analysis [3,5,66]. A complete description of a cost-effectiveness analysis is beyond the scope of this chapter; however, it is important to review the steps typically taken in such an analysis to better interpret the literature for use in your specific clinical or hospital situation. Undertaking a thorough cost-effectiveness analysis is quite complicated and usually is completed with the assistance of a healthcare economist. It is important to contrast the methods and results of a cost-effectiveness analysis to those of the more commonly used business case cost analysis, which we discuss in detail later.
Cost-Effectiveness Analysis Example
The method of completing a cost-effectiveness analysis can be broken down into several steps starting with the clear statement of the problem and the proposed interventions. For example, you may wish to compare several different interventions or strategies for reducing CVC-BSIs. These strategies might include an education program that improves CVC-insertion techniques, use of antibiotic-coated catheters, scheduled replacement of catheters, or elimination of femoral venous site catheters. Inherent in the initial framing of the problem is the determination of the perspective of the analysis, such as hospital or societal perspective.
The second step is then to develop a conceptual model for the infection of interest and the potential interventions [5]. A conceptual model allows the investigator to describe the full range of outcomes and costs that occur with the condition of interest and are potentially impacted by the interventions under study. In our example, patient outcomes attributable to CVC-BSI could include excess length of stay, ICU admission, increased antibiotic exposure, and associated mortality. Costs could include acute care hospital costs, outpatient treatment costs, and lost wages. In many instances, the conceptual framework in a cost-effectiveness analysis is a decision-analytic model in the form of a decision tree (Figure 16-1). Thus, the creation of a decision tree is commonly used to frame the conceptual model and complete the analysis. A decision analysis is not the only methodology available for completing a cost-effectiveness analysis; however, it is now the standard method used because it allows for a sensitivity analysis.
After a framework in the form of a decision tree is completed, the next step is to gather the data necessary to complete the analysis, such as the probability of each outcome and the expected decrease (or increase) in each outcome under each intervention. In our example, it would
P.243
be important to have an estimate of the daily probability of a CVC-BSI for each day that a CVC is in place. In addition, the probability and length of both excess hospital and ICU stays and mortality associated with BSI would be needed. Most importantly, one would need the proportion of CVC-BSIs prevented under each of the potential interventions and the costs of each intervention.
|
Figure 16-1 Hypothetical decision tree comparing use of antibiotic-impregnated catheters to a “do nothing” strategy in controlling central-venous catheter associate bacteremia. |
The data inputs for the decision analytic model can be gathered from existing published literature or from primary data collection using existing hospital administrative or clinical databases. Use of existing literature is important because investigators cannot be expected to complete every possible clinical trial to obtain data for the analysis. Investigators should preferentially use randomized clinical trials, followed by well-designed upper-level quasiexperimental trials and other observational studies [52]. Occasionally, estimates are generated by the use of expert opinion; however, this is typically discouraged and should be carefully examined during the sensitivity analysis.
When all of the necessary outcome probabilities and cost estimates are available, the decision tree can be solved and an estimate of the cost effectiveness of each intervention is completed. Basically, the completed analysis yields a net benefit (benefit of Intervention A minus benefit of Intervention B) and net cost (cost of Intervention A minus cost of Intervention B) of one intervention compared to another alternative intervention. The cost effectiveness is the ratio of net cost to net benefit (e.g., $5,000 per infection prevented, or $15,000 per life saved). In cost-utility analysis, outcomes are presented in QALYs so that the results would be reported in cost per QALY saved (e.g., $2,000 per QALY).
Decision model analysis, like all types of epidemiologic investigation, is associated with uncertainty in the results, particularly when many data inputs are from lower-level studies. It is likely that several input parameters used in the model will be uncertain or have wide confidence intervals. Importantly, a sensitivity analysis should be performed by varying the model's parameter data and model structure within expected ranges to confirm the model's predictions and assess under what assumptions (e.g., excess length of stay or mortality associated with a CVC-BSI or cost of antibiotic coated catheters) an intervention will be most cost effective in reducing the infections of interest. Sensitivity analyses improve the generalizability of the reported findings so that individual institutions or systems can determine under what conditions the intervention will be cost effective in their own unique hospital or healthcare system.
Performing a Business Case Analysis
Given the current reimbursement structure, IC programs often are cost centers, not revenue generating, so they are identified as potential areas for budget cuts [67]. In fact, many IC programs have faced downsizing in recent years [68,69]. Demonstrating value to administrators is increasingly important because health executives are faced with many initiatives and shrinking budgets [70]. To fend off downsizing, IC programs often must complete a business case economic analysis to initiate a new program or justify continuing a program during budget negotiations.
A business case analysis is a type of hospital perspective cost analysis because it typically leaves out patient outcomes. Broadly defined for use in a healthcare improvement intervention, a business case “exists if the entity that invests in the intervention realizes a financial return on its investment in a reasonable time-frame, using a reasonable rate of discounting” [6]. The reasonable return can be through profit, reduction in losses, or cost avoidance. In this instance, the purpose is to look purely at the dollar costs and benefits of an IC intervention or entire program to justify its existence to hospital administrators. The difficulty in making a business case cannot be overlooked because many IC programs lack the economic expertise necessary to complete such an analysis. Anyone considering a business case analysis should contact local institutions' finance administrators for assistance in using the available local cost data.
Often a certain intervention program has been in existence for several years and has kept rates of infections low. If these infections are now rare and no longer perceived as a problem, administrators might want to cut a program focused on controlling the infections, not realizing that the program is highly effective and even cost saving. The same difficulty arises when trying to initiate a new intervention program because it is easy to quantify the costs of a new program but often difficult to estimate the benefits, particularly when very few clinical trials are available to convince administrators and likely even fewer resources to complete studies at your own institution.
One partial solution to facilitate saving an existing program is to examine areas in which the intervention is not
P.244
in place and compare infection rates in those areas to areas in which the intervention is used. An example would be comparing CVC-BSI rates in a medical ICU where a prevention program exists to those in a surgical ICU that does not have a prevention program. Alternatively, if cost reductions force elimination of a specific program, it would be helpful to stagger the elimination of a program so that as infection rates rise in certain units where an intervention is eliminated, this evidence could be used to re-institute the program.
When an identified problem, new mandate, or new technology leads to the desire to introduce a new IC intervention, it is important to remember that this is the time to collect outcome, cost, and implementation data that will justify this intervention in the future if it faces elimination when the institutional will dissipates. To that end, it often is helpful from an analysis and more importantly from an implementation perspective to roll out a new intervention in a stepwise fashion. This allows comparisons to control populations (wards or ICUs where the intervention has not yet been implemented) using a higher-level quasiexperimental design [52].
Business Case Analysis Example
The steps of a business case analysis parallel those of the cost-effectiveness analysis described earlier. The initial step is framing the problem and developing a hypothesis regarding the potential solutions. For example, you may wish to implement an intervention to reduce SSIs in your hospital. To implement an intervention to reduce these infections, it might be necessary to hire additional staff for your IC department. Thus, you are faced with the task of convincing your hospital administration that the cost of an additional full-time employee (FTE) will be offset by the cost savings through reduced infections, including SSIs.
The next step is to determine the annual cost of the program, in this instance, the salary of an FTE including benefits. This is available from many sources, including your own institutional budgets or available on-line surveys [71]. For example, an FTE (ICP) might cost $75,000 annually.
You must now determine what costs can be avoided through reduced infections to determine whether the up-front cost of hiring a new ICP can be recouped during a reasonable period of time, usually the current fiscal year. Ideally, you might have data from your own institution that can be analyzed to determine whether SSIs decreased after hiring an ICP. Alternatively, the medical literature must be reviewed to see whether others have published data regarding a similar issue. For example, if 4,000 operations are completed at your institution annually and the current SSI rate is 2%, then 80 SSIs occur per year. If your experience or a literature review suggests that hiring an additional ICP would be expected to reduce SSIs by 25% through additional SSI surveillance including postdischarge surveillance, increased reporting of rates to surgeons, and improved timing of perioperative antibiotics then an effective ICP would directly prevent 20 SSIs.
After estimating the number of SSIs that could be prevented, the next step is to determine the costs associated with an SSI from your hospital's perspective. If hospital administrative data are readily available, an attributable cost of an SSI could be calculated as described earlier. Alternatively, a literature review might reveal that the average SSI costs $25,000 [1]. At this point, it might be tempting to multiply the expected SSIs prevented, 20, by the estimated costs per SSI and state that hiring an ICP will save $500,000 in SSI costs alone. However, a certain percent of these costs is currently reimbursed by third-party payers. Perhaps at your institution, 75% of costs are reimbursed, so the cost savings from preventing 20 SSI would fall to $125,000. A recent study found that profits on surgical patients fell from $3,288 when there were no complications to $755 when complications occurred [23]. Thus, the hospital still made money when complications occurred but lost a potential profit of approximately $2,500 per complication.
Completing the business case requires taking the estimated cost savings or additional profits and subtracting the costs of the up-front outlay, in this example, the salary and benefits of an ICP. In this example, the total gain to the hospital is estimated to be $50,000. Many IC interventions have multiple benefits. For instance, hand hygiene education that is increased in response to an Acinetobacter baumannii outbreak also would be expected to reduce MRSA and VRE infections [72]. To further make the business case for an additional ICP, one could include reduced infections and costs associated with other types of potentially preventable infections that an additional ICP could impact, such as ventilator-associated pneumonia.
Even though a business case analysis does not include the adverse consequences of HAIs, such as patient mortality, hospital administrators do respond to those issues. While patient safety cannot be the whole argument, some calculation of the patient safety improvement associated with the intervention should be included. If mortality associated with an SSI is 5%, preventing 20 SSIs is estimated to prevent one deaths. Additionally, preventing complications, such as SSIs, might be associated with reduced legal costs. These must be included in a proper business case and can influence hospital administration. Thus, a hospitals risk management group should be involved early in any quality improvement program economic analysis.
References
P.245
P.246