Cost Utility Analysis of Radiographic Screening for an Orbital Foreign Body before MR Imaging

A single case report early in the clinical application of MR imaging has led to great controversy regarding the need to screen patients before MR imaging. The patient was reported to have sustained an ocular injury from a retained ferromagnetic foreign body. No other adverse reactions to MR imaging due to an orbital foreign body have been documented in the literature. Despite case reports of human subjects unharmed by MR imaging in the presence of metallic orbital foreign bodies, recent investigators have concluded that “a history of occupational exposure to potential metallic ocular injury” necessitates radiographic orbital screening before MR imaging (1–6).

An increasing concern in health care delivery at present is cost. The cost of pre-imaging screening for a metallic foreign body may represent a significant portion of the overall cost of an MR imaging program at rates for orbital screening reported in the literature. The costs of overscreening can be clinical as well as financial; for example, MR resources in the community are used less effectively as a result of inefficient scheduling, canceled examinations, and delays, denying the benefit of MR imaging to some individuals who might require it. Similarly, the costs of underscreening might be financial as well as clinical; for instance, if an individual becomes partially disabled as a result of an ocular injury (7–11).

Cost-effectiveness analysis is a method whereby financial and clinical risks can be quantified and compared. By using explicit assumptions, the magnitude and direction of risks and benefits can be discussed rationally, and choices regarding radiographic screening for an orbital foreign body before MR imaging can be compared with other, less emotion-laden choices made daily in our society. Critical variables in the analysis can be identified when the results of the analysis change materially over a plausible range of the variable in question (12–24).

Methods

Results

Using base case values for the variables in our analysis, we estimated a cost per quality-adjusted life-year (QALY) of $328,580. In other words, if the cost of radiographic screening is $173, if the probability of a metallic orbital foreign body being ferromagnetic is .5, if the probability of its being of sufficient size and precise position so as to induce complete blindness in the affected eye is .25, if the sensitivity of radiographic screening for this type of foreign body is 90%, if an ophthalmologic examination would miss 10% of ocular foreign bodies, if the patient has a life expectancy of 30 years, and if the value of a dollar today is identical to the value of a dollar 30 years hence, then we must be willing to spend $328,580 per year of life adjusted to reflect the value of perfect health in order to justify current screening recommendations. Put another way, we must be willing to pay $2,464,350 today to avoid a blind eye event.

The analysis is sensitive to screening cost. The results in Table 3 show that a screening cost of about $25 would result in cost per QALY below $50,000. When a 5% discount rate is applied, this is no longer the case. Sensitivity analysis shows that the expected life span of the patient impacts the analysis significantly; the shorter the life span, the greater the cost of screening (Table 4). However, even with a life expectancy of 40 years, the cost per QALY is very high. As time preference for money increases, the radiographic screening program becomes less cost-effective. The prevalence of an ocular foreign body among the screened population is also an important variable (Table 5).

The efficacy of radiographic screening is not a critical variable (Table 6), and the probability of injury is most likely not a critical variable (Table 7). At an extremely radical assumption of 100% injury rate, the cost per QALY is over $70,000. The probability of a false-negative ophthalmologic examination is probably not critical (Table 8). Severity of injury does not appear to be an important variable (Table 9), nor is the probability of having a ferromagnetic foreign body in the eye (Table 10). The incremental cost of radiographic screening is not sensitive to medical or rehabilitation expenses (Table 11).

Discussion

Physicians operate under uncertainty every day. In our society, the frequency and severity of rare events with dramatic adverse outcomes are systematically overestimated and overemphasized. The social and clinical consequences of these errors have been explored, and means to avoid them have been discussed. Consider that the risk of reaction to MR contrast agents are at least an order of magnitude lower than those associated with iodinated contrast material. Despite this, a recent article states that “the indexes of suspicion …must be as rigorous for reactions associated with MR contrast agents as they are for reactions associated with iodinated contrast material.” (47) It is in this milieu that some current screening recommendations have developed (8, 47–57).

Methodological complexities abound in an analysis of this sort. Fortunately, the problem at hand is relatively straightforward. Only one event needs to be considered, the MR imaging examination, and all the costs, except lifelong monocular blindness, are incurred immediately. The incremental costs of clinical screening are de minimus, only one or two questions need to be asked, and this incremental cost is less than the incremental costs of arranging the radiographic screening examination, losing MR imaging availability, substituting less efficacious or more invasive diagnostic procedures, and rescheduling the MR imaging examination.

Decision tree analysis has been suggested as the ideal method for analyzing cost utility problems in medicine. This may be true for situations in which complex sequential chains of events are analyzed, in which succeeding decisions depend on those that have come before. However, here we have a single, discrete event, so a decision tree is not appropriate. Furthermore, the complexity of decision tree depictions may mask analytic deficiencies and distract even educated observers (26, 48).

Our base case estimates for the frequency and magnitude of risks associated with performing MR imaging in patients possibly harboring metallic foreign bodies were constructed in such a manner as to favor the hypothesis that radiographic screening is justified. For example, we chose a probability rate of possible injury associated with MR imaging in a patient with an orbital foreign body as 0.25. This rate was based on the two cases reported by Williamson et al in which patients had MR imaging and subsequent foreign body injury revealed at radiography, on another similar recent case in which no injury occurred, as well as on one case reported by Kelly et al in which an injury did occur. The rate of probability is most likely a great deal lower, since the sensitivity of orbital screening for detecting radiopaque foreign bodies is between 69% and 90%, and because patients' memories of remote occupational exposure must be faulty in many cases. Furthermore, in one survey, 5% of the MR imaging facilities polled had no orbital screening protocol. There has almost certainly been MR exposure in thousands of patients harboring metallic foreign bodies in whom no injuries resulted. The probability of injury is likely to be one per several thousand. Using a different analytical approach, Williamson et al estimated one in 2000 (1–4, 39, 42–44, 46, 58).

We assumed a base case discount rate of zero, which clearly favors the radiographic screening project. Since the costs are borne in year zero, when MR imaging is performed, and the costs of monocular blindness are borne in the future, this selection strongly biases our conclusion in favor of radiographic screening. The sensitivity analysis here is interesting. For example, recent studies of the medical care cost impacts of cigarette smoking used a variety of discount rates, and the discount rate of zero was abandoned in favor of nonzero rates. A recent comparison of multiple cost-effectiveness analyses normalized results to a 5% discount rate (27).

Furthermore, we ignore the radiation effects on the lens of the eye, on calvarial and spine bone marrow, and on the thyroid gland and brain. These effects would be seen only in a radiographically screened population and not in a clinically screened population. We concluded that the rates of these potential complications would be so low as to be negligible.

We assumed that complete loss of vision would result in the affected eye and that this would be immediate and permanent. We also assumed that all medical care to ameliorate blindness would be futile. Ratings of disability due to monocular blindness were based on complex formulas in which central and peripheral vision losses are taken into account and partial disabilities may be graded from 0% to 100%. Lesser degrees of ocular injury result in smaller fractions of disability (25, 59).

Complete monocular blindness is considered to result in 25% impairment of the visual system and 24% impairment of the whole person; 50% impairment of the worse eye is considered to represent 12% impairment of the visual system and 11% impairment of the whole person; and 25% impairment of the worse eye causes 6% impairment of the visual system and 6% impairment of the whole person (25).

We have assumed that medical care would be provided acutely to the injured individual and annually thereafter. We performed sensitivity analysis over a range of values, and the impact on the cost of radiographic screening was negligible. Our base case value of zero expenditure in medical and rehabilitation costs was based on the standard economic assumption that lost productivity represents the economic loss to society. After all, workers ultimately must, on average, produce the resources expended to care for themselves and others. The net impact on society as a whole must eventually be accounted for as the lost output to society manifest as lost productivity by the injured person. If labor markets are at all efficient, this ought to be accounted for in compensation to workers, on average. As we ran these medical costs to very high levels, the analysis was only minimally changed. This is because many individuals must be screened and only a very few are even potentially injured. Our assumptions with respect to medical care are extremely conservative, because we assume that no beneficial impact ever accrues to the patient. The disability is permanent, nonameliorable, and immediate (45).

Our base case estimates of an age of 45 years and a life expectancy of 30 years were derived from our clinical experience and from life tables. We used male life tables for this analysis because industrial exposure to metallic foreign bodies is vastly more common in men than in women. As social transformation in workforce composition continues, it is possible that these calculations would need to be reassessed, since greater longevity in females would result in longer life expectancy. Potential confounding variables that we ignored in this analysis are the near absence of significant exposure in pediatric populations and differential rates of eye protection use in contemporary versus remote occupational exposure (26).

Cost of screening is a critical value in our analysis. Our assessment of the cost of screening as the dollar charges for the examination therefore bears further scrutiny. We used a figure quoted from a recent paper strongly advocating screening based on occupational history alone (5). If the cost can be reduced to about one seventh of this level, the screening enterprise may be cost-effective. The level of $25 approximates the allowable Medicare fee for a single-view screening examination (46). It is highly probable that the costs of performing the examination are not fully reimbursed at this level of payment (personal communication). The figures for CT are included because a pre-MR imaging CT study had been available in the index case of ocular injury (4), and because the allowable Medicare fee for CT is close to the plain film charge reported (5, 46; and Personal communication, Unanimous declaration of California Managed Imaging Board of Directors, 1999).

The probability of injury is potentially a critical value as well. If 100% of patients with ocular foreign body injuries are blinded completely in the affected eye, the cost per QALY approaches the median household income, or the annual cost of dialysis. However, we know of only one case of injury and at least three documented cases in which persons with an orbital metallic foreign body were examined by MR imaging without injury. Furthermore, patients forget exposure, radiographic examination is not 100% sensitive, and some sites have no screening protocol. The probability of injury must be very much lower than the .25 we used in our base case. The true value for this variable is probably less than .0005 (1, 43).

The prevalence of a foreign body in the screened population is potentially a critical variable. We used the most recent published data in our base case, but if this is increased to 2.5%, the cost per QALY approaches the median income, or annual cost of dialysis (5, 43).

The concept of QALY is somewhat controversial. Some regard the valuation of human life by a continuously variable function as anathema. We adopt a neutral position. However, if there is no difference between a year of life in perfect health and a year of life disabled, then the entire discussion would be moot, and radiographic screening for MR imaging would be abandoned outright. The precise function for the assessment of disability is necessarily complex. It also changes over the course of one's life. For example, disabled patients consistently rate the quality of life with a given degree of disability higher than do those who are not disabled. We adopted the disability scales developed by the AMA and by California Workmen's Compensation, since these are widely used and have been vetted in a wide social milieu. Sensitivity analysis for this variable was not performed, since the discussion of plausible ranges may be theological as well as scientific. We did analyze the sensitivity of our conclusions to partial blindness injuries and found no critical impact (7, 25, 60–67).

The critical dollar value of the cost utility ratio, which changes the conclusion of the analysis, is necessarily arbitrary. We can compare the ratio calculated here with other published figures and draw conclusions based on the relative magnitude of the cost utility ratio. The median cost utility ratio for medical interventions is $19,000 per life-year saved. Another standard commonly used is the annual cost of renal dialysis, which is $35,000 to $45,000 in 1997 dollars (6, 10, 30–32, 67).

Cost-effectiveness analysis has been applied to many medical and nonmedical interventions designed to save lives and preserve health (68). As can readily be seen, the intervention under discussion here, radiographic screening for an orbital foreign body before MR imaging, based on recent recommendations, is very expensive relative to other interventions. Note also that the price of low-osmolality contrast media has declined markedly since the publication of the highest figure shown other than that for pre-MR imaging orbital screening. Therefore, the discrepancy between that figure and the cost of orbital screening is even more pronounced at present costs (Table 12).

Critical variables in cost-effectiveness analysis are those whose values, when varied over a plausible range, alter the cost per QALY sufficiently to alter the conclusion of the analysis. We identified several critical variables. If ophthalmologic evaluation for a metallic foreign body had a false-negative rate near 100%, the conclusion of our analysis would be altered. This seems unlikely, since the toxic effects of iron and steel in the eye are well known to ophthalmologists and the magnetic properties of these foreign objects actually facilitate their removal from the eye. The cost of screening is potentially a critical variable, but the level at which it becomes critical is probably below the cost of providing the service (personal communication). In populations in which the rate of ocular foreign body exposure exceeds that published in the literature, screening radiography may be cost-effective under some circumstances (35, 38, 40; and Personal communication, Unanimous declaration of California Managed Imaging Board of Directors, 1999).

Taking many years of clinical experience into account, we have adopted the following criteria in our practice. Occupational history by itself is not sufficient to mandate radiographic orbital screening. If a patient reports injury from an ocular metallic foreign body that was subsequently removed by a doctor or that resulted in negative findings on an eye examination, we perform MR imaging. To date, we have encountered no adverse clinical impact. Those persons with a history of injury and no subsequent negative eye examination are screened radiographically. We have performed approximately 100,000 MR imaging examinations under this protocol, so the upper bound of the 95% confidence limits for the rate of injury to the eye due to a ferromagnetic foreign body is less than three per 100,000, or 0.003% (69).

Conclusion

Radiographic screening before MR imaging on the basis of occupational exposure alone is not cost-effective. It is also probably not clinically necessary. The critical variables we identified that may affect the validity of this conclusion are cost of screening, effectiveness of ophthalmologic evaluation, probability of injury, and frequency of foreign body invasion of the orbit. To date, our approach has been clinically successful.