A Benefit–Cost Analysis of a Worksite Nurse Practitioner Program: First Impressions

Objective:This study aimed to assess the initial impact of an on-site nurse practitioner (NP) initiative on the healthcare costs (HCC) among 4284 employees and their dependents. Methods:The authors analyzed HCC by two methods. First, they compared annualized actual values for the first 6 months of the startup year (2004) with those projected for 2004 on the basis of claims paid in 2002 and 2003. Both aggregate and per-individual HCC were used as the basis for comparison. The difference in HCC between projected and observed values for 2004 was defined as the benefit of the NP program. In a second analysis, HCC were calculated using 2003 paid claims for major diagnostic categories (MDC). These HCC were compared with those that would have been incurred had off-site care been used for the (annualized) number of such patients cared for by the NP in 2004 with the same MDC. The cost of the NP program was used as the denominator in calculating the benefit-to-cost ratio using the savings in HCC estimated by the two previously mentioned methods. Results:Annualized cost of the NP program was $82,716. Savings in HCC using the first method were $1,313,756 per year, yielding a benefit-to-cost ratio of 15 to 1. Using the MDC analysis, the ratio was 2.4 to 1. This difference in ratios between the two estimates may partly be attributable to effects of other initiatives such as the wellness program and the Nurse Health Line. The latter was begun 10 weeks before the NP program, is available at all times, and is intended to minimize the need for workers and families to seek high-cost care at hospital emergency departments. Conclusions:The first 6 months of a new NP initiative yielded substantial reductions in HCC that warrant further analysis over longer periods of observation. However, the initial estimates may understate the aggregate value of the program because it may also reduce on-site injury and illness patterns and improve productivity, end points that were not assessed in this initial snapshot.

W ith the advent of higher medical care costs and productivity concerns increasing at more worksites in the past decade, a growing number of companies have developed in-house medical clinics with occupational health nurses, nurse practitioners, and other allied health professionals. These clinics exist in both midsized and large organizations. According to some of these companies, on-site medical clinics make sense to them in terms of both cost-containment and quality of health care provided to their employees. 1 For example, an external audit of one company's onsite clinics found that employees used fewer outpatient and inpatient services when adjusted by age, gender, and other demographics than their peers in the community. Moreover, the company's database showed the costs to run the clinic are significantly less than what the health plan would have been paying if employees were to have used community healthcare providers. 2 Similarly, other organizations report their healthcare costs have risen just 6% annually over the past 4 years, translating into corporate healthcare spending that is 17% less than industry-related averages; another organization reported its on-site clinics save the company more than $1 million per year. 3 In early 2004, officials of a southeastern U.S. industrial metal/plastics manufacturing firm and Carolinas Health care System (CHS) engaged in a contract to offer nurse practitioner (NP) services at the firm's 1) administrative, 2) plastics, and 3) foundry locations. On June 17, 2004, an NP was hired by CHS and placed at the firm to begin her official duties. The NP's official workload was distributed as follows: • 18 hours per week devoted to plastics division employees; • 18 hours per week devoted to foundry division employees; and • 4 hours per week devoted to administrative division employees.
The primary goal of the NP is to provide primary care services to employees in a timely and cost-efficient manner. The scope of her services includes writing prescriptions, administering medications, and establishing treatment modalities for chronic conditions. Currently, only employees are eligible to receive onsite medical care services from the NP at no out-of-pocket expense.
Before the NP placement, employees and dependents with health plan coverage obtained all medical care from off-site medical care providers. However, since April 1, 2004, a 24/7 nurse advice line provided by CHS has been available to employees and dependents. The nurse advice line is primarily designed to 1) assist employees in selecting appropriate onsite and off-site medical care and 2) eliminate unnecessary emergency department visits.
To determine the impact on the new NP placement on the firm's medical care expenses, CHS retained the services of Chenoweth & Associates, Inc. (C&A).

Materials and Methods
Based on the primary purpose of the NP initiative and the type of preand postinitiative data available to C&A, a benefit-cost analysis framework was used to evaluate the initial impact of the NP intervention. The primary purpose of benefit-cost analysis (BCA) is to determine whether a program is worth its cost. 4 In essence, BCA is an economic tool with an emphasis on estimating the monetary value of everything. The monetary value of a project rests on two fundamental postulates: Postulate 1: The social value of an intervention is the sum of the values of the project to the individual members of society (workplace). Postulate 2: The value of an intervention to an individual is equal to his (fully informed) willingness to pay for the intervention.
Making value judgments about the desirability of economic states is the thrust of welfare economics and the choice of a decision criterion is critical. A guiding rule in formulating criteria, at least in Western society, is that each individual's preferences must (somehow) count in the evaluation of alternative economic states. Although there are four popular decision criteria (unanimity, pareto superiority, majority rule, potential pareto superiority), the criterion used in BCA is the potential pareto superiority criterion. It states that an increase in general welfare occurs if those that are made better off and still from some change could, in principle, fully compensate those that are made worse off and still achieve in welfare improvement. Although this criterion provides the basis for the quantitative part of BCA, it poses problems in that potential compensation may not be actual compensation.
Obviously, BCA is most appropriate when both benefits and costs can be tangibly measured in monetary terms. Nevertheless, some researchers warn that quantification should not be the sole basis for performing a BCA. They contend that just because some important factors are not easily measured, they should not be ignored or given a lesser value than factors that can be measured. For example, how can you accurately quantify the human pain and suffering by people with severe back pain or chronic depression? In essence, BCA does not pretend to introduce rigor and quantification when data originate on subjectivity, imprecision, or where quantification is not feasible. However, when costs and benefits can be quantified, a BCA can be used to judge the worth of a single intervention or provide comparisons on two or more interventions.
Overall, BCA provides meaningful data to the extent that any benefits can be accurately measured. Yet, such noble benefits as human lives saved, preventing heart attacks, or easing chronic back pain are not easily translated into precise numbers. Interestingly, a human life was valued to be worth a mere $5000 nearly a century ago. Moreover, should a monetary value even be placed on a human life? Although it is possible to calculate the direct costs of treating a heart attack victim or to discount a person's future job earnings lost from a disability, try to imagine the technical and ethical implications of using a BCA beyond its intended scope.

Performing a Benefit-Cost Analysis
The cost side of a BCA involves calculating the costs of all resources used in planning and implementing an intervention. In contrast, the benefit side of the equation involves calculating the monetary value of any positive outcomes (preplacement medical costs vs postplacement medical costs) that can be quantified. Measuring benefits requires a number of different techniques. The effects of direct benefits are usually measurable using standard accounting reports and conventional financial analysis. However, the effects of indirect benefits on profitability can be very large, although difficult to prove using conventional cost-accounting.
Obviously, to prepare a workable BCA framework, it is essential to identify and measure benefits and costs. A sample listing of typical benefits and costs is as follows: Typical benefit items versus Typical cost items Fewer injuries/accidents vs personnel Fewer healthcare expenses vs facilities Fewer sick leave absences vs screening procedures Greater productivity vs equipment Presumably, calculating "direct" benefits associated with an intervention should be relatively simple. However, before any benefit can be calculated, evaluators must select benefit variables that are accessible and measurable, and feel confident that any benefit outcome is primarily the result of the intervention. After all costs and benefits have been identified and measured, the two categories are monetarily compared. In most cases, the goal is to determine the "net benefit" of a particular intervention. In essence, if the value of the benefits minus the value of the costs is positive, then the analysis would indicate that the intervention is financially "worth the effort." The net benefit of any intervention may be viewed in the following manner: where: ͚L$ (sometimes called the direct benefit) stands for the reduction in medical care expenses as a result of a reduction of disease or disability. For example, if the incidence of low back injury declines, then some of the spending on physicians and other medical care services will no longer be necessary and, thus, "saved" by the employer, society, or other payers.
͚GP stands for the increase in general productivity, leading to greater output and income. For example, if we reduce the incidence of low back injury, we also increase the performance capabilities of the persons involved (eg, uninjured workers are more productive than injured workers) so that they may continue to produce at desirable levels and earn their full paycheck.
͚PI stands for the gain in working income as a result of reduced illness and injury and their effects on absenteeism (lost income). Overall, GP and PI are the "indirect benefits." C stands for the cost of the intervention (eg, a nurse practitioner).
Using a BCA framework, the cost of the NP placement was compared with any reduction in the firm's actual 2004 medical costs versus the firm's projected 2004 medical care costs.

Results
The official date in which NP services were initiated at the firm was June 17, 2004. Thus, all direct costs associated with the NP placement were tracked from June 17, 2004, until December 31, 2004. Total direct costs for this timeframe were converted into a period of 12 months to provide an estimated annual cost. By factoring in all tangible costs associated with the NP placement and services to employees, a direct cost of $82,715.81 was computed on an annual basis.
Initially, it was necessary to establish an accurate baseline of the firm's costs in the pre-NP placement years (2002 and 2003). Essentially, the retrospective claims payments in these years gave evaluators a relative baseline in which to compute projected 2004 payments without the NP placement intervention. Thus, medical care claim charges and payment data for 2002 and 2003 were provided by the firm's benefits department who previously obtained these data from United HealthCare (the firm's health plan administrator in 2004). A review of these preplacement data revealed yearly comparisons as shown in Table 1.
Overall, Table 1 shows substantial increases in both utilization and cost patterns despite virtually no change in the number of covered lives. Although it is important to acknowledge total (gross) claims and charges, evaluators chose to provide per capita distributions because they reflect measurable outcomes at the individual level. By and large, there was very little difference between gross percentage changes versus per capita percentage differences from 2002 to 2003.
Once the preceding pre-NP placement utilization and cost patterns were quantified, evaluators devel-  Table 2.
The 2004 projections provide a glimpse of the firm's claims and payments without an NP. The next step was to compare the projected 2004 profile (without an NP) against the actual 2004 profile (with an NP in place). Table 3 provides a headto-head comparison between the two scenarios.
Overall, the preceding data show projected 2004 outcomes (without an NP) were considerably higher than actual 2004 outcomes in all major categories. Percentage differences between the respective 2004 profiles are as follows: • Actual payments were 18.5 lower than the projected norm; • Actual per capita claims were more than 24% lower than the projected utilization norm; and • Actual per capita payments were more than 14% lower than the projected payment norm.

Projected versus Actual Cost Differences
Overall, a cost difference of $1,313,756 exists between the projected 2004 medical payments (without the NP placement) versus actual 2004 payments (with an NP). Although it is possible to speculate that the preceding cost difference may be twice the listed amount-because the NP was in place only 6 months-it is impossible to determine an actual an-nualized cost difference until an evaluation is conducted after the NP has been in place for a full 12 months. Yet, this cost difference only reflects a direct cost difference tied exclusively to medical care payments incurred in two different scenarios. Interestingly, the cost difference does not factor in other "hidden costs" such as lost productivity if employees would have to leave work to seek healthcare services in off-site (community) healthcare settings. Nor does the cost difference factor in the number of lost time absences avoided by employees when they (would presumably) come to work because they have on-site healthcare services instead of staying home. These "hidden costs" have been studied by numerous organizations over the past decade. For example, one organization surveyed employees using the on-site medical clinic regarding: 1) whether they would have used an outside doctor if the corporate medical clinic were not available, and 2) how many days per year they estimated that they came to work because there was an on-site medical clinic when they would otherwise have stayed at home. The results indicated that, on average, employees who used the facility saved 3.3 days of absenteeism. In addition, nearly 70% of employees indicated that they would have sought attention elsewhere, suggesting that the presence of an on-site medical center does not induce demand. 5

Benefits versus Costs
Overall, there was a difference of $1,313,756 between the firm's projected 2004 payments versus the firm's actual 2004 payments. This "cost avoidance" represents a tangible and measurable "benefit" that can be subjected to the following benefit-cost comparison: The preceding benefit-to-cost ratio indicates the initial phase of the NP intervention generated a favorable level of cost-containment benefits at the firm. Yet, it should be noted that the 2004 claims and payment data were based on 11 months of actual data and, thus, did not include claims and payments for the month of December. Although C&A extrapolated 11 months of data into an annual (full year) projection, there is some evidence that claims and payments are higher in the winter months than in other months. 6   determine the most accurate benefit-to-cost ratio-within this initial methodological approach-would be to obtain all claims and payments relevant to the month of December. Statistically speaking, the preceding benefit-to-cost ratio of 15.88 to 1 would, at worst, be reduced no more than 1/12th based on the liberal assumption that utilization and payment patterns are twice as high in December as in preceding months. Even then, the ratio would be reduced to 14.55 to 1.
Like any project in which econometric analyses are involved, it is important to compare the initial costto-benefit findings with those from another evaluative approach to objectively judge how the independent approaches differ and what, if any, contrasting outcomes result. This project lends itself to such options that can provide decision-makers with at least two alternative methods to evaluate the impact of the NP intervention.

A Second Analysis
Unlike the preceding approach, which is a macroanalytical overview and comparison of projected versus actual claims and costs conducted on an aggregate (group) population, this second approach was a bit more microanalytical in that it takes into account: • Actual conditions reported by individual employees within each of the three worksite locations; • Estimated costs for the preceding conditions if they were treated off-site (without NP treatment); and • A cost comparison between treatments per specific condition without NP versus with the NP.
Key procedures used to factor in the preceding components were undertaken in the following order: was compared with the estimated cost of 2004 on-site treatments (with an NP) to determine a "costsavings benefit"; and 7. The "cost-savings benefit" (no. 6 preceding) was compared against projected 2004 medical costs (without an NP).
In performing the preceding analyses, it was essential for evaluators to identify discrepancies regarding the classification of pre-NP (2003) claims data versus the classification of conditions treated by the NP. For example, 2003 claims data provided by United Health-Care were by MDC. Yet, treatments rendered by the NP were formatted by International Classification of Disease (ICD). As a result of this discrepancy, evaluators had to assign the ICD-based treatments to their respective MDC (eg, ICD 401 was assigned to the MDC known as circulatory).
Because virtually all of the conditions treated by the NP appeared to be acute and minor ailments, evaluators decided to limit their review of 2003 claims to conditions costing no more than $500. Thus, by selecting relatively low cost claims (Ͻ$500), evaluators believed the average costs assigned to each MDC would not be artificially inflated by a small percentage of excessively high cost claims ("outliers"). In addition, only the most common MDCs were included in this analysis. A review of claims data indicated that 12 different MDCs comprised over 90% of all conditions treated by the NP during the study period (June 17, 2004, to November 12, 2004. Table 4 shows a listing of the average payment and most common MDC conditions treated by the NP during this timeframe in addition to estimated payments that the firm would have borne in 2004 without the NP. Overall, specific MDCs responsible for the greatest cost-savings potential were circulatory (25%), respiratory (17%), and skin and subcutaneous (13%).
The final computation conducted in this alternative analysis was a head-to-head dollar comparison of the annual cost for the NP compared with the estimated annual cost savings. Comparing direct costs tied to an annual placement of a NP at the firm with the estimated "cost savings" yields the following: Overall, the preceding microanalytical review of costs and benefits indicates the firm is saving $1 for every 42 cents spent for on-site NP services-or a savings-to-cost ratio of 2.38 to 1.
In summary, this microanalytical review of the NP placement indicates a more conservative, yet favorable, cost-to-benefit ratio than the initial macroanalytical approach.
Overall, the financial comparisons described and illustrated on the preceding pages reveal the following benefit-to-cost ratios and return-oninvestment percentages:

Discussion
The quest for control of healthcare costs and quality is leading some employers to establish on-site clinics and other initiatives that employ health professionals. A recent editorial in this journal 8 estimated that lost workdays for chronic medical conditions cost U.S. employers $73 billion per year in addition to costs of lost productivity as well as the outof-pocket medical expenses that they bear. Among strategies to cope with these increasing costs, a growing number of companies have introduced in-house medical clinics staffed by full-time nurse practitioners, occupational health nurses, physicians, and other health professionals. However, there is little published information on the balance between the cost of such initiatives and their benefits. Our analysis focused on the initial benefit-to-cost experience of a pilot program in which a NP was recruited to provide on-site care for minor illnesses, as well as improve management of chronic medical conditions of 1200 employees and 3084 dependents whose health insurance is subsidized by a three-site plastic pipe and foundry company. The new program shows promise for yielding positive monetary results. The size of the initial cost saving is sensitive to the method of analysis used. That is, the annualized savings in HCC was $1,313,756 using the projected HCC Our report has obvious weaknesses, the main one being the use of only 6 months of experience. Hence, further experience with the NP program is needed to refine this analysis and judge overall program impact. At this stage, our methodology did not enable us to apportion benefits from other interventions such as the advent of the CHS Nurse Health Line (NHL) and the company's Wellness Program. The latter program includes health risk appraisals with monetary inducements for participation, health fairs, on-site determination of serum lipoproteins and other analytes with reports to be given to personal physicians, and subsidized mammograms. However, we note that this program was already underway before the addition of the NP. It is possible that a future report will include an analysis of visits to hospital emergency departments and urgent care centers, enabling comparison of use patterns before the NHL (April 1, 2004) and NP (June 17, 2004) initiatives. Essentially, this expanded evaluation would permit us to see month-tomonth, 12-month comparisons both to projections and to previous year actual costs. It is also possible, however, that these initiatives may expe-  The firm's estimated cost savings if treatments were rendered by an on-site nurse practitioner rather than off-site healthcare providers.
MDC indicates major diagnostic category.
rience some adverse outcomes with negative monetary or other consequences that did not occur during this startup phase. Should such occur, the reactions on the part of attending physicians of the involved workers or family members could be quite negative. However, we endeavor to consult with personal physicians in all cases in which other than minor illnesses occur and to refer worker-patients promptly in such cases. Often, the workers do not have personal physicians, in which cases we identify such resources convenient to their homes and arrange for early appointments. The MDC analysis we used also carries certain limitations. For example, the $500 cutoff for claim costs may give an obvious disadvantage to the NP. The real savings of an on-site NP program may be in preventing claims that have escalated unnecessarily either by early treatment and prevention or by preventing trips to a hospital emergency room. Furthermore, these small claims represented a small percentage of total costs, thus raising doubt if an on-site NP program would, in actuality and over time, be limited to only impacting a small subset of total claims and costs.
As an additional limitation, our experience with the NP and NHL programs may lack general applicability, being based in three urban North Carolina locations of a single company engaged in the foundry and plastic business, and which had onsite occupational health nurses at both of its manufacturing sites. Many of the employees of the company live at substantial distances from the worksites, driving an hour or more each way. At times, long working hours make it difficult for individual workers to access health care offsite. Our findings may therefore not apply to other groups of workers with different logistics and personal habits. Other employers will need to make careful assessments of the benefits and possible disadvantages of on-site programs similar to the ones described here.
From the perspective of healthcare organizations, it may seem counterintuitive for such an entity to help design and implement lower-cost options for delivering such care as well as preventive medicine measures. However, it strikes us as essential that those entrusted with the care of the community must support efforts to improve such care and to optimize its benefit-to-cost function, and to do so in measurable ways. CHS therefore plans to continue its efforts to partner with employers in this regard, including optimizing the deployment of NP, NHL, and other health professionals at all sites where they practice their expertise.