Two months back (MLO. 2016;48(5):30-31), this series looked at some of the underlying complexities and examples of extant studies in assessing the cost-effectiveness of molecular diagnostics for oncology-related applications. In this month’s installment of The Primer, we’re going to expand this question to another common use of molecular diagnostics:
infectious diseases.
“Cost-effectiveness”: not as simple as it seems
At first glance, infectious diseases (ID) might seem to be a fairly straightforward field for this type of cost-benefit analysis. After all, if one imagines a scenario of a large clinical laboratory currently running a sample stream of statistically valuable size through a reference method, it must be possible to account direct assay costs per test for the reference method; and then apply a novel method such as an MDx approach to the same specimen stream; and calculate the same direct costs. Indeed, many papers which report on validation or pilot testing of new MDx methods present this sort of data. By accounting for the costs of the actual devices, reagents, and resources such as staff time and ancillary reagents used per generation of an accurate result, we can indeed generate one valid metric of cost-effectiveness. In this sort of comparison, if test A is 15 percent more costly than Test B but has a 20 percent higher accuracy rate, it’s judged as more “cost-effective.”
Note, however, the use of the term direct costs above. There are, of course, also indirect costs associated with infectious disease (such as costs of isolation, and costs related to secondary transmission cases arising from an undetected/untreated positive). Calculation of these has to rely more on models than on hard data. These models utilize both location invariant factors (such as false positive and negative rates for the tests in question) and location-specific factors (isolation costs, treatment costs). Larger societal costs such as lost productivity can also be incorporated, with associated differentials between two testing regimes. As with the oncology-related health economics we considered in the May issue, in-depth considerations of ID diagnostic economics employ concepts such as QALY (quality adjusted life years), ICER (incremental cost-effectiveness ratio), and Willingness to Pay thresholds. An explanation of these and their application is beyond the scope of this article; what is important to grasp from this is some idea of the underlying complexity of these analyses.
Two approaches to the problem
In spite of this, it’s a matter of necessity to have some sort of framework or method with which to make defensible assertions about the cost-effectiveness of various assay options when considering whether to adopt them to augment or replace extant methods. Without having the services of a healthcare economist to perform these analyses, what options does the reader have to do this?
Two approaches seem particularly feasible. The first is to just employ a direct costs-per-accurate-result metric as discussed above; while it is challenging, the estimation of this at a given locale is doable. If this is done and shows a very strong bias of one test modality being more cost-effective than others, we can perhaps accept this result with some confidence in the intuitive impression that unconsidered complicating factors, while influencing the exact values, are unlikely to change the net overall method preference.
Conversely, if direct cost-analysis results show only small differences in effectiveness between methods, then unconsidered factors are more likely to be of a deciding nature. The difficulty in this approach is that we have no good guidelines for what would constitute “small” versus “large” differences; significant differences of opinion are likely to exist. A subset of cases will exist where a proposed MDx assay is both demonstrably more accurate than traditional methods and less costly; in these cases, consideration of indirect costs is not strictly necessary, as those costs can only add to the net improved cost-benefit answer, not change its bias direction. However, most of the time we cannot both improve accuracy and reduce direct costs.
A second approach is to extrapolate from published studies based on more comprehensive analyses. The more points of similarity (jurisdiction, supplies and labor costs, identity of tests being compared) between a published study and the location of interest, the more applicable this extrapolation can be. Observation of which particular indirect costs were of largest magnitude in the study can also be informative. If only one or a few dominate the cost-benefit result, then these are likely the most critical indirect cost factors in similar settings. Finally, those published analyses which are most clear in describing their models employed may be most readily adapted to suit a new venue.
Review of literature
With this approach in mind, what sorts of cost-effectiveness studies regarding the use of MDx for ID can be found? While MDx methods in ID are now commonplace, a review of recent publications specifically addressing their cost-effectiveness turned up surprisingly few publications with hard numbers. This is probably due to the complexities enumerated above. A second observation: many of the publications to be found in a literature review relate to only a few newer MDx tests—in particular, ones with relatively high direct costs. On reflection, this is not very surprising; certainly, tests with high direct costs are the ones which most need in-depth financial justification. (This also is an affirmation of our intuitive conclusion above; if we assume that MDx tests in general have high accuracy rates, then it will be the ones which also have the highest direct costs which provide the smallest cost-per-accuracy benefit, and thus the ones where more in-depth analysis is essential to provide accurate assessment of whether this provides a cost benefit).
Instructive examples of some of these publications include the following, with annotated commentary and in no particular order:
- Goldenberg SD, Bacelar M, Brazier P, Bisnauthsing K, Edgeworth JD. A cost benefit analysis of the Luminex xTAG Gastrointestinal Pathogen Panel for detection of infectious gastroenteritis in hospitalised patients. Journal of Infection. 2015;70(5):504-511.Large MDx panel assays—useful for the single pass identification of an infectious disease etiology when there may be multiple pathogens likely to give a single clinical presentation—are a well-represented class of MDx with cost-benefit publications. Based on an eight-month, ~800-patient parallel specimen stream study approach in the UK, this publication provides a particularly good example of where an MDx assay appears to be cost-benefit prohibitive on direct costs alone; it reports for 800 patients, use of the MDx approach costs £22,283 more than alternate testing (or approximately $32,000). On further analysis, though, application of the MDx approach cut aggregate patient isolation days from 2,202 to 1,447, or a calculated cost savings of £66,765 ($96,000). Clearly, when indirect costs are figured in, the use of MDx not only improves healthcare outcomes by faster accurate diagnoses and less isolation, but also is more economical to the health system.
Examined in light of our thoughts above, what extrapolations could be gleaned from this paper for another site looking to introduce some form of GI MDx for ID? First, isolation costs were a significant indirect factor and should be accounted for in some way. (This is a generalizable observation for ID economics; since isolation is very costly, it’s frequently a major indirect factor for any disease where it is part of the infection control or treatment strategy). Second, this paper illustrates another point raised above: while the monetary cost values are persuasive, for extrapolation purposes it’s even more useful to have the days of isolation highlighted, as this allows for correction to localized isolation costs.
- Schroeder LF, Robilotti E, Peterson LR, Banaei N, Dowdy DW. Economic evaluation of laboratory testing strategies for hospital-associated Clostridium difficile infection. Journal of Clinical Microbiology. 2014;52(2):489-496.
This study refers to a GI setting like our prior example, but it is based around a simplex assay as opposed to a multiplex panel. This publication involves highly detailed models with results indicating that stand-alone simplex PCR testing for the pathogen was preferable to traditional or mixed diagnostic approaches, for situations where system costs of a missed positive (false negative) exceeded $6,900. The authors then proceed to apply Monte Carlo simulations to estimate this system cost, with a median result of $14,000; stand-alone PCR thus appears to be the most economical approach in an “average” setting.
Extrapolations from this paper are fairly straightforward. As the authors provide the estimated sensitivity and specificity of their C. difficile PCR, substitution of those values with ones applicable to an assay under consideration allows reasonably direct comparisons of outcomes.
- Millman AJ, Dowdy DW, Miller CR, et al. Rapid molecular testing for TB to guide respiratory isolation in the U.S.: a cost-benefit analysis. PLoS One. 2013;20(8):e79669.
This study, focusing on the application of a single-target sample-to-answer MDx for TB detection compared with traditional AFB smear microscopy, provides an estimated cost benefit of use of MDx of $533,520 per year for a hypothetical laboratory cohort. Extrapolations from this paper are aided by a particularly clear presentation of the epidemiological and diagnostic costs and factors used in the models. For a lab considering this or a similar test, review of those values to see how closely they agree with local values will help in understanding how applicable the values from this study are in another context. Readers will also find the statement “The majority of cost savings arose from reductions of length of stay in respiratory isolation.” As mentioned above, this is a reliably recurring theme.
So in conclusion…
Space constraints preclude us from considering any more examples in this article. However, it is hoped that the reader will have gained an appreciation of the complexities involved in accurate health economic analyses even in the “simple” context of ID. Direct costs analysis plus at least some estimation of any isolation costs may be a starting point, while those looking to employ an assay which has already had a detailed published economic analysis may find an extrapolation approach fruitful. The simplest cases are those that can readily demonstrate both lower direct costs and improved results accuracy by switching to a new method, as they can safely dispense with indirect considerations.
John Brunstein, PhD, is a member of the MLO Editorial Advisory Board. He serves as President and Chief Science Officer for British Columbia-based PathoID, Inc., which provides consulting for development and validation of molecular assays.
